(1), Electric Rhino R68S soft routing evaluation
1. Appearance and key configuration analysis
.1 R68S key configurations Electric Rhino R68S uses Rockchip RK-3568 ARM processor, 22nm process, 4-core Cortex A55 architecture, clocked at 2Ghz, supports up to 8G memory;
RK3568 has built-in 1 QSGMII, 2 PCIE 3.0X1 channels, and 2 Gigabit Ethernet interfaces, so it can be used on a router to achieve 2x2.5Gb+2x1000Mb 4 network port combination.
R68S core configuration is shown in the figure below:
core configuration
CPU The two chips next to are memory and flash memory chip . The
LPDDR4 memory chip comes from Rayson, the model is RS1G32LF4D2BDS-53BT, and the capacity is 4GB (the 2GB version is 70 yuan cheaper); the
eMMC flash memory chip comes from FORESEE (Shenzhen Jiangbo Longqi Bottom), the model is NCEMASKG-16G, the capacity is 16GB; there is a Gigabit Ethernet switching chip in the upper left corner of the
processor, the model is RTL8211F. When you see this crab logo, you know it is from Realtek ;
does not have 4 Network port? What about the other 3 switching chips?
is on the back of the motherboard.
The two 2.5Gb switching chips are also from Realtek, the model is RTL8125BG:
The back of the motherboard
Realtek's 8125B, Intel 's I225 B3 and Qualcomm 's QCA8081 are currently the most common .5G switching chip on routers and NAS;
.2 R68S appearance engineering machine was sent with only a bare board.
Because I wanted to test the temperature inside the weak current box , so a few days later the boss sent another shell (for trial molding, not the final version).
The shell I received is made of plastic (there is also a CNC aluminum shell, the price is about 100 yuan), and there are a lot of heat dissipation holes.
The official version of the casing will be made with a frosted effect
The casing and motherboard
Because the openings of the casing are not equipped with dustproof cotton, so if you put the R68S on the desktop, you still need to consider the dust problem. After all, dust is the first concern of electronic equipment. A killer:
Front
There are also a large number of heat dissipation holes on the two sides of the machine:
Side of the machine
If you, like me, place the R68S in a weak current box, then the hollow design of the shell is actually very beneficial to heat dissipation.
But if you use R68S on the desktop, I recommend buying a CNC aluminum casing, which is dustproof and waterproof, and has good heat dissipation. The
CNC shell looks like this:
CNC shell preliminary model
.3 Power supply R68S can be purchased as a set with power supply or as a stand-alone version.
I actually tested it and found that can also work normally with a 12V 1A optical modem power adapter.
Of course, if you need to connect an external hard drive, I suggest you use a 12V 1.5A power adapter.
Speaking of external hard drives, the USB 3.0 performance of R68S is still very strong, and it can read and write up to about 120MB. If you want to connect a hard drive to the R68S to watch movies, you don’t have to worry about the original high-bitrate movie being stuck.
2. Performance and stability test
For an soft router , performance and stability are equally important.
.1 encryption and decryption performance directly look at the table:
Zhihu@仙鱼
CPU CoreMARK
AES...
CHA...
AES hard decoding
R68S (RK3568)
7633
03101
2885
is
N1(S905D)
8704
5377
5712
is
R4S (RK3399)
8275
36089
7899
is
J1900
3985
6550
9185
no
J4125
9931
31907
51102
is
The encryption and decryption performance of R68S is much stronger than the N1 of ARM architecture , but it lags behind the 6-core R4S (RK3399) by about 20%. The CPU CoreMARK of
R68S is about 6000 points behind J1900. However, J1900 does not support hardware AES decoding after all, and R68S still has some advantages in this regard. If
is compared with J4125, which also supports AES hard decoding, R68S lags behind by more than half.
Although the R68S is not excellent in terms of scores, in actual use, the performance is more than sufficient.
.2 NAT performance NAT performance mainly tests the broadband speed. The tested topology is as follows:
NAT test
The test rate is as follows:
40/100
It can be seen that if the transmission loss is removed, the R68S has reached the limit of Gigabit broadband.
At this time, the CPU load is around 45%.
.3 LAN forwarding performance LAN forwarding capability mainly tests the forwarding rate between the two 2.5G LAN ports of the R68S. I did a total of 3 different tests. The topologies of
different tests are as follows:
tests
- .3.1 iPerf3 test
R68S and iPerf3 client installed on the computer respectively, ThinkPAD X13 (2.5G USB 3.0 network card) is directly connected to the 2.5G network port of R68S. The test results are:
iPerf3
The downlink rate is close to 2.1Gb, which may be limited by the performance of the USB network card. The uplink rate is slightly worse at 1.94Gb.
At this time, the CPU load is about 30%.
- .3.2 LAN TO LAN test (2.5G)
.5G QNAP NAS installed the intranet SPEEDTEST client and connected to the R68S at the same time as the laptop . The test results are as follows:
LAN TO The downlink rate of LAN
can reach around 2.3Gb, which is close to the limit, but the uplink rate still lags behind. In multiple tests, the fastest rate only reached 1.85Gb.
At the same time, I also tried large file transfer: the reading rate of
can be stable at 210MB, and the peak value can reach 242MB.
However, the writing speed is relatively unstable and fluctuates around 150MB. This may be due to a bottleneck in the hard disk speed, or the network card uplink is inherently slow.
LAN file transmission speed
- .3.2 LAN to wifi6 test (2.5g)
lan to wifi6
AX5400 direct connecting NAS, the maximum rate of WIFI6 is 1850/6 24. After the R68S repost, the rate decreases slightly, but the decrease is 5%, which is acceptable.
- .3.4 LAN Passing Test Summary
, if it is used in R68S DOCKER, the resumption of the local area network bridge will be greatly reduced. The firmware of the TML6R68S is fooled by the LEAN God, and it is used for 10 days Judging from the situation, stability is still very good.
running time
R68s I have been thrown in a weak electric box.
At present, Zhejiang's indoor temperature is almost about 27 degrees, and the R68S nude standby temperature is about 47 degrees. After the
shell arrived, I observed that the down temperature was soaring to about 53 degrees.
But the temperature has risen not only R68S. In the past two days, my Redmi AX5400 router, even if it is placed on the TV cabinet, the shell is already very hot. When
, when high temperature, the indoor temperature may be about 31 degrees, the temperature of the weak electric box will be higher, and I will continue to update the temperature monitoring of the R68S.
Zhejiang has been in high temperature for more than 40 days, and the indoor temperature is basically about 32 degrees. The R68S temperature of
is basically about 58 degrees in the weak electric box, and it is still very stable. It has not occurred and has not occurred.
三、玩机教程
这里主要讲一下刷机和部分插件安装的教程
.1 重置与刷机R68S带有重置按钮，如果系统混乱了，或者进不去路由器设置页面了，那么按住重The set button can restore the factory settings (under plug -in), which is still very convenient.
has to blow the R68S flashing machine, which can be completed in a few seconds. Compared with the N1 flashing, it is really different.
, but convenient and convenient, the tutorial still has to be learned.
) Install Risin Microcarmament and flashing toollink: https: //pan.baidu.com/s/1ptxq1yfjytccclhcy068aw? PWD = 6tnl extraction code: 6t NL
firmware Select
flashing order
can hear the prompt sound on the computer after about two seconds, and the device will be prompted on the development tool.
flashing machine
.2 upgrade. Tong's update is enough to use web upgrade:
web upgrade
web The upgrade requires the "Sysupgrade" image. The file suffix name is generally TAR.
.2 plug-in installationmachine comes with a pure version of firmware built into the factory. If you need to use certain firmware, you can upload the IPK or install it with the SSH command.
For example, the Alibaba Cloud Disk WebDAV plug-in that is very needed by audio and video players is not built-in when it leaves the factory. We can use the SSH tool or the TYYD copy command in OpenWRT to install it (copy one by one and press Enter):
wget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/aliyundrive-webdav_1.3.2-1_aarch64_generic.ipkwget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/luci-app-aliyundrive-webdav_1.3.2_all.ipkwget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/luci-i18n-aliyundrive-webdav-zh-cn_1.3.2-1_all.ipkopkg install aliyundrive-webdav_1.3.2-1_aarch64_generic.ipkopkg install luci-app-aliyundrive-webdav_1.3.2_all.ipkopkg install luci-i18n-aliyundrive-webdav-zh-cn_1.3.2-1_all.ipkAfter the installation is completed, refresh it and there will be a plug-in in the service:
Alibaba Cloud Disk
4. Summary and purchase suggestions
Electric Rhino R68S soft router, for me Personally it's very suitable.
Because I already have a finished NAS with a 2.5G network port, and the router usually uses AP mode, so for me, I only need a soft router with a 2.5G network port as a gateway.
The current soft routing market is very large. The price of J4125's 2.5G soft routing has also been reduced to about 600 yuan. If you have more needs, such as the standard all-in-one gameplay of Aikuai split + OpenWRT bypass + unraid, then R68S may Unable to meet your request.
But if you are like me and only need a pure soft router, then R68S is still very suitable for you.
(2), NanoPi R5S soft routing evaluation
NanoPi R5S unboxing
NanoPi with metal case SSD and cooling pad After unboxing R5S
, I found that the router has been assembled. They also came with 6 rubber feet and a piece of
NanoPi R5S
NanoPi R5S has a microSD card slot on one side, and the rear panel has a USB-C port for power supply, a WiFi antenna hole (this antenna hole can also be inserted into GPIO, UART console and other cables), two 2.5 GbERJ45 LAN port, a Gigabit Ethernet WAN port and HDMI video output .
NanoPi R5S mask key
On the other side of the NanoPi R5S we will find a mask button for firmware upgrades, while on the front panel there are four LED lights for the "System" and Ethernet ports, and two USB 3.0 port.
NanoPi R5S disassembly
Generally, the main reasons for wanting to disassemble are as follows: out of curiosity, to install M.2 NVMe SSD, to solder SPI flash memory, to connect some GPIO, RTC battery or to debug UART to TTL plate. It's easy to take it apart, just loosen the four screws that need to be loosened.
After taking it apart, you can see the bottom of the circuit board with the M.2 Key M slot, the space occupied by the SPI flash memory (right), and the Samsung KLM8G1GETF-B041 eMMC 5.1 flash memory 8GB motherboard.
SPI flash and M.2 socket on the NanoPi R5S board
Before removing the circuit board from the case, four more screws need to be loosened.
NanoPi R5S SBC
Rayson RS512M32LM4 D2BDS is a 2GB LPDDR4X memory chip. I found the RTL8211F (GbE) and
Metal Shell is the NANOPI of the CPU radiator. NANOPI R5s has a cooling pad that directly contacts metal shells, which can help achieve the best cooling effect.
uses 2.5GBE USB encrypted dogs and UP XTreme I11 mini PC for test settings. B encrypted dog has encountered some performance problems, but now this problem has been solved. Because Realtek sent me another RTL6RTL8156BG. TML7UP XTREME I11 mini PC transmission data , and perform full dual -work test, Its test results are 2.34Gbps/2.29Gbps.
I also use the same settings for testing, but this time the middle thing is replaced by NANOPI R5s.
NANOPI Taiwang switch , via
Xiaomi AX6000 router to connect the WAN port of the NANOPI R5S router to the Internet. This can prevent me from installing some installation on the equipment. Software package. Although the NANOPI R5s can take better photos on the top of the TP-Link switch, these two devices are very hot. I don't recommend doing this. Because my Ethernet cable was short, I had to move the router to the table for testing.OPENWRT and IPERF3 benchmark test
FRIENDLYWRT is already pre -installed on the router, so it can be used when the box is opened. You can also use "root" as a user and "Password" as a password to immediately access the LUCI interface or SSH. This is indeed very convenient, but it is not safe, and may violate laws in some countries . Anyway, it is best to change the password at least when you use it for the first time.
Friendlywrt status
Friendlywrt 3.0-RC1 and Linux 5.10.66 core. It is less than 250MB of RAMs used in idle and using default settings. Because the system is equipped with 2GB RAM, these ram are enough.
Friendlywrt network
L6. ROOT partition has 6.7 GB available in memory, and 920 KB is used in the ROOT partition. When all interfaces start at DHCP, you can obtain the IP address correctly. The device on the LAN can also be accessed by hostname.lan and obtained some IPV6 address.
Starting with the iperf3 benchmark, I first ran "iperf3 -s" on the NanoPi R5S and ran the following commands on the laptop:
- Download: (View the results of the Rx test on the NanoPiR5S)
iperf3 -t 60 -c 192.168.2.1 -i 10Connecting to host 192.168.2.1, port 5201[ 5] local 192.168.2.130 port 48782 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate Retr Cwnd[ 5] 0.00-10.00 sec 2.23 GBytes 1.92 Gbits/sec 0 1.69 MBytes[ 5] 10.00-20.00 sec 2.02 GBytes 1.74 Gbits/sec 0 1.69 MBytes[ 5] 20.00-30.00 sec 2.33 GBytes 2.00 Gbits/sec 0 2.64 MBytes[ 5] 30.00-40.00 sec 1.66 GBytes 1.42 Gbits/sec 0 2.64 MBytes[ 5] 40.00-50.00 sec 2.62 GBytes 2.25 Gbits/sec 0 2.64 MBytes[ 5] 50.00-60.00 sec 2.01 GBytes 1.73 Gbits/sec 0 2.64 MBytes- - - - - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr[ 5] 0.00-60.00 sec 12.9 GBytes 1.84 Gbits/sec 0 sender[5] 0.00-60.05 sec 12.9 GBytes 1.84 Gbits/sec receiveriperf Done.
- Upload (View the results of Tx test on NanoPiR5S):
$ iperf3 -t 60 -c 192.168.2.1 -i 10 -RConnecting to host 192.168.2.1, port 5201Reverse mode, remote host 192.168.2.1 is sending[ 5] local 192.168.2.130 port 48786 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate[ 5] 0.00-10.00 sec 1.29 GBytes 1.11 Gbits/sec[ 5] 10.00-20.00 sec 1.31 GBytes 1.12 Gbits/sec[ 5] 20.00-30.00 sec 1.33 GBytes 1.14 Gbits/sec[ 5] 30.00-40.00 sec 1.27 GBytes 1.09 Gbits/sec[ 5] 40.00-50.00 sec 1.30 GBytes 1.12 Gbits/sec[ 5] 50.00-60.00 sec 1.30 GBytes 1.12 Gbits/sec- - - - - - - - - - - - - - - - - - - - - - - - - -[ID] Interval Transfer Bitrate Retr[5] 0.00-60.06 sec 7.80 GBytes 1.12 Gbits/sec 0 sender[ 5] 0.00-60.00 sec 7.80 GBytes 1.12 Gbits/sec receiveriperf Done.
It can be seen from the above that this is not exactly the 2.35 advertised by FriendlyElec. Gbps and 1.85 Mbps because in this configuration I'm seeing 1.84 Gbps and 1.12 Gbps. There are also some changes on the Rx side when looking at the 10 second transfer test report. Fortunately, no retransmission issues were found.Now we have the same attempt on another WAN port on Nanopi R5s, and run from UP XTreme I11: TML31
download (RX): Devkit@UPX -I11: ~ $ iPerf3 -T 60 -C 192.168.2.1-i 10Connecting to host 192.168.2.1, port 5201 [5] Local 192.168.2.207 Port 52052 Connect to 192.168.2.1 port 5201 [ID] internsfer bitrate Retr Cwnd [5] 0.00 -10.00 SEC 2.59 GBITS/SEC 0 1.67 MBYTES [ 5] 10.00-20.00 SEC 2.62 GBYTES 2.25 GBITS/SEC 0 1.93 MBytes [5] 20.00-30.00 SEC 2.60 GBITS/SEC 0 1.93 MBYTES [5] 30.00-40.00 SEC 2.47 Gbytes 2.12 Gbits/sec 0 1.93 MBytes [5] 40.00-50.00 sec 2.43 GBytes 2.08 Gbits/sec 0 1.93 MBytes[ 5] 50.00-60.00 sec 2.45 GBytes 2.10 Gbits/sec 0 4.90 MBytes- - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr [5] 0.00-60.00 SEC 15.2 GBITS/Secits/Secret 0 SENDER [5] 0.00-60.00 SEC 15.1 GBITS/Secits/Secits TML42
- upload (TX);
Devkit @Devkit@ Upx -I11: ~ $ iPerf3 -T 60 -C 192.168.2.1 -i 10 -Ronnecting to host 192.168.2.1, Port 5201reverse Mode, Remote Host 192.168.2.1 is sending [5] Local 192.168.2.20.2020 7 port 52056 connected to 192.168. 2.1 Port 5201 [ID] Interval Transfer Bitrate [5] 0.00-10.00 SEC 1.31 GBITES 1.13 GBITS/SEC [5] 10.00-20.00 SEC 1.29 GBITS/SEC [5] 20.30.00 SEC 1.32222222222222 GBYTES 1.14 GBITS/SEC [5 ] 30.00-40.00 sec 1.30 GBytes 1.11 Gbits/sec[ 5] 40.00-50.00 sec 1.33 GBytes 1.14 Gbits/sec[ 5] 50.00-60.00 sec 1.27 GBytes 1.09 Gbits/sec- - - - - - - - - - - - - - - - - - - - --- [ID] Interval Transfer Bitrate Retr [5] 0.00-60.00 SEC 7.82 GBYTES 1.12 GBITS/SEC 0 SENDER [5] 0.00-60.00 SEC 7.82 GBITS/Secrets/Secret ERIPERF DONE. Whendownload (RX), it seems better to run at 2.17 Gbps; but when uploading (TX), it still runs slowly at 1.12 GBPS.
Let's take a look at the speed of using two WAN ports at the same time? One is to run "iPerf3 -S" on the UP Xtreme I11, and the other port runs the following commands on the notebook:
TML0TML7TML3
8
TML6
0htmm l7
1
2
3
4
5
6jaufranc@cnx -laptop -4: ~ iPerf3 -T 60 -C 192.168.2.207 -i 10htmlml 7
Connecting to host 192.168.2.207, Port 5201
[5] Local 192.168.2.130 Port 49762 Connect to 192.168.2.207 Port 5201
[ID] Interval Transfer Bitrate Retr CWNDTML3
[5] 0.00-10.00 SEC 1.39 GBITS/SEC 0 3.15 MBYTESTML3
[5] 10.00 -20.00 SEC 1.42 GBITS/SEC 0 3.15 MBYTES
[5] 20.00-30.00 SEC 1.44 GBITS/SEC 0 3.15 MBYTES
[5] 30.00-40.00 SEC 1.42 GBITES 1.22 GBITS/SEC 0 3.15 MBYTES
[5] 40.00-50.00.00 SEC 1.42 GBYTES 1.22 GBITS/SEC 0 3.15 MBYTES
[5] 50.00-60.00 SEC 1.39 GBITS/SEC 0 3.15 l3
[5] 0.00-60.00 SEC 8.49 GBITS/Secits/SEC 0 SENDER- -------------------- -
[id] interval transfer bitrate retr
[5] 0.00-60.05 SEC 8.48 GBYTES 1.21 GBITS/Sec itiverTML3IPERF 4
This data should be within the expected range It is because we have obtained the (TX) data previously lower; when we 1.21 Gbps, it is still a little higher than the 1.12 GBPS when we only use the upload (TX).We tried again:
$ iPerf3 -T 60 -C 192.168.207 -i 10 -RConnecting 192.168.2.207, por T 5201reverse Mode, Remote Host 192.168.2.207 is sending [5] Local 192.168.2.130 Port 49766 Connected to 192.68.2.207 Port 5201 [ID] Interval Transfer Bitrate [5] 0.00-10.00 SEC 2.04 GBITS/SEC [5] 10.00-20.00 SEC 2.04 GBITS/SE C [5] 20.00-30.00 SEC 2.04 GBYTES 1.75 GBITS /sec[ 5] 30.00-40.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 40.00-50.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 50.00-60.00 sec 2.05 GBytes 1.76 Gbits/sec- - - - - - - - - - - - - - - - - - - --- [ID] Interval Transfer Bitrate Retr [5] 0.00-60.05 SEC 12.3 GBITES 1.75 GBITS/SEC 0 SENDER [5] 0.00-60.00 SEC 12.3 GBITS/GBITS/ SeciverIperf Done.
is 1.75 Gbps this time. I don't know what happened or how this happened.followed me with a full dual work, testing the CPU, memory and temperature reported in LUCI.
NanoPi R5S的CPU负载
由上可看到其CPU负载介于2.0和2.5之间。
NANOPI R5s The temperature
CPU #0 is relatively low, CPU #1 to #3 Occupy most resources.In a room with an ambient temperature of 28 ° C, its temperature has never exceeded 60 ° C.
Friendlywrt use rate TML12
does Change. The performance of the full dual -work period is: 1.66 Gbps and 736 Mbps.The result of
is a bit disappointing. I am now intending to switch to Friendlycore (Ubuntu Core) to check whether similar results will be obtained, and then perform further testing. If you want me to evaluate other contents of OpenWrt, you can tell me in the comment area.
is further tested to switch to the FriendlyCore image based on Ubuntu 20.04, because I am more familiar with the Debian -based operating system, and some tools cannot run on OpenWRT. Note that the performance of the test results is still not ideal. This is why I call the prediction test. With the use and adjustment software of more and more people, the situation should be improved in the next few months.
OPENWRT optimization?
Now we do not discuss the mirror -based image based on Ubuntu. Because Friendelec (Guangzhou Friendship Electronic Technology Co., Ltd.) informed me that they had added some optimization, so I upgraded the updated version of Friendlywrt and tested this image: TML3. TML6 we did some of the new mirror image Optimized, such as network card interrupt settings, uninstalled support, etc.
" So I downloaded the" RK3568888 found on Google Drive on Google Drive -eflasher-Friendlywrt-20220526.img.gz ", and then record it with USBImager to , Then Boot to the router. After, it will automatically record the image into the EMMC flash. If you connect to the display, you can operate according to the results. After the operation is completed, remove the Microsd card and restart the router.
can be checked by connecting the HDMI display (as shown above) or viewing the LED on the device. This process is very fast, just a few seconds installed on the EMMC flash memory.
This new version of the image is mainly on 40-net-smpinity files.pre-installed before In FriendlyWrt, it looks like this:
0
1
2
3friendlyelec,nanopi-r5s)
set_interface_core 8 "eth0"
echo 7 /sys/class/net/eth0/queues/rx-0/rps_cpus
set_interface_core 2 "eth1-0"
set_interface_core 4 "eth1-16"
set_interface_core 4 "eth1-18"
echo b /sys/class/net/eth1/queues/rx-0/rps_cpus
set_interface_core 4 "eth2-0"
set_interface_core 2 "eth2-16"
set_interface_core 2 "eth2-18"
echo 9 /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
and the new ht The ml640-net-smp-affinity files are indeed different:
0
1
2
3friendlyelec,nanopi-r5s)
set_interface_core 8 "eth0"
echo f /sys/class/net/eth0/queues/rx-0/rps_cpus
set_interface_core 4 "eth1-0"
set_interface_core 4 "eth1-16"
set_interface_core 4 "eth1-18"
echo b /sys/class/net/eth1/queues/rx-0/rps_cpus
set_interface_core 2 "eth2-0"
set_interface_core 2 "eth2-16"
set_interface_core 2 "eth2-18"
echo d /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
Willy Tarreau also explained the reason for the changes made to the eth1 interface:
It involves RPS. They receive this IRQ (interrupt request) on core 2 and redistribute the incoming traffic into core 0, core 1, core 3. This is the correct way to use RPS. However, to achieve this, you must manually assign a network performance testing tool iperf and observe the first saturated core. If you use ksoftirqd to saturate core 2 first, then make sure iperf is runnable on any of the other 3 cores. If core2 is slightly idle, then try setting iperf on it. If putting iperf on top of it causes ksoftirqd to pop up, then they will hinder each other, and the user will be more willing to change the RPS settings to help free up another core and use it for iperf.
I actually didn't try this method before testing and switching to Ubuntu.When I try to use the new FriendlyWrt image, I get even worse results:
0
1
2
3
4
5
6
7
8
9
$ iperf3 -t 60 -c 192.168.2.1 -i 10
Connecting to host 192.168.2.1, port 5201
[ 5] local 192.168.2.130 port 49590 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 1.92 GBytes 1.65 Gbits/sec 0 1.62 MBytes
[ 5] 10.00-20.00 sec 1.91 GBytes 1.64 Gbits/sec 10 2.34 MBytes
[ 5] 20.00-30.00 sec 1.90 GBytes 1.63 Gbits/sec 0 2.61 MBytes
[ 5] 30.00-40.00 sec 1.85 GBytes 1.59 Gbits/sec 4 1.30 MBytes
[ 5] 40.00-50.00 sec 1.88 GBytes 1.61 Gbits/sec 1 1.06 MBytes
[ 5] 50.00-60.00 sec 1.76 GBytes 1.51 Gbits/sec 2 868 KByte /sec 17 sender
[ 5] 0.00-60.05 sec 11.2 GBytes 1.60 Gbits/sec receiver
iperf Done.
$ iperf3 -t 60 -c 192.168.2.1 -i 10 -R
Connecting to host 192.168.2.1, port 5201
Reverse mode, remote host 192.168.2.1 is sending
[ 5] local 192.168.2.130 port 49594 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.22 GBytes 1.05 Gbits/sec
[ 5] 10.00-20.00 sec 1.36 GBytes 1.17 Gbits/sec
[ 5] 20.00-30.00 sec 1.31 GBytes 1.12 Gbits/sec
[ 5] 30.00-40.00 sec 1.46 GBytes 1.26 Gbits/sec
[ 5] 40.00-50.00 sec 1.47 GBytes 1.26 Gbits/sec
[ 5] 50.00-60.00 sec 1.46 GBytes 1.26 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.05 sec 8.29 GBytes 1.19 Gbits/sec 1 sender
[ 5] 0.00-60.00 sec 8.29 GBytes 1.19 Gbits/sec receiver
iperf Done.
Therefore, this issue has to be re-examined.M.2 NVMe SSD installation in NanoPi R5S
I purchased an APACER AS2280 (AP256GAS2280P4-1) PCIe Gen 3.0 x4 SSD a while ago and it can achieve sequential read speeds of up to 1,800 MB/s and up to 1,100 on the right hardware Sequential write speed in MB/s.
Apacer M.2 2280 PCIe SSD
The installation process was easy as I only had to loosen four screws to remove the bottom cover, install the SSD and then secure it in place with the screws they provided.
M.2 NVMe SSD in NanoPi R5S Installation
Installing Ubuntu 20.04 FriendlyCore on NanoPi R5S
I first tried to install FriendlyCore using the eflasher mirror.
Installed FriendlyCore using eflasher mirror This is normal for mirroring. I tried again, going into the elasher UI settings by clicking "Finish", but it still didn't work.
Eflasher Installation FriendlyCore
Therefore, I downloaded the "SD" image to help boot directly from the microSD card and run the OS from there. After doing this, operation is still normal. However, if you plan to use the NanoPi R5S for multiple purposes and are also looking forward to using the desktop environment in the Ubuntu 20.04 image, you may be disappointed because the HDMI output can currently only be used to access the terminal.
Ubuntu 20.04 FriendlyElec Login
FriendlyCore System Information
You can find the startup log at CNX Software Pastebin. I logged in via SSH using pi/pi credentials (username/password) and upgraded the system to the latest packages:
sudo apt update
sudo apt dist-upgrade
Now we run some commands to get system information:
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ cat /etc/lsb-release
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=20.04
DISTRIB_CODENAME=focal
DISTRIB_DESCRIPTION="Ubuntu 20.04.4 LTS"
pi@FriendlyELEC:~$ uname -a
Linux FriendlyELEC 5.10.66 #219 SMP PREEMPT Fri Apr 22 18:20:21 CST 2022 aarch64 aarch64 aarch64 GNU/Linux
pi@FriendlyELEC:~$ free -mh
total used free shared buff/cache available
Mem: 1.9Gi 150Mi 1.7Gi 3.0Mi 114Mi 1.7Gi
Swap: 0B 0B 0B
pi@FriendlyELEC:~$ df -mh
Filesystem Size Used Avail Use% Mounted on
udev 969M 0 969M 0% /dev
tmpfs 197M 480K 196M 1% /run
overlay 27G 1013M 26G 4% /
tmpfs 981M 0 981M 0% /dev/shm
tmpfs 5.0M 4.0K 5.0M 1% /run/lock
tmpfs 981M 0 981M 0% /sys/fs/cgroup
tmpfs 197M 0 197M 0% /run/user/1000
Everything looks fine except that the NVMe drive is not automounted.(1), Electric Rhino R68S soft routing evaluation
1. Appearance and key configuration analysis
.1 R68S key configurationsElectric Rhino R68S uses Rockchip RK-3568 ARM processor, 22nm process, 4-core Cortex A55 architecture, clocked at 2Ghz, supports up to 8G memory;
RK3568 has built-in 1 QSGMII, 2 PCIE 3.0X1 channels, and 2 Gigabit Ethernet interfaces, so it can be used on a router to achieve 2x2.5Gb+2x1000Mb 4 network port combination.
R68S core configuration is shown in the figure below:
core configuration
CPU The two chips next to are memory and flash memory chip . The
LPDDR4 memory chip comes from Rayson, the model is RS1G32LF4D2BDS-53BT, and the capacity is 4GB (the 2GB version is 70 yuan cheaper); the
eMMC flash memory chip comes from FORESEE (Shenzhen Jiangbo Longqi Bottom), the model is NCEMASKG-16G, the capacity is 16GB; there is a Gigabit Ethernet switching chip in the upper left corner of the
processor, the model is RTL8211F. When you see this crab logo, you know it is from Realtek ;
does not have 4 Network port? What about the other 3 switching chips?
is on the back of the motherboard.
The two 2.5Gb switching chips are also from Realtek, the model is RTL8125BG:
The back of the motherboard
Realtek's 8125B, Intel 's I225 B3 and Qualcomm 's QCA8081 are currently the most common
.2 R68S appearance.2 NAT Performanceengineering machine was sent with only a bare board.
Because I wanted to test the temperature inside the weak current box , so a few days later the boss sent another shell (for trial molding, not the final version).
The shell I received is made of plastic (there is also a CNC aluminum shell, the price is about 100 yuan), and there are a lot of heat dissipation holes.
The official version of the casing will be made with a frosted effect
The casing and motherboard
Because the openings of the casing are not equipped with dustproof cotton, so if you put the R68S on the desktop, you still need to consider the dust problem. After all, dust is the first concern of electronic equipment. A killer:
Front
There are also a large number of heat dissipation holes on the two sides of the machine:
Side of the machine
If you, like me, place the R68S in a weak current box, then the hollow design of the shell is actually very beneficial to heat dissipation.
But if you use R68S on the desktop, I recommend buying a CNC aluminum casing, which is dustproof and waterproof, and has good heat dissipation. The
CNC shell looks like this:
CNC shell preliminary model
.3 Power supply.1 plus decomposition performanceR68S can be purchased as a set with power supply or as a stand-alone version.
I actually tested it and found that can also work normally with a 12V 1A optical modem power adapter.
Of course, if you need to connect an external hard drive, I suggest you use a 12V 1.5A power adapter.
Speaking of external hard drives, the USB 3.0 performance of R68S is still very strong, and it can read and write up to about 120MB. If you want to connect a hard drive to the R68S to watch movies, you don’t have to worry about the original high-bitrate movie being stuck.
2. Performance and stability test
For an soft router , performance and stability are equally important.
directly look at the table:
Zhihu@m
CPU Coremarkml7
AES ...
h3CHA ...
AES hard solution
R68S (RK3568)
l24
is
N1 (S905D)
8 704
is L6 24 R4S (RK3399)
36089
is
3j1900
No
J4125
51102It is
R68S. The decryption performance is much stronger than the N1 of the ARM architecture , but about 20%of R4S (RK3399) behind 6 cores.
R68S CPU Coremark is behind J1900 about 6000 points. However, after all, the J1900 does not support hardware AES decoding, and R68S is still advantageous in this regard. If
is compared with J4125, which also supports AES hard solution, R68S is less than half behind. Although
is not excellent in scores, it is more than enough in actual use.
NAT Performance mainly tests broadband rates. The test topology is as follows:
NAT test
test rate as follows:
can be seen that if the transmission loss is removed, the R68S has reached the limit of the Gigabit Broadband.
is about 45%at this time.
.3 LAN forwarding performanceLAN forwarding capability The main test of the forwarding rate between the two 2.5G LAN ports of R68S, I did a total of 3 different tests.
R68S and computer are installed on the iPerf3 client, ThinkPad X13 (2.5G USB 3.0 network card) Direct connection of 2.5G network port of R68S, test results:
i6i Perf3downward The rate is close to 2.1GB, which may be limited by the performance of the USB network card. The uplink rate is slightly worse, 1.94GB.
is about 30%at this time.
, I also tried large file transmission:
read rate can be stable at 210MB, and the peak can run to 242MB.
is not stable when writing speed. It will fluctuate up and down at 150MB. It may be a bottleneck at the hard disk rate, or it may be slow.
LAN file transmission speed
HT The AX200 network card of ml0thinkpad X13 can achieve 2402MB negotiation rate under Redmi AX5400 e -version router, and the laptop computer uses wifi most of the time, so LAN TO WIFI6 test is also necessary:
lan to wifi6
AX5400 direct connecting NAS, the maximum rate of WIFI6 is 1850/6 24. After the R68S repost, the rate decreases slightly, but the decrease is 5%, which is acceptable.
Overall, R68S's LAN forwarding ability is still quite possible. 2.5G switch.
, if it is used in R68S DOCKER, the resumption of the local area network bridge will be greatly reduced. The firmware of the TML6R68S is fooled by the LEAN God, and it is used for 10 days Judging from the situation, stability is still very good.
running time
R68s I have been thrown in a weak electric box.
At present, Zhejiang's indoor temperature is almost about 27 degrees, and the R68S nude standby temperature is about 47 degrees. After the
shell arrived, I observed that the down temperature was soaring to about 53 degrees.
But the temperature has risen not only R68S. In the past two days, my Redmi AX5400 router, even if it is placed on the TV cabinet, the shell is already very hot. When
, when high temperature, the indoor temperature may be about 31 degrees, the temperature of the weak electric box will be higher, and I will continue to update the temperature monitoring of the R68S.
Zhejiang has been in high temperature for more than 40 days, and the indoor temperature is basically about 32 degrees. The R68S temperature of
is basically about 58 degrees in the weak electric box, and it is still very stable. It has not occurred and has not occurred.
三、玩机教程
这里主要讲一下刷机和部分插件安装的教程
.1 重置与刷机R68S带有重置按钮，如果系统混乱了，或者进不去路由器设置页面了，那么按住重The set button can restore the factory settings (under plug -in), which is still very convenient.
has to blow the R68S flashing machine, which can be completed in a few seconds. Compared with the N1 flashing, it is really different.
, but convenient and convenient, the tutorial still has to be learned.
) Install Risin Microcarmament and flashing toollink: https: //pan.baidu.com/s/1ptxq1yfjytccclhcy068aw? PWD = 6tnl extraction code: 6t NL
firmware Select
flashing order
can hear the prompt sound on the computer after about two seconds, and the device will be prompted on the development tool.
flashing machine
.2 upgrade. Tong's update is enough to use web upgrade:
web upgrade
web The upgrade requires the "Sysupgrade" image. The file suffix name is generally TAR.
.2 plug -in installationmachine is built -in pure version of the firmware. If you need to use certain firmware, you can upload IPK or SH command installation.
. For example, the Alibaba Cloud Plate WEBDAV plug -in that the audio and video player is very needed is not built in the factory. We can use the SSH tool or OPENWRT Tyyd to enter the command installation (replicate the car one by one):
wget https://github.com/messense /Aliyundrive-BDAV/releases/download/v1.3.2/aliyundrive-webdav_1.3.2-1_Aarch64_gneric.ipkwget https://github.com/messense /Aliyundrive-Webdav/releases/download/v1.3.2/luci- app-aliyundrive- webdav_1.3.2_all.ipkwget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/luci-i18n -webdav-zh- cn_1.3.2-1_all.ipkopkg Install Aliyyundrive-Webdav_1. 3.2-1_AARCH64_GENERIC.IPKOPKGG Install Luci-ALIYUNDrive-Webdav_1.3.2_ALL.IPKOPKG Install Luci-I18n-WEBDAV-ZH-CN_1.3.2-1_ All.ipkis refreshed after installation is complete, and the service has plug -in:
Alibaba Cloud Plate4. Summary and purchase suggestions
Electric Rhino R68S soft routing, which is very suitable for me.
is because I already have the finished NAS with a 2.5G network port. The router usually uses an AP mode, so for me, you only need to have a soft route with 2.5G mesh.
The current soft routing market is very rolled, and the 2.5G soft route of J4125 has also reduced the price to about 600 yuan. If you have more needs, such as love fast diversion+OpenWRT bypass+UNRAID's standard All in one gameplay, then R68S may be possible Can't meet your requirements.
But if you just need a pure soft route, then R68S is still very suitable for you.
(2), NANOPI R5s soft routing evaluation
NANOPI R5s
with metal shell ssd and sanoPi R R 5S
I found that the router has been assembled. They also accompanied 6 rubber foot pads and a section of
NANOPI R5STML12
NANOPI is there. With a USB-C port for power supply, a WIFI antenna hole (this antenna hole can also insert GPIO, UART CONSOLE and other cables), two 2.5 2.5 GBE RJ45 LAN port, a Gigabit Ethernet WAN port and HDMI video output .
NANOPI R5s cover key
We will find one for solidarity The cover button for upgrade, four on the front panel for LED lights for "system" and Ethernet ports, and two USB 3.0 port.NANOPI R5s Disassemble
. Generally, the main reason for disassembling is the following:: Out of curiosity, to install M.2 NVMe, to welded SPI flash memory, you need to connect some GPIO, RTC battery, or adjust the UART to TTL to TTL Test plate. If you want to open it, it is easy to open. You can easily disassemble the four screws that need to be released.
can see the bottom of the circuit board with the M.2 KEY M slot, the SPI flash memory occupation space (right), and the Samsung KLM8G1GETF-B041 EMMC 5.1 flash memory 8GB.
NANOPI R5S SPI flash memory and M.2 socket
need to remove the four screws before taking the circuit board from the shell.
NANOPI R5S Sbc
R6RAYSON RS512LM4 D2BDS is a 2GB LPDDR4X memory chip. RTL8211F (GBE) and
Metal Shell is the NANOPI of the CPU radiator. NANOPI R5s has a cooling pad that directly contacts metal shells, which can help achieve the best cooling effect.
uses 2.5GBE USB encrypted dogs and UP XTreme I11 mini PC for test settings. B encrypted dog has encountered some performance problems, but now this problem has been solved. Because Realtek sent me another RTL6RTL8156BG. TML7UP XTREME I11 mini PC transmission data , and perform full dual -work test, Its test results are 2.34Gbps/2.29Gbps.
I also use the same settings for testing, but this time the middle thing is replaced by NANOPI R5s.
NANOPI Taiwang switch , via
Xiaomi AX6000 router to connect the WAN port of the NANOPI R5S router to the Internet. This can prevent me from installing some installation on the equipment. Software package. Although the NANOPI R5s can take better photos on the top of the TP-Link switch, these two devices are very hot. I don't recommend doing this. Because my Ethernet cable was short, I had to move the router to the table for testing.OPENWRT and IPERF3 benchmark test
FRIENDLYWRT is already pre -installed on the router, so it can be used when the box is opened. You can also use "root" as a user and "Password" as a password to immediately access the LUCI interface or SSH. This is indeed very convenient, but it is not safe, and may violate laws in some countries . Anyway, it is best to change the password at least when you use it for the first time.
Friendlywrt status
Friendlywrt 3.0-RC1 and Linux 5.10.66 core. It is less than 250MB of RAMs used in idle and using default settings. Because the system is equipped with 2GB RAM, these ram are enough.
Friendlywrt network
L6. ROOT partition has 6.7 GB available in memory, and 920 KB is used in the ROOT partition. When all interfaces start at DHCP, you can obtain the IP address correctly. The device on the LAN can also be accessed by hostname.lan and obtained some IPV6 address.
Starting with the iperf3 benchmark, I first ran "iperf3 -s" on the NanoPi R5S and ran the following commands on the laptop:
- Download: (View the results of the Rx test on the NanoPiR5S)
iperf3 -t 60 -c 192.168.2.1 -i 10Connecting to host 192.168.2.1, port 5201[ 5] local 192.168.2.130 port 48782 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate Retr Cwnd[ 5] 0.00-10.00 sec 2.23 GBytes 1.92 Gbits/sec 0 1.69 MBytes[ 5] 10.00-20.00 sec 2.02 GBytes 1.74 Gbits/sec 0 1.69 MBytes[ 5] 20.00-30.00 sec 2.33 GBytes 2.00 Gbits/sec 0 2.64 MBytes[ 5] 30.00-40.00 sec 1.66 GBytes 1.42 Gbits/sec 0 2.64 MBytes[ 5] 40.00-50.00 sec 2.62 GBytes 2.25 Gbits/sec 0 2.64 MBytes[ 5] 50.00-60.00 sec 2.01 GBytes 1.73 Gbits/sec 0 2.64 MBytes- - - - - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr[ 5] 0.00-60.00 sec 12.9 GBytes 1.84 Gbits/sec 0 sender[5] 0.00-60.05 sec 12.9 GBytes 1.84 Gbits/sec receiveriperf Done.
- Upload (View the results of Tx test on NanoPiR5S):
$ iperf3 -t 60 -c 192.168.2.1 -i 10 -RConnecting to host 192.168.2.1, port 5201Reverse mode, remote host 192.168.2.1 is sending[ 5] local 192.168.2.130 port 48786 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate[ 5] 0.00-10.00 sec 1.29 GBytes 1.11 Gbits/sec[ 5] 10.00-20.00 sec 1.31 GBytes 1.12 Gbits/sec[ 5] 20.00-30.00 sec 1.33 GBytes 1.14 Gbits/sec[ 5] 30.00-40.00 sec 1.27 GBytes 1.09 Gbits/sec[ 5] 40.00-50.00 sec 1.30 GBytes 1.12 Gbits/sec[ 5] 50.00-60.00 sec 1.30 GBytes 1.12 Gbits/sec- - - - - - - - - - - - - - - - - - - - - - - - - -[ID] Interval Transfer Bitrate Retr[5] 0.00-60.06 sec 7.80 GBytes 1.12 Gbits/sec 0 sender[ 5] 0.00-60.00 sec 7.80 GBytes 1.12 Gbits/sec receiveriperf Done.
It can be seen from the above that this is not exactly the 2.35 advertised by FriendlyElec. Gbps and 1.85 Mbps because in this configuration I'm seeing 1.84 Gbps and 1.12 Gbps. There are also some changes on the Rx side when looking at the 10 second transfer test report. Fortunately, no retransmission issues were found.Now we have the same attempt on another WAN port on Nanopi R5s, and run from UP XTreme I11: TML31
download (RX): Devkit@UPX -I11: ~ $ iPerf3 -T 60 -C 192.168.2.1-i 10Connecting to host 192.168.2.1, port 5201 [5] Local 192.168.2.207 Port 52052 Connect to 192.168.2.1 port 5201 [ID] internsfer bitrate Retr Cwnd [5] 0.00 -10.00 SEC 2.59 GBITS/SEC 0 1.67 MBYTES [ 5] 10.00-20.00 SEC 2.62 GBYTES 2.25 GBITS/SEC 0 1.93 MBytes [5] 20.00-30.00 SEC 2.60 GBITS/SEC 0 1.93 MBYTES [5] 30.00-40.00 SEC 2.47 Gbytes 2.12 Gbits/sec 0 1.93 MBytes [5] 40.00-50.00 sec 2.43 GBytes 2.08 Gbits/sec 0 1.93 MBytes[ 5] 50.00-60.00 sec 2.45 GBytes 2.10 Gbits/sec 0 4.90 MBytes- - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr [5] 0.00-60.00 SEC 15.2 GBITS/Secits/Secret 0 SENDER [5] 0.00-60.00 SEC 15.1 GBITS/Secits/Secits TML42
- upload (TX);
Devkit @Devkit@ Upx -I11: ~ $ iPerf3 -T 60 -C 192.168.2.1 -i 10 -Ronnecting to host 192.168.2.1, Port 5201reverse Mode, Remote Host 192.168.2.1 is sending [5] Local 192.168.2.20.2020 7 port 52056 connected to 192.168. 2.1 Port 5201 [ID] Interval Transfer Bitrate [5] 0.00-10.00 SEC 1.31 GBITES 1.13 GBITS/SEC [5] 10.00-20.00 SEC 1.29 GBITS/SEC [5] 20.30.00 SEC 1.32222222222222 GBYTES 1.14 GBITS/SEC [5 ] 30.00-40.00 sec 1.30 GBytes 1.11 Gbits/sec[ 5] 40.00-50.00 sec 1.33 GBytes 1.14 Gbits/sec[ 5] 50.00-60.00 sec 1.27 GBytes 1.09 Gbits/sec- - - - - - - - - - - - - - - - - - - - --- [ID] Interval Transfer Bitrate Retr [5] 0.00-60.00 SEC 7.82 GBYTES 1.12 GBITS/SEC 0 SENDER [5] 0.00-60.00 SEC 7.82 GBITS/Secrets/Secret ERIPERF DONE. Whendownload (RX), it seems better to run at 2.17 Gbps; but when uploading (TX), it still runs slowly at 1.12 GBPS.
Let's take a look at the speed of using two WAN ports at the same time? One is to run "iPerf3 -S" on the UP Xtreme I11, and the other port runs the following commands on the notebook:
TML0TML7TML3
8
TML6
0htmm l7
1
2
3
4
5
6jaufranc@cnx -laptop -4: ~ iPerf3 -T 60 -C 192.168.2.207 -i 10htmlml 7
Connecting to host 192.168.2.207, Port 5201
[5] Local 192.168.2.130 Port 49762 Connect to 192.168.2.207 Port 5201
[ID] Interval Transfer Bitrate Retr CWNDTML3
[5] 0.00-10.00 SEC 1.39 GBITS/SEC 0 3.15 MBYTESTML3
[5] 10.00 -20.00 SEC 1.42 GBITS/SEC 0 3.15 MBYTES
[5] 20.00-30.00 SEC 1.44 GBITS/SEC 0 3.15 MBYTES
[5] 30.00-40.00 SEC 1.42 GBITES 1.22 GBITS/SEC 0 3.15 MBYTES
[5] 40.00-50.00.00 SEC 1.42 GBYTES 1.22 GBITS/SEC 0 3.15 MBYTES
[5] 50.00-60.00 SEC 1.39 GBITS/SEC 0 3.15 l3
[5] 0.00-60.00 SEC 8.49 GBITS/Secits/SEC 0 SENDER- -------------------- -
[id] interval transfer bitrate retr
[5] 0.00-60.05 SEC 8.48 GBYTES 1.21 GBITS/Sec itiverTML3IPERF 4
This data should be within the expected range It is because we have obtained the (TX) data previously lower; when we 1.21 Gbps, it is still a little higher than the 1.12 GBPS when we only use the upload (TX).We tried again:
$ iPerf3 -T 60 -C 192.168.207 -i 10 -RConnecting 192.168.2.207, por T 5201reverse Mode, Remote Host 192.168.2.207 is sending [5] Local 192.168.2.130 Port 49766 Connected to 192.68.2.207 Port 5201 [ID] Interval Transfer Bitrate [5] 0.00-10.00 SEC 2.04 GBITS/SEC [5] 10.00-20.00 SEC 2.04 GBITS/SE C [5] 20.00-30.00 SEC 2.04 GBYTES 1.75 GBITS /sec[ 5] 30.00-40.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 40.00-50.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 50.00-60.00 sec 2.05 GBytes 1.76 Gbits/sec- - - - - - - - - - - - - - - - - - - --- [ID] Interval Transfer Bitrate Retr [5] 0.00-60.05 SEC 12.3 GBITES 1.75 GBITS/SEC 0 SENDER [5] 0.00-60.00 SEC 12.3 GBITS/GBITS/ SeciverIperf Done.
is 1.75 Gbps this time. I don't know what happened or how this happened.followed me with a full dual work, testing the CPU, memory and temperature reported in LUCI.
NanoPi R5S的CPU负载
由上可看到其CPU负载介于2.0和2.5之间。
NANOPI R5s The temperature
CPU #0 is relatively low, CPU #1 to #3 Occupy most resources.In a room with an ambient temperature of 28 ° C, its temperature has never exceeded 60 ° C.
Friendlywrt use rate TML12
does Change. The performance of the full dual -work period is: 1.66 Gbps and 736 Mbps.
This result is a bit disappointing, I now plan to switch to FriendlyCore (Ubuntu Core) to check if I will get similar results, and then perform further tests. If you want me to review other aspects of OpenWrt, you can tell me in the comment area.
took the test a step further and switched to the FriendlyCore image based on Ubuntu 20.04 because I was more familiar with Debian-based operating systems and some tools did not run on OpenWrt. Note that the performance shown in the test results is still not very ideal, which is why I call it a pre-test. As more people use and adjust the software, things should improve in the next few months.
OpenWrt optimization?
Let’s not discuss Ubuntu-based images for now. Because FriendElec (Guangzhou Friendly Electronic Technology Co., Ltd.) told me that they added some optimizations, so I upgraded to the newer version of FriendlyWrt and tested this image:
This is what they said: " We have made some optimizations to the new image, such as network card interrupt settings, uninstall support, etc. ” So I downloaded it on Google. "rk3568-eflasher-friendlywrt-20220526.img.gz" found on Drive, then use USBImager to burn it to the microSD card, and then boot into the router.
After doing this, it will automatically burn the image into the eMMC flash memory. If you connect a monitor, you can follow the results. After the operation is completed, remove the microSD card and restart the router.
Once this is done, you can check the status by connecting an HDMI display (shown above) or by looking at the LEDs on the device. The process is very fast and installation to the eMMC flash memory only takes a few seconds.
This new version of the image is mainly , which has made changes to the 40-net-smp-affinity file.In the previously pre -installed Friendly, it looks like this:
0
1H7H7
2
3FriendlyElec, Nanopi-R5s)
set_interface_core 8 "ETH0"
ECHO 7/SYS/Class/NET/ETH0/Queues/RPS_CPS
SET_IN Terface_core 2 "ETH1-0"Set_interface_core 4 "ETH1-16"
Set_interface_core 4 "ETH1-18"
EML6E Cho B/SYS/Class/NET/ETH1/Queues/RX-0/RPS_CPushml7
SET_INTERFACE_CORE 4 "ETH2-0"
Set_interf ACE_CORE 2 "ETH2-16"
SET_INTERFACE_CORE 2 "ETH2- 18"
echo 9 /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
而新的
0
1
2
3friendlyelec,nanopi-r5s)
set_interface_core 8 "eth0"
echo f /SYS/Class/Net/Eth0/Queues/RX-0/RPS_CPUS
SET_INTERFACE_COR Core 4 "ETH1-16" Hilml7
SET_INTERFACE_COR -0/rps_cpus
set_interface_core 2 "eth2-0"
set_interface_core 2 "eth2-16"
set_interface_core 2 "eth2-18"
echo d /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
Willy Tarreau也解释了Reasons for changing the ETH1 interface:
It involves rps. They received this IRQ (interrupt request) on Core 2 and re -assigned the flow to Core 0, Core 1, and Core 3. This is the correct method of using RPS. However, to reach this point, you must manually distribute a network performance testing tool iPerf and observe the first saturated Core. If you first use KSOFTIRQD to saturate Core 2, then make sure that iPerf can run on any of the other three cores. If Core2 is slightly idle, try to set an iPerf on it. If you put the iPerf on it will pop up KSoftirqd, then they will hinder each other, and users will be more willing to change the RPS settings to help release another Core and use it for IPERF.
Before testing and switching to Ubuntu, I actually didn't try this method.When I tried to use the new Friendlywrt mirror, the result was worse: $ iPerf3 -T 60 -C 192.168.2.1 -i 10
Connecting to host 192.168.2.1, Port 5201
[5] LOCAL 192.168.2.130 Port 49590 Connect to 192.168. 2.1 Port 5201
[ID] interval transfer bitrate Retr CWND
[5] 0.00-10.00 SEC 1.92 GBYTES 1.65 GBITS/ SEC 0 1.62 MByteml7
[5] 10.00-20.00 SEC 1.91 GBYTES 1.64 GBITS/SEC 10 2.34 MBYTES
[5] 20.00-30.00 s EC 1.90 GBytes 1.63 Gbits/sec 0 2.61 MBytes
[ 5] 30.00-40.00 sec 1.85 GBytes 1.59 Gbits/sec 4 1.30 MBytes
[ 5] 40.00-50.00 sec 1.88 GBytes 1.61 Gbits/sec 1 1.06 MBytes
[ 5] 50.00-60.00 sec 1.76 GBytes 1.51 GBITS/SEC 2 868 KBYTES
- ------------------
[ID] intermfer bitrate TR[5] 0.00-60.00 SEC 11.2 Gbytes 1.61 Gbits/Secret 17 SENDER
[5] 0.00-60.05 SEC 11.2 GBYTES 1.60 GBITS/Secire
IPERF ml7$ iPerf3 -T 60 -C 192.168.2.1 -i 10 -RTML3
Connecting to host 192.168.2.1, Port 5201
reverse mode, remote host 192.168 .2.1 is sending
[5] LOCAL 192.168.2.130 Port 49594 Connected to 192.168.2.1 Port 5201
[ID] Internet Val Transfer Bitraate[5] 0.00-10.00 SEC 1.22 GBYTES 1.05 GBITS/SEC
[5] 10.00-20.00 SEC 1.36 s 1.17 GBITS /sec
[5] 20.00-30.00 SEC 1.31 GBYTES 1.12 Gbits/Sec
[5] 30.00-40.00 SEC 1.26 gbites 1.26 gbites 1.26 gbites S/SEC[5] 40.00-50.00 SEC 1.47 GBYTES 1.26 GBITS/Sec
[5] 50.00-60.00 SEC 1.46 Gbytes 1.26 Gbits/Sec- - - - ---------------------
[id] interval 7[5] 0.00-60.05 SEC 8.29 GBITES 1.19 GBITS/SEC 1 SENDER
[5] 0.00-60.00 SEC 8.29 GBYTES 1.19 GBITS/Sec itiver
IPERF Done.
So, this problem has to be re -examined.NANOPI R5s M.2 NVME SSD Install
I bought Apacer AS2280 (AP256GAS2280P4-1) PCIE Gen 3.0 X4 SSD not long ago. It can achieve the correct hardware on the correct hardware. Reading speed and as high as 1,100 as high as 1,800 mb/s The sequence of MB/s is written.
APACER M.2 2280 PCIE SSD
L6 installation process is simple because I only need to loosen four four four four. The screws can be removed, installed SSD, and then fixed it with the screws they provided.
M.2 NVMe SSD in NanoPi R5S Installation
Installing Ubuntu 20.04 FriendlyCore on NanoPi R5S
I first tried to install FriendlyCore using the eflasher mirror.
Installed FriendlyCore using eflasher mirror This is normal for mirroring. I tried again, going into the elasher UI settings by clicking "Finish", but it still didn't work.
Eflasher Installation FriendlyCore
Therefore, I downloaded the "SD" image to help boot directly from the microSD card and run the OS from there. After doing this, operation is still normal. However, if you plan to use the NanoPi R5S for multiple purposes and are also looking forward to using the desktop environment in the Ubuntu 20.04 image, you may be disappointed because the HDMI output can currently only be used to access the terminal.
Ubuntu 20.04 FriendlyElec Login
FriendlyCore System Information
You can find the startup log at CNX Software Pastebin. I logged in via SSH using pi/pi credentials (username/password) and upgraded the system to the latest packages:
sudo apt update
sudo apt dist-upgrade
Now we run some commands to get system information:
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ cat /etc/lsb-release
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=20.04
DISTRIB_CODENAME=focal
DISTRIB_DESCRIPTION="Ubuntu 20.04.4 LTS"
pi@FriendlyELEC:~$ uname -a
Linux FriendlyELEC 5.10.66 #219 SMP PREEMPT Fri Apr 22 18:20:21 CST 2022 aarch64 aarch64 aarch64 GNU/Linux
pi@FriendlyELEC:~$ free -mh
total used free shared buff/cache available
Mem: 1.9Gi 150Mi 1.7Gi 3.0Mi 114Mi 1.7Gi
Swap: 0B 0B 0B
pi@FriendlyELEC:~$ df -mh
Filesystem Size Used Avail Use% Mounted on
udev 969M 0 969M 0% /dev
tmpfs 197M 480K 196M 1% /run
overlay 27G 1013M 26G 4% /
tmpfs 981M 0 981M 0% /dev/shm
tmpfs 5.0M 4.0K 5.0M 1% /run/lock
tmpfs 981M 0 981M 0% /sys/fs/cgroup
tmpfs 197M 0 197M 0% /run/user/1000
Everything looks fine except that the NVMe drive is not automounted.Now we find more details with INXI: @Friendlyelec: ~ $ inxi -fc0SYSYSTEM:
TML3HOST: Friendlyelec Kernel: 5.10.66 AARCH 64 Bits: 64 Console: TTY 0Distro: Ubuntu 20.04.4 LTS (FOCAL)
MACHINE:
Type: ARM Device System: Friendly NANOPI R5s Details: N/A
Serial: 8CBFE79E107C459C
BATRY:
html-1: test_battery charge: 100% Condition: n/A
CPU:
toPology: Quad core A variant: cortex-a55 bits: 64 Type: mcpSpeed: 408 MHz Min/Max: 408/1992 MHz Core Speeds (MHz): 1: 1992 2: 1992Device-: Display-Subsystem Driver: RockChip_drm V: N/A
Device-2: Mali-Bifrost Driver: Mali v: n/aDevice-3: RK3568-DW-HDMI Driver: dwhdmi_rockChip v: n/a
Display: Server: x. x. x. ORG 1.20.8 Driver: DWHDMI_RockChip Ty: 80x24TML3MESSAGE: Advanced Graphics Data UnavaiLe. Display
AUDIO:
Device-1: RK3568-DW-HDMI Driver: DWHDMI_RockChip
Device-2: Simple-Audio-Card Driver: Asoc_simple_card
Device-3: Simple-Audio-Card Driver: n/A
ML0 Device-4: Simple-Audio-Card Driver: ASOC_SIMPLE_CARD
Sound Server : ALSA V: K5.10.66
Network:
Device-1: Realtek RTL8125 2.5GBE Driver: ml0if: ETH1 State: DOWN MAC: E2: 1D: 62: A1: 1A: Ca
Device-2: Realtek RTL8125 2.5GBE Driver: R8125
IF: ETH1 State: DOWN MAC: E2: 1D: 62: A1: 1A: Ca
Device-3: RK3568-GMAC Driver: RK_GMAC_DWMACI6F-ID -1: ETH0 State: Up Speed: 1000 Mbps Duplex: FullTML3MAC: 1d: 62: A1: 1A: Ca
IF-ID-2: ETH2 State: DOWN MAC: 12: BF: 2B: D6: 4B: E0
DRIVES:H6H6H6H6H6H6hl6LOCAL Storage: Total: 274.88 GIB USED: 1012.3 mib (0.4%)
id-: /DEV /MMCBLK0 Model : SD16G SIZE: 2912 Gib
ID-2: /DEV /MMCBLK2 Model: 8GTF4R SIZE: 7.28 gib
id-3: /dev /nvme00 n1 Vendor: Apacer Model: AS2280P4 256GB SIZE: 238.47 Gib
PARTION: TML3
id-: 26.48 Gib use d: 1012.3 MIB (3.7%) FS: Overlay
SOURCE: ERR-102
TML3Sensors: SYSTEM Te Mperatures: CPU: 46.1 C Mobo: N/AFAN Speeds (RPM): N/A
Il6Info:PRO sets: 130 uptime: 7M Memory: 1.92 Gib use: 211.4 mib ( 10.8%)
Init: systemd Shell: bash inxi: 3.0.38
只有eth0 WAN端口打开了，eth1/eth2 2.5GbE端口是关闭的，根本没有显示配置。The FriendlyElec side seems to be mainly focused on the FriendlyWrt mirror, and they told me that optimizations haven't been implemented on FriendlyCore yet, so most people are probably still using FriendlyWrt because it's easier to configure network and router settings. I saw that the Apacer AS2280P4 SSD was actually detected, but it wasn't formatted out of the box, so I had to format it with mkfs.ext4.NanoPi R5S基准测试
现在我们通过在路由器上运行SBC Bench的方式来对CPU进行基准测试，希望尽可能可以发现一些问题：$ sudo /bin/bash ./sbc-bench.sh -c
[sudo] password for Pi:
WARNING: DMESG OUTPUT Does Not Contain Early Boot Messages Whitml7
Help in IDentifying Re Details.IT is Recoming to Reboot Now and then Execute The Benchmarks.
Press [Ctrl]- [C] To Stop or [Enter] to Continue.
AVERAGE LOAD and/or CPU Utilization too high (too much background activity). Waiting ...
too busy for benchmarking: 07:21:06 up 3 min, 1 user, LOA, LOA, LOA, LOA, LOA, LOA D Average: 0.41, 0.27, 0.11, CPU: 3%
TOO Busy For BenchMarking: 07:21:11 up 3 min, 1 user, load average: 0.38, 0.26, 0.11, CPU: 1%
TML3TOO BUSYMARKING: 07:21:16 up 3 min, 1 user, load average: 0.35, 0.26, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:21 up 3 min, 1 user, load average: 0.32, 0.25, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:26 up 3 min, 1 user, load average: 0.29, 0.25, 0.11, CPU: 1%
too Busy for Bundmarking: 07:21:31 up 3 min, 1 user, load average: 0.27, 0. 24, 0.10, CPU: 1%
SBC- Bench v0.9.7
Installing Needed Tools. This may take some time. Done.checking cpufreq opp. DONE (Results Will Be Available in 20-28 Minutes) .ECUTING TINYMEMBENCH. Done.
Ecuting Ram Lam ATENCY TESTER. Done.ECUTING Openssl Benchmark. Done. .executing 7-zip Benchmark. Done.
Ecuting cpuminer. 5 More minutes to wait. ml3l6l6l6l6l6l6l7
Memory Performance: MEMCPY: 2800.5 MB /s
memSet: 6191.5 mb/s (0.2%)
cpuminer Total Scores (5 Minutes Execution): 6.87,6.86,6.85,6.83,6.82,6.79 kH/STML0
OPENSSL Results:
Type 16 bytes 64 bytes 1024 bytes 1024 bytes 8192 bytes 16384 bytes
AES-128-CBC 173609.37K 509936.75k 972013.31k 1264387.07K 1392645.46khtmmm L3
AES-128-CBC 175451.26k 506569.66k 973690.71k 1264628.74k 1382845.10k 1393180.67k
AES-CBC 166539.51k 448796.48k 790104.58k 970846.55k 1040621.57k 1046462.46k
aes-192-cbc 168407.31k 451709.25k 792148.91k 970579.63k 1041061.21k 1046375.08k
aes-256-cbc 159430.38k 412822.74k 676804.10k 809129.64k 857347.41k 861137.58k
aes-256-cbc 162313.43k 412763.39k 677746.94k 809317.38k 857642.33k 861334.19k
Unable To Upload Full Test Results. Please Copy & Paste the Below Stuff to Pastebin.com And
PROVIDE The URL. OTTLING and SWAPING PLEASE.
I started it almost after BOOT, so the DMESG output should be complete (refer to the specific reference. The previous BOOT is loaded), but some information is missing in the script.The full output of the sbc-bench.sh script can be found at at pastebin We see that the "1992" MHz advertised frequency is tested in reality as 1845 MHz, so I feel like some optimization could be done here.
NVMe benchmark
I use iozone 3 for NVMe The SSD was tested 3 times, 1 of which was a 100MB file:
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 100M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
Iozone: Performance Test of File I/O
Version $Revision: 3.489 $
Compiled for 64 bit mode.
Build: linux
Include fsync in write timing
O_DIRECT feature enabled
Auto Mode
File size set to 102400 iozone -e -I -a -s 100M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
Output is in kBytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 kBytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
02400 4 34994 53668 30431 30385 30136 59719
02400 16 102031 130543 80174 80125 79796 133162
02400 512 300692 296328 276975 291837 276681 313464
02400 1024 309822 340026 308900 326826 306102 339059
02400 16384 357975 392544 369753 391219 370336 390004iozone test complete.
and then a 500MB file:
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 500M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
and finally a 1GB file:
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 1000M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
024000 4 35105 58082 30395 30447 27458 60895
024000 16 102421 135279 80210 80314 74596 133579
024000 512 300759 314704 282743 283883 277911 313413
024000 1024 329840 337468 318228 319091 317641 337714
024000 16384 383289 385247 382642 382850 382870 381344
After a series of operations, the results are more or less consistent in all three tests without much change.In the end, I got a read-write speed of about 380MB/s, which was far lower than the original promotional writing and reading speed of SSD, and the results of ODROID-M1. I think this is because the PCIe 2.0 X1 interface is used in this design, not the PCIE Gen 3.0 X2 interface used in Hardkernel board.Below is the output of LSPCI, for reference only:
.5GBE interface configuration and benchmark test
TML3
0
1
2
3
4L6 15
PI@FriendlyElec:/Media/NVME0N1 $ Sudo LSPCI -VTML3
Subsystem: Phison Electronics Corporation Device 5013
FLAGS: Bus Master, Fast Devsel, Laten Cy 0, IRQ 87
MEML6Memory at 380900000 (64-bit, Non-Prefetchable) [SIZE = 16K]
capabilities: [80] e0] e0] XPress Endpoint, MSI 00CAPabilities: [ D0] MSI-X: ENABLE+ Count = 9 Masked-
Capabilities: [E0] ENable-Count = 1/8 Maskable+ capabilities: [F8] Power Management Version 3TML3
CAPABILIES: [100] Lanetcy Tolerance l6capabilities: [110] l1 PM SubstateS
CAPABILIES: [200] Advanced Error Reportingml7
CAPABILITIES: [300] Secondary PCI EXPRER ss
kernel driver in use: nvme
ML6NANOPI R5s router
is configured with ETH0 Gigabit Ethernet "WAN" interface. Therefore, we must manually configure two 2.5GBE ports. I used the same test platform with the same "first part of Friendlywrt Evaluation", namely Ubuntu 20.04 laptop. Then, I connected REALTEK RTL8156BG USB 3.0 to 2.5GBE crypto dog to ETH1, UP XTREME I11 mini PC is connected to ETH2. I did not use the bridge interface like in Friendlywrt, but configured two different subnets: ETH1 was 192.168.2.0, ETH2 was 192.168.3.0.
现在我们在/etc/network/interfaces.d/中创建两个新文件：
- eth1
auto eth1
iface eth1 inet static
address 192.168.2.1
network 192.168.2.0
netmask 255.255.255.0
broadcast 192.168.2.255
- eth2
auto eth2
iface eth2 inet static
address 192.168.3.1
network 192.168 .3.0
netmask 255.255.255.0
broadcast 192.168.3.255
现在安装DHCP服务器
sudo apt install isc-dhcp-server
使用我们的两个子网编辑/etc/dhcp/dhcpd.conf文件：
subnet 192.168.2.0 netmask 255.255.255.0 {
RANGE 192.168.2.100 192.168.2.200;
Option Routers 192.168.2.1;
}
SUBNET 192.168.3.0 Netmask 255.255.255.0 {TML0RANGE 192.168.3.100 192.168.3.200 ;
Option Routers 192.168.3.1;
}
Before restarting the DHCP server:
sudo SystemCTL RESTART ISC-DHCP-Server
At this time, notebook computers and mini PCs should be able to get them from the nanopi R5s on their respective subnets. IP address. Now we can start the benchmark test on the interface.Download iperf3 using eth1 connected to laptop and then from R5S perspective receive:
0
1
2
3
4
5
6$ iperf3 -t 60 -c 192.168.2.1 -i 10
Connecting to host 192.168.2.1, port 5201
[ 5] local 192.168.2.130 port 59822 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 2.28 GBytes 1.96 Gbits/sec 42 1.41 MBytes
[ 5] 10.00-20.00 sec 2.02 GBytes 1.74 Gbits/sec 0 1.61 MBytes
[ 5] 20.00-30.00 sec 1.72 GBytes 1.48 Gbits/sec 0 1.62 MBytes
[ 5] 30.00-40.00 sec 1.87 GBytes 1.61 Gbits/sec 0 1.62 MBytes
[ 5] 40.00-50.00 sec 1.89 GBytes 1.62 Gbits/sec 0 1.70 MBytes
[ 5] 50.00-60.00 sec 2.06 GBytes 1.77 Gbits/sec 21 1.66 MByte /sec 63 sender
[ 5] 0.00-60.04 sec 11.8 GBytes 1.69 Gbits/sec receiver
iperf Done.
This is better than the 1.85 I got in OpenWrt Gbps is slower, and there is some retransmission of content.During the transfer I also use l sbc-bench.sh monitoring system:
0
1
213
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
The system did run at its advertised maximum frequency during testing, and I didn't see any obvious issues here.I also used EthTool to check some information and statistical information: @friendlyelec: ~ $ Sudo ETHTOOL -I ETH1TML3DRIVER: R8125
Version: 9.008.00-napi
FIRMWARE-VERSION: TML3EML6Expans:
BUS-INFO: 0000: 01: 00.0
SUPPPPOTS -StATISTICS: YESTML7SUPPPOS-TEST: no
SUPPPROM-ACCESS: NO
supports-register-dump: YES
SUPPOPPORTS-FLAGS: no
PI@Friendlyelec: ~ Sudo ETHTOOL -S ETH1NIC Statistics:
TX_PACKETS: 451228
RX_PACKETS: 95 69147TX_ERRORS: 0
RX_ERRORS: 0RX_MISSED: 0l3align_error: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 9569102
broadcast: 45
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 31676089
rx_octets: 14506385933
rx_multicast64: 0
tx_unicast64: 451214
tx_broadcast64: 2
tx_multicast64: 12
tx_pause_on: 570
tx_pause_off: 570
tx_pause_all: 1140
tx_deferred: 0
tx_late_collision: 0
tx_all_collision: 0
tx_aborted32: 0
align_errors32: 0
rx_frame_too_long: 0
rx_runt: 0
rx_pause_on: 0
rx_pause_off: 0
rx_pause_all: 0
RX_UNKNOWN_OPCODE: 0
RX_MAC_ERROR: 0htmmmmm l7tx_underrun32: 0
RX_MAC_MISSED: 31
RX_TCAM_DROOPED: 0tdu: 0
rdu: 570
Really got some RX_MAC_MISSED.Now let's test it in reverse:
0
1
2
3
4
5
6
7$ iperf3 -t 60 -c 192.168.2.1 -i 10 -R
Connecting to host 192.168.2.1, port 5201
Reverse mode, remote host 192.168.2.1 is sending
[ 5] local 192.168.2.130 port 59826 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.75 GBytes 1.50 Gbits/sec
[ 5] 10.00-20.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 20.00-30.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 30.00-40.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 40.00-50.00 sec 1.94 GBytes 1.67 Gbits/sec
[ 5] 50.00-60.00 sec 1.94 GBytes 1.67 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.04 sec 11.5 GBytes 1.64 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.5 GBytes 1.64 Gbits/sec receiver
iperf Done.
This looks much better than OpenWrt (1.12Gbps).
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
0
1
2
3
4TML3
6Devkit@UPX -I11: ~ $ iPerf3 -T 60 -C 192.168.3.1 -i10
Connecting T o host 192.168.3.1, port 5201[5] LOCAL 192.168.3.100 Port 37794 Connected to 192.16888 .3.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 10.00-20.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 20.00-30.00 SEC 2.73 GBYTES 2.35 Gbits/SEC 0 1.81 MBytes
[5] 30.00-40.00 SEC 2.73 GBITS/SEC 0 2.90 l3
[5] 40.00-50.00 SEC 2.73 GBYTES 2.35 GBITS/SEC 0 4.37 MBYTES
[5] 50.00-60.00 SEC 2.73 GBYTES 2.35 GBITS/SEC 0 4.37 MBYTES- - - - - - - - - - - --------
Tra on Retr
[5] 0.00-60.00 SEC 16.4 Gbytes 2.35 Gbits/SEC 0 SENDER
[5] 0.00-60.00 SEC 16.4 GBYTES 2.35 GBITS/Secret
IPERF .
Oh, great! This is the first time I have received a 2.35 Gbps transmission speed, so it still looks hopeful! @Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % CPU % SYS % USR % Nice % IRQ TEMP
Unless I get wrong, 25%of IRQ means that a core can be used to handle these.Now let's try:
0
1
2
3
4
5
6
7devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.3.1 -i 10 -R
Connecting to host 192.168.3.1, port 5201
Reverse mode, remote host 192.168.3.1 is sending
[ 5] local 192.168.3.100 port 37800 connected to 192.168.3.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.92 GBytes 1.65 Gbits/sec
[ 5] 10.00-20.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 20.00-30.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 30.00-40.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 40.00-50.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 50.00-60.00 sec 1.84 GBytes 1.58 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.01 sec 11.1 GBytes 1.59 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.1 GBytes 1.59 Gbits/sec receiver
iperf Done.
The result is 1.59 Gbps, not quite perfect, but still better than OpenWrt.
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
^C
The CPU is running at full speed again, and it is far from 100% utilization, so I think the problem should lie elsewhere. Now we can use ethtool again to check the eth2 information and statistics.
0
1$ sudo ethtool -i eth2
driver: r8125
version: 9.008.00-NAPI
firmware-version:
expansion-rom-version:
bus-info: 0001:11:00.0
supports-statistics: yes
supports-test: no
supports-eeprom-access: no
supports-register-dump: yes
supports-priv-flags: no
0
1
2
3
4
5
6
7
8
9
sudo ethtool -S eth2
NIC statistics:
tx_packets: 8506609
rx_packets: 12553353
tx_errors: 0
rx_errors: 0
rx_missed: 0
align_errors: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 12553209
broadcast: 144
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 12543719502
rx_octets: 18471602900
rx_multicast64: 0
tx_unicast64: 8503557
tx_broadcast64: 3035
tx_multicast64: 17
tx_pause_on: 35
tx_pause_off: 35
tx_pause_all: 70
tx_deferred: 0
tx_late_collision: 0
tx_all_collision: 0
tx_aborted32: 0
align_errors32: 0
rx_frame_too_long: 0
rx_runt: 0
rx_pause_on: 0
rx_pause_off: 0
rx_pause_all: 0
rx_unknown_opcode: 0
rx_mac_error: 0
tx_underrun32: 0
rx_mac_missed: 335
rx_tcam_dropped: 0
tdu: 0
rdu: 35
In the test results here, I found that there are more rx_mac_missed. So I guess there should be some tweaks to improve performance.But according to my previous pair of RTL8156B adjustment settings, adjustment settings are really tricky, and experienced people seem to be unanimous on the specific adjustment option. I mainly refer RTL8156/8125 driver Realtek engineers, as well as some online experts in readers.is configured between the two 2.5GBE interfaces nat
Since the 2.5GBE interface cannot reach the best cooperation with IPERF3, I don't worry about testing the performance of the router in Friendlywrt, but there are still a few people asking me. So next I will show how I configure NAT in Ubuntu 20.04, and I will continue to test NAT performance. Remember that it will definitely improve it in a few weeks or months.
is here, we need to enable IP forwarding and NAT. The instruction I use is adapted from a post on NETWORKREVERSE .
编辑/etc/sysctl.conf以启用IP转发（取消注释以下行）：
net.ipv4.ip_forward=1
应用更改：
sudo sysctl -p
现在我们启用NAT：
sudo iptables ! -o lo -t nat -A Postrouting -j Masquerade
We can PING on the laptop of 192.168.2.0 subnet. Treme i11:
jaufranc@CNX-LAPTOP-4: ~ Ping 192.168.3.100
ping 192.168.3.100 (192.168.3.100 ) 56 (84) bytes of data.
SUDO APT Install iPtables-Persistent
SUDO SH -C 'iPtables-save /ETABLES.v4'
I tried IPERF3 between UP XTreme I11 and my laptop.
0
1
2
3
4
5
6
7
8
9
jaufranc@cnx-laptop-4:~$ iperf3 -t 60 -c 192.168.3.100 -i 10
Connecting to host 192.168.3.100, port 5201
[ 5] local 192.168.2.130 port 59430 connected to 192.168.3.100 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 914 MBytes 767 Mbits/sec 355 1011 KBytes
[ 5] 10.00-20.00 sec 912 MBytes 765 Mbits/sec 324 1.23 MBytes
[ 5] 20.00-30.00 sec 917 MBytes 769 Mbits/sec 124 1.09 MBytes
[ 5] 30.00-40.00 sec 915 MBytes 767 Mbits/sec 150 942 KBytes
[ 5] 40.00-50.00 sec 915 MBytes 767 Mbits/sec 78 1.22 MBytes
[ 5] 50.00-60.00 sec 919 MBytes 771 Mbits/sec 64 1.03 MBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.00 sec 5.36 GBytes 768 Mbits/sec 1095 sender
[ 5] 0.00-60.06 sec 5.36 GBytes 767 Mbits/sec receiver
iperf Done.
jaufranc@cnx-laptop-4:~$ iperf3 -t 60 -c 192.168.3.100 -i 10 -R
Connecting to host 192.168.3.100, port 5201
Reverse mode, remote host 192.168.3.100 is sending
[ 5] local 192.168.2.130 port 59434 connected to 192.168.3.100 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.09 GBytes 935 Mbits/sec
[ 5] 10.00-20.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 20.00-30.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 30.00-40.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 40.00-50.00 sec 1.09 GBytes 939 Mbits/sec
[ 5] 50.00-60.00 sec 1.09 GBytes 937 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.05 sec 6.55 GBytes 937 Mbits/sec 973 sender
[ 5] 0.00-60.00 sec 6.55 GBytes 937 Mbits/sec receiver
iperf Done.
has a transmission speed of 768 Mbps in one direction and 937 Mbps in the other direction.@Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % cpu % sys % usr % nice % IRQ TEMP
1: 48: 1992MHz 0.90 25% 0.90 25% 0 % 0% 0% 0% 0% 25% 54.4 ° C
H6
If you use SBC-Bench.sh, the running frequency will be 1992 (Actually 1845 MHz), and 25%IRQ should be just It means that a Core has been completely used to deal with IRQ.is displayed according to the MPSTAT command, which should be processed by Core #0.
l3
TML3H0H0
$ MPSTAT -P All -I Sum
Linux 5.10.66 (F Riendlyelec) 06/05/22 _aarch64_ (4 cpu)
can be confirmed by using TOP and HTOP.
NANOPI
I just have a wall power meter. 6 Free -5.1 W
IPERF3 to ETH1 - 6.3 to 6.7w NAT test -6.2Wht between laptop and mini computers Ml7 It was tested after installing Ubuntu 20.04 and NVMe SSD on this product.
I am still using OpenWrt and without SSD NANOPI R5s for testing:
free L32IPERF3 - 6.0 to 6.2 W Note that because I use wall -mounted power meters, this value will include the power adapter (Khadas Vim4 usb. -C power adapter) efficiency loss. And this value may be higher than the value of using USB-C power timing. This should also be optimized by setting.
The final conclusion
's evaluation today is here. The optimization part should actually include "changing the firmware to make the RockChip kernel run in 1992 MHz", "Adjustment of various settings related to PCIe and Ethernet settings" and so on. Most of them are not very familiar.
(3), THT Linglong M NUC brush X86 soft route evaluation
foreword
many friends to eat X86 soft routes, or they are ready to step into X86 soft routing pit! First of all, it is necessary to clarify whether you are willing to spend time to toss. You can consider buying the finished OPNENWRT router to experience it. You can also buy a router that brushes the Merlin firmware or brush the PANDORABOX (Pandora) firmware to play.prepares related network devices
a suitable 2.5G port switch and X86 soft route with a 2.5G power port, a Type-C end of a 2.5G mesh or external wired network card at USB-A end.
TSMC Linglong M NUC is a mini machine that is very suitable for X86 soft routing. It has two 2.5GE power outlets, two SO-DIMM memory slots and a PCIE 3.0 NVME SSD M.2 slot. It is recommended to buy bare metal, bring up memory and NVMe SSD. Here are two 4GB Micron DDR4 3200MHz memory bar and 1 Intel Optane M10 16GB (Ao Teng Memory).
is installed with memory and NVME SSD.
can be used to use it.
Xike is a new network equipment brand. This is a two -layer management switch with 8 2.5G electric ports and 4 million -dimensional light ports to support port aggregation and VLAN division.
Friends bought this Xike switch to help me upgrade the network of my home, intending to be a 2.5G internal network environment in the whole house.
begins to adjust the device, which is still very simple for experienced people.
can see that the 10G optical port of the XIKE SKS7300-8GPY4XGS switch has been turned up, and the connection status is 10G Full rate. For friends who have 10G SFP modules, they can use 10G LAN. However, at this stage, 2.5G LAN is more suitable for the actual experience needs of the public. If you upgrade to a 10G optical port, the overall hardware cost will be significantly higher several times. For small and medium-sized enterprises, this switch is very suitable for use as an access switch connected to a 10G core switch. It is still more recommended for home users who are in need of 2.5G electrical ports and friends who like to mess with 2.5G LAN to buy this XIKE switch.
For users who have a GPON Stick ONU module, plugging it directly into the SFP interface of this SKS7300-8GPY4XGS switch can save an optical modem. It effectively solves the problem of insufficient accumulation of equipment in the weak current box, and can also improve the space inside the weak current box and reduce the problem of heat accumulation between devices.
What should I do if my computer (desktop and laptop) does not have a 2.5G network port? You can buy a superior USB to RJ45 2.5G wired network card. Connect to the computer through the Type-C end, and the RJ45 network port on the other end has a 2.5GbE (2500Mbps) rate.
Take a USB flash drive to make a boot disk
I generally like to use the tool BalenaEtcher to make a USB flash drive. It is simple and easy to operate! Supports ISO, DMG, IMG, IMG.GZ and other image file formats to create boot disks.
makes the Sea Spider V9 boot disk (flash)
. Go to the Sea Spider official website and find the download area in the service and support. Scroll to the upgrade address at the bottom, find V9.0, and select Download Now next to it to download the Sea Spider V9 firmware. Note that the ISO image file is downloaded.
Sea Spider V9 system
default login address: 192.168.0.1. The account and password are set by yourself after flashing the machine.
You can view the current information overview on the overview. The instrument icon shows CPU usage, memory usage, and disk usage. Below is the monitoring of the real-time total traffic of the intranet, and on the right is some key information. The LAN port already in use by
can identify the network port model: Realtek RTL8125 2.5GbE. The network port icon also identifies 2.5Gb.
Nigou can view the current memory usage in the overview. After flashing and initializing the system, you can see that 10.5% is used by default and 835.86 MB of storage space has been used. Total memory size: 7.6GB (8GB total memory), remaining 6.8GB available.
You can also view the CPU load in the overview. The
interface load can also view the data of the LAN port. The LAN port here is running in full-duplex mode, 2500Mb/s (that is, the 2.5GbE Gigabit port rate).
You can see the motherboard/CP information in the hardware information. Basically, the relevant information displayed is complete.
Here is the 16GB Optane memory, available capacity: 13.41GB, you can check the partitioning of the hard disk storage.
Both 2.5GbE Gigabit network ports can be recognized normally.
Haispider has module management, including Docker container modules.
Basically these modules are very complete for soft routing, including VPN IPsec, PHP suite, IGMP multicast proxy, etc.
FTP service.
file storage and sharing.
Comment: Sea Spider V9 requires purchasing relevant authorizations to fully use all functions. Fortunately, there is a long trial period. Relatively many functions are more suitable for network workers. For ordinary novice users, the operation may be very difficult, and some professional terms are convoluted and difficult to understand.
Make iStore OS boot disk (flash)
Go to KoolCenter official website to download iStore OS firmware. Find the iStore OS directory in the firmware list and enter it.
Considering that the current new X86 soft routing is booted by UEFI, so select X86_64_EFI to download the iStore OS firmware.
Generally has Date (Date) in the bottom list. Select the latest Istore OS firmware.
istore OS
In fact, ISTORE OS is the in -depth customization of Chinese people. Compared with the original OpenWrt, the gameplay (plug -in) and Docker gameplay.
homepage will have a legend with traffic statistics, IP addresses, and network interface status. It looks clear at a glance.
pulls down to see disk information, storage services, docker, download services and remote domain names. The playability is still very high.
can be seen at the bottom to see the system information. The dog recognizes the CPU temperature, CPU usage rate, and memory usage.
network wizard, this is the best direction function for Xiaobai users.
has system information. You can see the computer model, processor architecture and other related information.
can also see the use of memory at the overview.
Soft Route is difficult to master a disk management. Here ISTORE OS directly completes a graphical Diskman disk management, which reduces the difficulty of novice operation and enhances the simpleness of use.
istore OS built -in Docker
is definite It is difficult to install plug -ins. Many beginners or little white users who have just started. The iStore OS directly built -in Docker components for easy deployment of containerization. And supporting CPU and memory configuration settings in advanced settings, which is very user -friendly.
ISTORE OS Unique App Store
For wanting to study the smart home experience, you can consider Home Assistant.
is easy to be suitable for miniature home deployment data service centers. Many tutorials on the Internet will not be discussed in depth.
If you can't afford a dedicated NAS device, in fact, soft routes can also act as simple NAS, but most of the X86 soft routes are not easy to use. There are two tools that can be solved directly here to solve this problem. For friends who want to build an audio -visual center: Jellyfin and NAS Tools. If you can use it, it can definitely easily make the X86 soft routing into a NAS.
ARIA2 downloader. If you can use it, it is also a very practical download tool. For the two broadband at home, in order to avoid the broadband during the work or during the business trip, try to deploy the net Xinyun to earn some benefits.
I actually prefer HOMEBOX inner mesh speed measuring. This is better experienced by the Speedtest speed measuring server. I will show the experience of the HOMEBOX intranet speed measuring experience later.
DISKMAN disk management is definitely a plug -in worth installed. Friends who use QBITTORRENT downloaders will use it to use it. The specific use methods and tutorials have more detailed professional introduction on the tutorial network. If you are interested in the research, go to the study.
KMS server can solve the problem that Windows and Office prompts not activated.
installation plug -in is as simple and convenient as installing the app, which is also the biggest ease of use of Isstore OS.
I use most of these four plug -ins. Among them, multi -line multi -broadcast is suitable for broadband overlay areas. One sentence can run to 600m in one sentence, 500M superposition can run 1000m but do not need to upgrade broadband packages and extra money. It is worthwhile.
KMS server is enabled. Basically, local Windows and Office have no longer encountered unauthorized situations.
This is the HomeBox intranet speed test, which can directly measure the speed of the intranet. I use Teclast Linglong M NUC’s 2.5GbE Gigabit network port, CAT6 (Category 6e) network cable, and Type-C to RJ45 2.5GbE external wired network card, which can achieve a 2500Mbps full-link LAN (intranet) experience .
Summary
Users with basic network engineering skills or a relatively good understanding of the Internet can consider starting X86 soft routing. Sea Spider V9 actually prefers traditional network equipment, combining functions such as switches, routers, and AC controllers. iStore OS is suitable for the general public. It is more playable and easier to get started. I personally recommend flashing the iStore OS firmware to use it, which is basically the Router OS in openwrt that I have used and experienced well.
FriendlyElec seems to focus on Friendlywrt mirror. They told me that it has not been optimized on Friendlycore, so most people may use Friendlywrt because it is easier to configure network and router settings. I saw that Apacer AS2280P4 SSD was actually detected, but it was formatted without the box, so I could only use mkfs.ext4 to format it.(1), Electric Rhino R68S Soft Route Evaluation
I. Appearance and key configuration Analysis of
.1 R68S Key Configuration.2 NAT performanceElectric Rhino R68S RK-3568 ARM Processor, 22nm process, 4 -core Cortex A55 Architecture, main frequency 2GHz, supports the maximum 8G memory;
RK3568 built -in 1 QSGMII, 2 PCIE 3.0x1 channels, also with 2 Gigabit Ether interfaces, so you can use 42.5GB+2x1000MB 4 4 Gigabit Ether interface. Net port combination.
R68S core configuration is shown in the figure below:
HTM core configuration CPU is the memory and html. 6 flash memory chip .
LPDDR4 memory chip is from Rayson, the model is RS1G32LF4D2BDS- 53BT, the capacity is 4GB (2GB version is 70 yuan); Resee (under Shenzhen Jiang Bolong), the model is NCEMASKGG-16G, the capacity is 16GB;
There is a Gigabit Ethernet exchange chip in the upper left corner of the processor with a model of RTL8211F. When you see this crab logo, you know that from Riyu ;
does n’t there be 4 mesh ports? What about the remaining 3 exchanges?
is on the back of the motherboard.
Two 2.5GB exchange chips also comes from Riyu, the model is RTL8125BG:
motherboard
Ruyu's 8125B, Intel 's I225 B3 and Qualcomm QCA8081 are currently the most common.2 R68s Appearance.1 encryption and decryption performanceengineering machine is sent, there is only one piece of bare plate.
is because I want to do the temperature test in weak electric box , so the boss sent a shell after a few days (for test model, non -final version).
I received is made of plastic material (also CNC aluminum shell, the price will be about 100 yuan), and there are a lot of heat dissipation holes.
official version shell will be made into a frosted effect
HT There is no dust -proof cotton, so if you use R68S to put the desktop, you still need to consider the problem of dust. After all, dust is the first of electronic equipment. One killer:
front
HTM has a large amount of heat dissipation holes in the two sides of the 2 side:
machine side HT
If you will put R68s in a weak electric box like me, then the hollow design of the shell is actually very beneficial to heat dissipation.
But if you put R68S on the desktop, I also recommend buying a CNC aluminum shell, that is, dustproof and waterproof, heat dissipation is okay.
cnc shell is about length. Single version.
I actually tested it, and the 12V 1A light cat power adapter can also work normally.
Of course, if you have the need to connect to external hard disks, I suggest that it is best to put a 12V 1.5A power adapter.
When it comes to external hard disks, the USB 3.0 performance of R68S is still very strong. Reading and writing can run to about 120MB. If you want to watch movies on the R68S, you don't have to worry about the original disk high -code movie will be stuck.
2. Performance and stability Test
For a soft routing , performance and stability are equally important.
directly look at the table:
Zhihu@仙鱼
CPU CoreMARK
AES...
CHA...
AES hard decoding
R68S (RK3568)
03101
is
N1(S905D)
3
8704
is
2 R4S (RK3399)
36089
is
J1900
no
J4125
51102is
The encryption and decryption performance of R68S is much stronger than the N1 of ARM architecture , but it lags behind the 6-core R4S (RK3399) by about 20%. The CPU CoreMARK of
R68S is about 6000 points behind J1900. However, J1900 does not support hardware AES decoding after all, and R68S still has some advantages in this regard. If
is compared with J4125, which also supports AES hard decoding, R68S lags behind by more than half.
Although the R68S is not excellent in terms of scores, in actual use, the performance is more than sufficient.
NAT performance mainly tests the broadband speed. The tested topology is as follows:
NAT test
The test rate is as follows:
It can be seen that if the transmission loss is removed, the R68S has reached the limit of Gigabit broadband.
At this time, the CPU load is around 45%.
.3 LAN forwarding performanceLAN forwarding capability mainly tests the forwarding rate between the two 2.5G LAN ports of the R68S. I did a total of 3 different tests. The topologies of
R68S and iPerf3 client installed on the computer respectively, ThinkPAD X13 (2.5G USB 3.0 network card) is directly connected to the 2.5G network port of R68S. The test results are:
iPerf3
The downlink rate is close to 2.1Gb, which may be limited by the performance of the USB network card. The uplink rate is slightly worse at 1.94Gb.
At this time, the CPU load is about 30%.
LAN TO The downlink rate of LAN
can reach around 2.3Gb, which is close to the limit, but the uplink rate still lags behind. In multiple tests, the fastest rate only reached 1.85Gb.
At the same time, I also tried large file transfer: the reading rate of
can be stable at 210MB, and the peak value can reach 242MB.
However, the writing speed is relatively unstable and fluctuates around 150MB. This may be due to a bottleneck in the hard disk speed, or the network card uplink is inherently slow.
LAN file transfer speed
Thinkpad The X13's AX200 network card can achieve a 2402Mb negotiation rate under the Redmi AX5400 e-sports router, and the laptop uses WIFI most of the time, so the LAN TO WIFI6 test is also very necessary:
LAN TO WIFI6
AX5400 is directly connected to the NAS. The maximum rate that can be reached by using WIFI6 is 1850/624. The rate drops slightly after being forwarded by R68S, but the drop is only 5%, which is acceptable.
Overall, the LAN forwarding capability of R68S is quite good. If you don’t have many 2.5G devices, you can still save a 2.5G switch by using R68S.
However, if you install docker on R68S, the LAN bridge forwarding performance will drop significantly. You must add an switch :
.4 Temperature and StabilityThe firmware of R68S is designed by the master lean, and it only lasts 10 days. Judging from the situation, the stability is still very good.
running time
R68S I have been using it in a weak current box.
The current indoor temperature in Zhejiang is about 27 degrees, and the standby temperature of the R68S bare board is about 47 degrees. After the
case arrived, I observed that the standby temperature soared to about 53 degrees.
However, it’s not just the R68S that’s rising in temperature. In the past two days, the casing of my Redmi AX5400 router has been very hot even if it is placed on the TV cabinet.
When the temperature is high, the indoor temperature may be around 31 degrees, and the temperature of the weak current box will be even higher. I will continue to update the temperature monitoring of R68S.
Zhejiang has been experiencing high temperatures for more than 40 days, and the indoor temperature is basically around 32 degrees.
The temperature of R68S in the weak current box is basically around 58 degrees, and it is very stable in use, with no crashes or interruptions.
3. Tutorial on playing with the machine
Here we mainly talk about the tutorial on flashing the machine and installing some plug-ins
.1 Resetting and flashing the machineR68S has a reset button. If the system is confused, or you cannot enter the router settings page, then press and hold the reset button. It is very convenient to restore the factory settings (when plugged in) by pressing the button for 5 seconds.
In addition, you need to flash the R68S, which can be completed in a few seconds. Compared with the N1 flash machine, it is really a world of difference.
However, convenience is convenience, and you still need to learn the tutorial.
) Install Rockchip micro-driver and flash toollink: https://pan.baidu.com/s/1pTxQ1yfjYTccClHcy068aw?pwd=6tnl Extraction code: 6tnl
firmware selects
Flashing sequence
You will hear a prompt on the computer in about two seconds, and the development tool will also prompt that the device has been found.
Erase Except Flash
. Flashing
.2 WEB upgradeOf course, only when a major version is upgraded or the internal partition is changed, you need to use to brush . For ordinary updates, just use WEB upgrade:
WEB upgrade
WEB upgrade requires the "sysupgrade" image, and the file suffix name is general It's all tar.
.2 plug -in installationmachine is built -in pure version of the firmware. If you need to use certain firmware, you can upload IPK or SH command installation.
. For example, the Alibaba Cloud Plate WEBDAV plug -in that the audio and video player is very needed is not built in the factory. We can use the SSH tool or OPENWRT Tyyd to enter the command installation (replicate the car one by one):
wget https://github.com/messense /Aliyundrive-BDAV/releases/download/v1.3.2/aliyundrive-webdav_1.3.2-1_Aarch64_gneric.ipkwget https://github.com/messense /Aliyundrive-Webdav/releases/download/v1.3.2/luci- app-aliyundrive- webdav_1.3.2_all.ipkwget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/luci-i18n -webdav-zh- cn_1.3.2-1_all.ipkopkg Install Aliyyundrive-Webdav_1. 3.2-1_AARCH64_GENERIC.IPKOPKGG Install Luci-ALIYUNDrive-Webdav_1.3.2_ALL.IPKOPKG Install Luci-I18n-WEBDAV-ZH-CN_1.3.2-1_ All.ipkis refreshed after installation is complete, and the service has plug -in:
Alibaba Cloud Plate4. Summary and purchase suggestions
Electric Rhino R68S soft routing, which is very suitable for me.
is because I already have the finished NAS with a 2.5G network port. The router usually uses an AP mode, so for me, you only need to have a soft route with 2.5G mesh.
The current soft routing market is very rolled, and the 2.5G soft route of J4125 has also reduced the price to about 600 yuan. If you have more needs, such as love fast diversion+OpenWRT bypass+UNRAID's standard All in one gameplay, then R68S may be possible Can't meet your requirements.
But if you just need a pure soft route, then R68S is still very suitable for you.
(2), NANOPI R5s soft routing evaluation
NANOPI R5s
with metal shell ssd and sanoPi R R 5S
I found that the router has been assembled. They also accompanied 6 rubber foot pads and a section of
NANOPI R5STML12
NANOPI is there. With a USB-C port for power supply, a WIFI antenna hole (this antenna hole can also insert GPIO, UART CONSOLE and other cables), two 2.5 2.5 GBE RJ45 LAN port, a Gigabit Ethernet WAN port and HDMI video output .
NANOPI R5s cover key
We will find one for solidarity The cover button for upgrade, four on the front panel for LED lights for "system" and Ethernet ports, and two USB 3.0 port.NANOPI R5s Disassemble
. Generally, the main reason for disassembling is the following:: Out of curiosity, to install M.2 NVMe, to welded SPI flash memory, you need to connect some GPIO, RTC battery, or adjust the UART to TTL to TTL Test plate. If you want to open it, it is easy to open. You can easily disassemble the four screws that need to be released.
can see the bottom of the circuit board with the M.2 KEY M slot, the SPI flash memory occupation space (right), and the Samsung KLM8G1GETF-B041 EMMC 5.1 flash memory 8GB.
NANOPI R5S SPI flash memory and M.2 socket
need to remove the four screws before taking the circuit board from the shell.
NANOPI R5S Sbc
R6RAYSON RS512LM4 D2BDS is a 2GB LPDDR4X memory chip. RTL8211F (GBE) and
NanoPi R5S with a metal casing that acts as a CPU cooler.
Test setup using 2.5GbE USB dongle and UP Xtreme i11 mini PC
You may remember that I had some performance issues with not too long ago when using the RTL8156B USB dongle , but now this issue has been resolved. Because Realtek sent me another RTL8156BG, I was previously using iperf3 on my laptop from UP Xtreme When i11 mini PC transmits data and performs full duplex test, its test result is 2.34Gbps/2.29Gbps.
I am now using the same settings for testing, but this time the middle stuff is replaced with NanoPi R5S.
NanoPi R5S
TP-Link switch here is only used as a Gigabit Ethernet switch , and the NanoPi is connected through Xiaomi AX6000 router The R5S router's WAN port connects to the internet, which prevents me from having to install some software packages on the device. While placing the NanoPi R5S on top of the TP-Link switch resulted in better photos, both devices get very hot and I don't recommend this. Since my Ethernet cable was very short, I had to move the router to a table to test it. The
OpenWrt and iperf3 benchmarks
FriendlyWrt come pre-installed on the router, so they are ready to use right out of the box. It is also possible to immediately access the LuCI interface or SSH using "root" as user and "password" as password. This is indeed very convenient, but it is not very safe, and may be against the law in some countries. Anyway, it is best to change the password at least the first time you use it. The status of
FriendlyWrt
FriendlyWrt is based on the kernel of OpenWrt 22.03.0-rc1 and Linux 5.10.66. It uses less than 250MB of RAM when idle and using default settings, which is more than enough since the system comes with 2GB of RAM.
FriendlyWrt network
I haven't connected SSD yet, so I only mounted the root partition. The available memory is 6.7 GB, and the root partition uses 920 KB. All interfaces can correctly obtain IP addresses through DHCP when starting up. Devices on the LAN can also be accessed through hostname.lan and obtain some IPv6 addresses.
Start from the IPERF3 benchmark test. I first run the "iPerf3 -S" on the NANOPI R5S and run the following commands on the laptop:
download: (NANO NANO NANO View the results of the RX test on PIR5S) iperf3 -T 60 -C 192.168.2.1 -i 10Connecting to host 192.168.2.1, port 5201[ 5] local 192.168.2.130 port 48782 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate Retr Cwnd[ 5] 0.00-10.00 sec 2.23 GBytes 1.92 Gbits/sec 0 1.69 MBYTES [5] 10.00-20.00 SEC 2.02 GBYTES 1.74 Gbites/SEC 0 1.69 MBYTES [5] 20.00-30.00 SEC 2.33 GBITS/SEC 0 2.64 MBYTES [5] 30.00-40.00 SEC 1.66 ES 1.42 GBITS/SEC 0 2.64 MBYTES [ 5] 40.00-50.00 sec 2.62 GBytes 2.25 Gbits/sec 0 2.64 MBytes[ 5] 50.00-60.00 sec 2.01 GBytes 1.73 Gbits/sec 0 2.64 MBytes- - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr [5] 0.00-60.00 SEC 12.9 GBITES 1.84 GBITS/SEC 0 Sender [5] 0.00-60.05 SEC 12.9 Gbites/Secits/Secits E.
- (on NANOPIR5S Check the results of the TX test):
$ iPerf3 -T 60 -C 192.168.2.1 -RConnecting to host 192.168.2.1, Port 5201Reverse Mode, Remote Host 192. 168.2.1 is sending [5] LOCAL 192.168.2.130 Port 48786 Connect TO 192.168.2.1 port 5201 [ID] Interval Transfer Bitrate [5] 0.00-10.00 SEC 1.29 GBYTES 1.11 GBITS/Secain [5] 10.00-20.00 SEC 1.31 GBITS/SEC [5] 20.00-30 .00 SEC 1.33 GBYTES 1.14 GBITS/ sec[ 5] 30.00-40.00 sec 1.27 GBytes 1.09 Gbits/sec[ 5] 40.00-50.00 sec 1.30 GBytes 1.12 Gbits/sec[ 5] 50.00-60.00 sec 1.30 GBytes 1.12 Gbits/sec- - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr [5] 0.00-60.06 SEC 7.80 GBYTES 1.12 GBITS/Secits [5] 0.00-60.00 SEC 7.80 GBITS/SEC Receiverf Done.
is not exactly the above 2.35 GBPS and 1.85 Mbps when it is propagated by Friendlyelec, because in this configuration, I see 1.84 Gbps and 1.12 Gbps. When viewing the 10 -second transmission test report, RX has also changed. Fortunately, no problem was found.Now we try the same with another WAN port on NanoPi R5S and run the command from UP Xtreme i11:
- Download (Rx):
devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.2.1 -i 10Connecting to host 192.168.2.1, port 5201[ 5] local 192.168.2.207 port 52052 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate Retr Cwnd[ 5] 0.00-10.00 sec 2.59 GBytes 2.22 Gbits/sec 0 1.67 MBytes[ 5] 10.00-20.00 sec 2.62 GBytes 2.25 Gbits/sec 0 1.93 MBytes[ 5] 20.00-30.00 sec 2.60 GBytes 2.24 Gbits/sec 0 1.93 MBytes[ 5] 30.00-40.00 sec 2.47 GBytes[ 5] 50.00-60.00 sec 2.45 GBytes 2.10 Gbits/sec 0 4.90 MBytes- - - - - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr[ 5] 0.00-60.00 sec 15.2 GBytes 2.17 Gbits/sec 0 sender[5] 0.00-60.00 sec 15.1 GBytes 2.17 Gbits/sec receiveriperf Done.
- upload (Tx);
devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.2.1 -i 10 -RConnecting to host 192.168.2.1, port 5201Reverse mode, remote host 192.168.2.1 is sending[ 5] local 192.168.2.207 port 52056 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate[ 5] 0.00-10.00 sec 1.31 GBytes 1.13 Gbits/sec[ 5] 10.00-20.00 sec 1.29 GBytes 1.11 Gbits/sec[ 5] 20.00-30.00 sec 1.32 GBytes 1.14 Gbits/sec[ 5] 30.00-40.00 sec 1.30 GBytes 1.11 Gbits/sec[ 5] 40.00-50.00 sec 1.33 GBytes 1.14 Gbits/sec[ 5] 50.00-60.00 sec 1.27 GBytes 1.09 Gbits/sec- - - - - - - - - - - - - - - - - - - - - - - - - -[ID] Interval Transfer Bitrate Retr[5] 0.00-60.00 sec 7.82 GBytes 1.12 Gbits/sec 0 sender[5] 0.00-60.00 sec 7.82 GBytes 1.12 Gbits/sec receiveriperf Done.when downloading (Rx) at 2.17 The Gbps operation seems to be better; but when uploading (Tx), it is still running very slow at 1.12 Gbps.
Let's take a look at the speed of using two WAN ports at the same time? One is to run "iPerf3 -S" on the UP Xtreme I11, and the other port runs the following commands on the notebook:
TML0TML7TML3
8
TML6
0htmm l7
1
2
3
4
5
6jaufranc@cnx -laptop -4: ~ iPerf3 -T 60 -C 192.168.2.207 -i 10htmlml 7
Connecting to host 192.168.2.207, Port 5201
[5] Local 192.168.2.130 Port 49762 Connect to 192.168.2.207 Port 5201
[ID] Interval Transfer Bitrate Retr CWNDTML3
[5] 0.00-10.00 SEC 1.39 GBITS/SEC 0 3.15 MBYTESTML3
[5] 10.00 -20.00 SEC 1.42 GBITS/SEC 0 3.15 MBYTES
[5] 20.00-30.00 SEC 1.44 GBITS/SEC 0 3.15 MBYTES
[5] 30.00-40.00 SEC 1.42 GBITES 1.22 GBITS/SEC 0 3.15 MBYTES
[5] 40.00-50.00.00 SEC 1.42 GBYTES 1.22 GBITS/SEC 0 3.15 MBYTES
[5] 50.00-60.00 SEC 1.39 GBITS/SEC 0 3.15 l3
[5] 0.00-60.00 SEC 8.49 GBITS/Secits/SEC 0 SENDER- -------------------- -
[id] interval transfer bitrate retr
[5] 0.00-60.05 SEC 8.48 GBYTES 1.21 GBITS/Sec itiverTML3IPERF 4
This data should be within the expected range It is because we have obtained the (TX) data previously lower; when we 1.21 Gbps, it is still a little higher than the 1.12 GBPS when we only use the upload (TX).We tried again:
$ iPerf3 -T 60 -C 192.168.207 -i 10 -RConnecting 192.168.2.207, por T 5201reverse Mode, Remote Host 192.168.2.207 is sending [5] Local 192.168.2.130 Port 49766 Connected to 192.68.2.207 Port 5201 [ID] Interval Transfer Bitrate [5] 0.00-10.00 SEC 2.04 GBITS/SEC [5] 10.00-20.00 SEC 2.04 GBITS/SE C [5] 20.00-30.00 SEC 2.04 GBYTES 1.75 GBITS /sec[ 5] 30.00-40.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 40.00-50.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 50.00-60.00 sec 2.05 GBytes 1.76 Gbits/sec- - - - - - - - - - - - - - - - - - - --- [ID] Interval Transfer Bitrate Retr [5] 0.00-60.05 SEC 12.3 GBITES 1.75 GBITS/SEC 0 SENDER [5] 0.00-60.00 SEC 12.3 GBITS/GBITS/ SeciverIperf Done.
is 1.75 Gbps this time. I don't know what happened or how this happened.followed me with a full dual work, testing the CPU, memory and temperature reported in LUCI.
NanoPi R5S的CPU负载
由上可看到其CPU负载介于2.0和2.5之间。
NANOPI R5s The temperature
CPU #0 is relatively low, CPU #1 to #3 Occupy most resources.In a room with an ambient temperature of 28 ° C, its temperature has never exceeded 60 ° C.
Friendlywrt use rate TML12
does Change. The performance of the full dual -work period is: 1.66 Gbps and 736 Mbps.
This result is a bit disappointing, I now plan to switch to FriendlyCore (Ubuntu Core) to check if I will get similar results, and then perform further tests. If you want me to review other aspects of OpenWrt, you can tell me in the comment area.
took the test a step further and switched to the FriendlyCore image based on Ubuntu 20.04 because I was more familiar with Debian-based operating systems and some tools did not run on OpenWrt. Note that the performance shown in the test results is still not very ideal, which is why I call it a pre-test. As more people use and adjust the software, things should improve in the next few months.
OpenWrt optimization?
Let’s not discuss Ubuntu-based images for now. Because FriendElec (Guangzhou Friendly Electronic Technology Co., Ltd.) told me that they added some optimizations, so I upgraded to the newer version of FriendlyWrt and tested this image:
This is what they said: " We have made some optimizations to the new image, such as network card interrupt settings, uninstall support, etc. ” So I downloaded it on Google. "rk3568-eflasher-friendlywrt-20220526.img.gz" found on Drive, then use USBImager to burn it to the microSD card, and then boot into the router.
After doing this, it will automatically burn the image into the eMMC flash memory. If you connect a monitor, you can follow the results. After the operation is completed, remove the microSD card and restart the router.
Once this is done, you can check the status by connecting an HDMI display (shown above) or by looking at the LEDs on the device. The process is very fast and installation to the eMMC flash memory only takes a few seconds.
This new version of the image is mainly , which has made changes to the 40-net-smp-affinity file.In the previously pre -installed Friendly, it looks like this:
0
1H7H7
2
3FriendlyElec, Nanopi-R5s)
set_interface_core 8 "ETH0"
ECHO 7/SYS/Class/NET/ETH0/Queues/RPS_CPS
SET_IN Terface_core 2 "ETH1-0"Set_interface_core 4 "ETH1-16"
Set_interface_core 4 "ETH1-18"
EML6E Cho B/SYS/Class/NET/ETH1/Queues/RX-0/RPS_CPushml7
SET_INTERFACE_CORE 4 "ETH2-0"
Set_interf ACE_CORE 2 "ETH2-16"
SET_INTERFACE_CORE 2 "ETH2- 18"
echo 9 /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
而新的
0
1
2
3friendlyelec,nanopi-r5s)
set_interface_core 8 "eth0"
echo f /SYS/Class/Net/Eth0/Queues/RX-0/RPS_CPUS
SET_INTERFACE_COR Core 4 "ETH1-16" Hilml7
SET_INTERFACE_COR -0/rps_cpus
set_interface_core 2 "eth2-0"
set_interface_core 2 "eth2-16"
set_interface_core 2 "eth2-18"
echo d /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
Willy Tarreau也解释了Reasons for changing the ETH1 interface:
It involves rps. They received this IRQ (interrupt request) on Core 2 and re -assigned the flow to Core 0, Core 1, and Core 3. This is the correct method of using RPS. However, to reach this point, you must manually distribute a network performance testing tool iPerf and observe the first saturated Core. If you first use KSOFTIRQD to saturate Core 2, then make sure that iPerf can run on any of the other three cores. If Core2 is slightly idle, try to set an iPerf on it. If you put the iPerf on it will pop up KSoftirqd, then they will hinder each other, and users will be more willing to change the RPS settings to help release another Core and use it for IPERF.
Before testing and switching to Ubuntu, I actually didn't try this method.When I try to use the new FriendlyWrt image, I get even worse results:
0
1
2
3
4
5
6
7
8
9
$ iperf3 -t 60 -c 192.168.2.1 -i 10
Connecting to host 192.168.2.1, port 5201
[ 5] local 192.168.2.130 port 49590 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 1.92 GBytes 1.65 Gbits/sec 0 1.62 MBytes
[ 5] 10.00-20.00 sec 1.91 GBytes 1.64 Gbits/sec 10 2.34 MBytes
[ 5] 20.00-30.00 sec 1.90 GBytes 1.63 Gbits/sec 0 2.61 MBytes
[ 5] 30.00-40.00 sec 1.85 GBytes 1.59 Gbits/sec 4 1.30 MBytes
[ 5] 40.00-50.00 sec 1.88 GBytes 1.61 Gbits/sec 1 1.06 MBytes
[ 5] 50.00-60.00 sec 1.76 GBytes 1.51 Gbits/sec 2 868 KByte /sec 17 sender
[ 5] 0.00-60.05 sec 11.2 GBytes 1.60 Gbits/sec receiver
iperf Done.
$ iperf3 -t 60 -c 192.168.2.1 -i 10 -R
Connecting to host 192.168.2.1, port 5201
Reverse mode, remote host 192.168.2.1 is sending
[ 5] local 192.168.2.130 port 49594 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.22 GBytes 1.05 Gbits/sec
[ 5] 10.00-20.00 sec 1.36 GBytes 1.17 Gbits/sec
[ 5] 20.00-30.00 sec 1.31 GBytes 1.12 Gbits/sec
[ 5] 30.00-40.00 sec 1.46 GBytes 1.26 Gbits/sec
[ 5] 40.00-50.00 sec 1.47 GBytes 1.26 Gbits/sec
[ 5] 50.00-60.00 sec 1.46 GBytes 1.26 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.05 sec 8.29 GBytes 1.19 Gbits/sec 1 sender
[ 5] 0.00-60.00 sec 8.29 GBytes 1.19 Gbits/sec receiver
iperf Done.
Therefore, this issue has to be re-examined.M.2 NVMe SSD installation in NanoPi R5S
I purchased an APACER AS2280 (AP256GAS2280P4-1) PCIe Gen 3.0 x4 SSD a while ago and it can achieve sequential read speeds of up to 1,800 MB/s and up to 1,100 on the right hardware Sequential write speed in MB/s.
Apacer M.2 2280 PCIe SSD
The installation process was easy as I only had to loosen four screws to remove the bottom cover, install the SSD and then secure it in place with the screws they provided.
NANOPI R5s Installation
Riendlycore
I first tried to install Friendlycore with EFLASHER mirror.
use EFLASHER mirror to install FriendlyCore
I started the router, but later I noticed that the WAN interface link was not displayed on the TP-Link switch. Only the power supply LED lit up, and the power supply LED light up Mirror is normal. I tried again and settled through the EFLASHER UI settings by clicking "to complete", but still not.
eflasher Install F Riendlycore
Start directly from the Microsd card and run the operating system from there. After doing so, it is normal to run. However, if you intend to use the NANOPI R5s for a variety of uses and look forward to using the desktop environment in Ubuntu 20.04 mirrors, it may be disappointed because HDMI output can only be used to access the terminal.
UBuntu 20.04 FRIENDLELEC login TML3
Friendlycore system information
You can find the startup log on CNX Software Pastebin.我使用pi/pi凭据（用户名/密码）通过过了SSH登录，并将系统升级到了最新软件包：
sudo apt update
sudo apt dist-upgrade
现在，我们运行一些命令来获取系统信息：@FriendlyELEC:~$ cat /etc/lsb-releaseDistRib_id=ubuntu
Distrib_release=20.04
h6h6distrib_codename = Focal
DistRib_Descripting = "Ubuntu 20.04.4 LTS"
PI@Friendlye LEC: ~ $ Uname -ALinux Friendlyelec 5.10.66 #219 SMP PRIMPT FRI APR 22 18:20:21 cst 2022 AARCH64 aarch64 aarch64 GNU/Linux
pi@FriendlyELEC:~$ free -mh
total used free shared buff/cache available
Mem: 1.9Gi 150Mi 1.7Gi 3.0Mi 114Mi 1.7Gi
Swap: 0B 0B 0B
pi@FriendlyELEC:~$ df -mh
Filesystem Size Used Avail Use% Mounted on
udev 969M 0 969M 0% /dev
tmpfs 197M 480K 196M 1% /run
overlay 27G 1013M 26G 4% /
tmpfs 981M 0 981M 0% /dev/shm
tmpfs 5.0M 4.0K 5.0M 1% /run/lock
tmpfs 981M 0 981M 0% /sys/ fs/cgroup
tmpfs 197M 0 197M 0% /run/user/1000
除了NVMe驱动器没有自动挂载，一切看起来都还不错。Now we find more details with INXI: @Friendlyelec: ~ $ inxi -fc0SYSYSTEM: TML3HOST: Friendlyelec Kernel: 5.10.66 AARCH 64 Bits: 64 Console: TTY 0
Distro: Ubuntu 20.04.4 LTS (FOCAL)
MACHINE:
Type: ARM Device System: Friendly NANOPI R5s Details: N/A
Serial: 8CBFE79E107C459C
BATRY:
html-1: test_battery charge: 100% Condition: n/A
CPU:
toPology: Quad core A variant: cortex-a55 bits: 64 Type: mcpSpeed: 408 MHz Min/Max: 408/1992 MHz Core Speeds (MHz): 1: 1992 2: 1992Device-: Display-Subsystem Driver: RockChip_drm V: N/A
Device-2: Mali-Bifrost Driver: Mali v: n/aDevice-3: RK3568-DW-HDMI Driver: dwhdmi_rockChip v: n/a
Display: Server: x. x. x. ORG 1.20.8 Driver: DWHDMI_RockChip Ty: 80x24TML3MESSAGE: Advanced Graphics Data UnavaiLe. Display
AUDIO:
Device-1: RK3568-DW-HDMI Driver: DWHDMI_RockChip
Device-2: Simple-Audio-Card Driver: Asoc_simple_card
Device-3: Simple-Audio-Card Driver: n/A
ML0 Device-4: Simple-Audio-Card Driver: ASOC_SIMPLE_CARD
Sound Server : ALSA V: K5.10.66
Network:
Device-1: Realtek RTL8125 2.5GBE Driver: ml0if: ETH1 State: DOWN MAC: E2: 1D: 62: A1: 1A: Ca
Device-2: Realtek RTL8125 2.5GBE Driver: R8125
IF: ETH1 State: DOWN MAC: E2: 1D: 62: A1: 1A: Ca
Device-3: RK3568-GMAC Driver: RK_GMAC_DWMACI6F-ID -1: ETH0 State: Up Speed: 1000 Mbps Duplex: FullTML3MAC: 1d: 62: A1: 1A: Ca
IF-ID-2: ETH2 State: DOWN MAC: 12: BF: 2B: D6: 4B: E0
DRIVES:H6H6H6H6H6H6hl6LOCAL Storage: Total: 274.88 GIB USED: 1012.3 mib (0.4%)
id-: /DEV /MMCBLK0 Model : SD16G SIZE: 2912 Gib
ID-2: /DEV /MMCBLK2 Model: 8GTF4R SIZE: 7.28 gib
id-3: /dev /nvme00 n1 Vendor: Apacer Model: AS2280P4 256GB SIZE: 238.47 Gib
PARTION: TML3
id-: 26.48 Gib use d: 1012.3 MIB (3.7%) FS: Overlay
SOURCE: ERR-102
TML3Sensors: SYSTEM Te Mperatures: CPU: 46.1 C Mobo: N/AFAN Speeds (RPM): N/A
Il6Info:PRO sets: 130 uptime: 7M Memory: 1.92 Gib use: 211.4 mib ( 10.8%)
Init: systemd Shell: bash inxi: 3.0.38
只有eth0 WAN端口打开了，eth1/eth2 2.5GbE端口是关闭的，根本没有显示配置。FriendlyElec seems to focus on Friendlywrt mirror. They told me that it has not been optimized on Friendlycore, so most people may use Friendlywrt because it is easier to configure network and router settings. I saw that Apacer AS2280P4 SSD was actually detected, but it was formatted without the box, so I could only use mkfs.ext4 to format it.NanoPi R5S基准测试
现在我们通过在路由器上运行SBC Bench的方式来对CPU进行基准测试，希望尽可能可以发现一些问题：$ sudo /bin/bash ./sbc-bench.sh -c
[sudo] password for Pi:
WARNING: DMESG OUTPUT Does Not Contain Early Boot Messages Whitml7
Help in IDentifying Re Details.IT is Recoming to Reboot Now and then Execute The Benchmarks.
Press [Ctrl]- [C] To Stop or [Enter] to Continue.
AVERAGE LOAD and/or CPU Utilization too high (too much background activity). Waiting ...
too busy for benchmarking: 07:21:06 up 3 min, 1 user, LOA, LOA, LOA, LOA, LOA, LOA D Average: 0.41, 0.27, 0.11, CPU: 3%
TOO Busy For BenchMarking: 07:21:11 up 3 min, 1 user, load average: 0.38, 0.26, 0.11, CPU: 1%
TML3TOO BUSYMARKING: 07:21:16 up 3 min, 1 user, load average: 0.35, 0.26, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:21 up 3 min, 1 user, load average: 0.32, 0.25, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:26 up 3 min, 1 user, load average: 0.29, 0.25, 0.11, CPU: 1%
too Busy for Bundmarking: 07:21:31 up 3 min, 1 user, load average: 0.27, 0. 24, 0.10, CPU: 1%
SBC- Bench v0.9.7
Installing Needed Tools. This may take some time. Done.checking cpufreq opp. DONE (Results Will Be Available in 20-28 Minutes) .ECUTING TINYMEMBENCH. Done.
Ecuting Ram Lam ATENCY TESTER. Done.ECUTING Openssl Benchmark. Done. .executing 7-zip Benchmark. Done.
Ecuting cpuminer. 5 More minutes to wait. ml3l6l6l6l6l6l6l7
Memory Performance: MEMCPY: 2800.5 MB /s
memSet: 6191.5 mb/s (0.2%)
cpuminer Total Scores (5 Minutes Execution): 6.87,6.86,6.85,6.83,6.82,6.79 kH/STML0
OPENSSL Results:
Type 16 bytes 64 bytes 1024 bytes 1024 bytes 8192 bytes 16384 bytes
AES-128-CBC 173609.37K 509936.75k 972013.31k 1264387.07K 1392645.46khtmmm L3
AES-128-CBC 175451.26k 506569.66k 973690.71k 1264628.74k 1382845.10k 1393180.67k
AES-CBC 166539.51k 448796.48k 790104.58k 970846.55k 1040621.57k 1046462.46k
aes-192-cbc 168407.31k 451709.25k 792148.91k 970579.63k 1041061.21k 1046375.08k
aes-256-cbc 159430.38k 412822.74k 676804.10k 809129.64k 857347.41k 861137.58k
aes-256-cbc 162313.43k 412763.39k 677746.94k 809317.38k 857642.33k 861334.19k
Unable To Upload Full Test Results. Please Copy & Paste the Below Stuff to Pastebin.com And
PROVIDE The URL. OTTLING and SWAPING PLEASE.
I started it almost after BOOT, so the DMESG output should be complete (refer to the specific reference. The previous BOOT is loaded), but some information is missing in the script.sbc-bench.sh脚本的完整输出可以在pastebin上找到，我们看到“1992”MHz广告频率在现实中被测试为1845 MHz，因此我觉得可以在这里进行一些优化。
NVME reference test
. Three tests for NVMe SSD, one of which is 100MB of files: @friendlyelec:/Media/NVMe0n1 $ Sudo IOZONE -E -A -A -R 4K -R 16K -R 512K -R 1024K -R 16384k -i 0 -i 1 -i 2
IOZONE: Performance Test of File I/O
Vision $ Revision: 3.489 $
Compiled for 64 bit mode.
build: linuxinnClude fsync in write timingml7
ml0 O_Direct Feature Enabled
Auto Mode
File SIZE SET TO 102400 0RCORD SIZE 4 KBRCON SIZE 16 KB
RECORD SIZE 512 KBRCORD SIZE 1024 KB
RCORD SIZE 16384 KBTML0Command Line Used: IOZONE -E -I -A -S 100M -R 4k -R 16K -R 512K -R 1024K -R 16384K -i 1 -i 1 -I 2
OUTPUTPut is in KBYTES/ SECTime Resolution = 0.000001 seconds.
Processor Cache SIZE SET TO 1024
Processor Cache Line SIZE SET TO 32 bytes.TML3
File Stride Size Set to 17 * Record Size.
RANDOM RANDOM BKWD Record Stride
KB ReCLEN WRITE READ REREAD Read Rite Read Fwrite Frewrite FREAD FREREAD
02400 4 34994 53668 304385 30136 59719
02431 130174 801254 801254 79796 133162
02400 512 300692 276975 291837 276681 313464TML3
024 309822 340026 3 3 3 3 3 08900 326826 306102 339059
02400 16384 357975 392544 369753 391219 370336 Zone TST TML20
HTM ml7
PI@ FriendlyElec:/Media/NVMe0n1 $ Sudo IOZONE -E -A -A -A -R 4K -R 16K -R 512K -R 1024K -R 16384K -I 1 -I 2
RANDOM RANDOM BKWD Cord StrideKB ReClen Write Read REREAD Read Write Read Rewrite Read Fwrite Frewrite Fread Freestml7
was finally a 1GB of 1GB文件：
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 1000M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
random random bkwd record stride
kB ReClen Write READ REREAD Read Write Read Read Fwrite FWRITE FREAD FREREAD
024000 4 35105 58082 30395 304444 7 27458 60895
024000 16 1024279 80210 80314 74596 13357TML0
02459 314704 282282282282 743 283883 277911 313413
024000 1024 329840 318228 319091 31764137714
024000 16 384 383289 385247 382642 382850 382870 381344
After the series of operations, the results are more or less consistent in all three tests, without much change.In the end, I got a read-write speed of about 380MB/s, which was far lower than the original promotional writing and reading speed of SSD, and the results of ODROID-M1. I think this is because the PCIe 2.0 X1 interface is used in this design, not the PCIE Gen 3.0 X2 interface used in Hardkernel board.Below is the output of LSPCI, for reference only:
.5GBE interface configuration and benchmark test
TML3
0
1
2
3
4L6 15
PI@FriendlyElec:/Media/NVME0N1 $ Sudo LSPCI -VTML3
Subsystem: Phison Electronics Corporation Device 5013
FLAGS: Bus Master, Fast Devsel, Laten Cy 0, IRQ 87
MEML6Memory at 380900000 (64-bit, Non-Prefetchable) [SIZE = 16K]
capabilities: [80] e0] e0] XPress Endpoint, MSI 00CAPabilities: [ D0] MSI-X: ENABLE+ Count = 9 Masked-
Capabilities: [E0] ENable-Count = 1/8 Maskable+ capabilities: [F8] Power Management Version 3TML3
CAPABILIES: [100] Lanetcy Tolerance l6capabilities: [110] l1 PM SubstateS
CAPABILIES: [200] Advanced Error Reportingml7
CAPABILITIES: [300] Secondary PCI EXPRER ss
kernel driver in use: nvme
ML6NANOPI R5s router
is configured with ETH0 Gigabit Ethernet "WAN" interface. Therefore, we must manually configure two 2.5GBE ports. I used the same test platform with the same "first part of Friendlywrt Evaluation", namely Ubuntu 20.04 laptop. Then, I connected REALTEK RTL8156BG USB 3.0 to 2.5GBE crypto dog to ETH1, UP XTREME I11 mini PC is connected to ETH2. I did not use the bridge interface like in Friendlywrt, but configured two different subnets: ETH1 was 192.168.2.0, ETH2 was 192.168.3.0.
现在我们在/etc/network/interfaces.d/中创建两个新文件：
- eth1
auto eth1
iface eth1 inet static
address 192.168.2.1
network 192.168.2.0
netmask 255.255.255.0
broadcast 192.168.2.255
- eth2
auto eth2
iface eth2 inet static
address 192.168.3.1
network 192.168 .3.0
netmask 255.255.255.0
broadcast 192.168.3.255
现在安装DHCP服务器
sudo apt install isc-dhcp-server
使用我们的两个子网编辑/etc/dhcp/dhcpd.conf文件：
subnet 192.168.2.0 netmask 255.255.255.0 {
RANGE 192.168.2.100 192.168.2.200;
Option Routers 192.168.2.1;
}
SUBNET 192.168.3.0 Netmask 255.255.255.0 {TML0RANGE 192.168.3.100 192.168.3.200 ;
Option Routers 192.168.3.1;
}
Before restarting the DHCP server:
sudo SystemCTL RESTART ISC-DHCP-Server
At this time, notebook computers and mini PCs should be able to get them from the nanopi R5s on their respective subnets. IP address. Now we can start the benchmark test on the interface.Download iperf3 using eth1 connected to laptop and then from R5S perspective receive:
0
1
2
3
4
5
6$ iperf3 -t 60 -c 192.168.2.1 -i 10
Connecting to host 192.168.2.1, port 5201
[ 5] local 192.168.2.130 port 59822 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 2.28 GBytes 1.96 Gbits/sec 42 1.41 MBytes
[ 5] 10.00-20.00 sec 2.02 GBytes 1.74 Gbits/sec 0 1.61 MBytes
[ 5] 20.00-30.00 sec 1.72 GBytes 1.48 Gbits/sec 0 1.62 MBytes
[ 5] 30.00-40.00 sec 1.87 GBytes 1.61 Gbits/sec 0 1.62 MBytes
[ 5] 40.00-50.00 sec 1.89 GBytes 1.62 Gbits/sec 0 1.70 MBytes
[ 5] 50.00-60.00 sec 2.06 GBytes 1.77 Gbits/sec 21 1.66 MByte /sec 63 sender
[ 5] 0.00-60.04 sec 11.8 GBytes 1.69 Gbits/sec receiver
iperf Done.
This is better than the 1.85 I got in OpenWrt Gbps is slower, and there is some retransmission of content.During the transmission process, I also used the L SBC-bench.sh monitoring system: @friendlyelec: ~ $ suno ./sbc-mrockChip RK3568 (35682000), Kernel: Aarch64, Userland: ARM64
CPU SYSFS TOPOLOGY (Clusters, CPUFREQ Members, CLOCKSEDS)
CPUFREQ MIN MAX
CPU Cluser Policy Speed Speed Speed Speed Speed L0
0 0 408 1992 Cortex-A55 / R2P0
Time CPU load %SYS %USR %IRQ TEMP
H3H3
This system does run at the maximum frequency mentioned in the publicity period during the test period. I didn't see any obvious problems here.I also used EthTool to check some information and statistical information: @friendlyelec: ~ $ Sudo ETHTOOL -I ETH1TML3DRIVER: R8125
Version: 9.008.00-napi
FIRMWARE-VERSION: TML3EML6Expans:
BUS-INFO: 0000: 01: 00.0
SUPPPPOTS -StATISTICS: YESTML7SUPPPOS-TEST: no
SUPPPROM-ACCESS: NO
supports-register-dump: YES
SUPPOPPORTS-FLAGS: no
PI@Friendlyelec: ~ Sudo ETHTOOL -S ETH1NIC Statistics:
TX_PACKETS: 451228
RX_PACKETS: 95 69147TX_ERRORS: 0
RX_ERRORS: 0RX_MISSED: 0l3align_error: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 9569102
broadcast: 45
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 31676089
rx_octets: 14506385933
rx_multicast64: 0
tx_unicast64: 451214
tx_broadcast64: 2
tx_multicast64: 12
tx_pause_on: 570
tx_pause_off: 570
tx_pause_all: 1140
tx_deferred: 0
tx_late_collision: 0
tx_all_collision: 0
tx_aborted32: 0
align_errors32: 0
rx_frame_too_long: 0
rx_runt: 0
rx_pause_on: 0
rx_pause_off: 0
rx_pause_all: 0
RX_UNKNOWN_OPCODE: 0
RX_MAC_ERROR: 0htmmmmm l7tx_underrun32: 0
RX_MAC_MISSED: 31
RX_TCAM_DROOPED: 0tdu: 0
rdu: 570
Really got some RX_MAC_MISSED.Now we are tested in turn: $ iPerr3 -T 60 -C 192.168.2.1 -i 10 -RConnecting to host 192.168.2.1, Port 5201
revery MO MO MO DE, Remote Host 192.168.2.1 is Sending[5] LOCAL 192.168.2.130 Port 59826 Connect to 192.168.2.1 Port 5201
[ID] Interval Transfer
[5] 0.00-10.00 SEC 1.75 Gbites/S ec
[5] 10.00-20.00 SEC 1.95 GBYTES 1.67 GBITS/Sec
[5] 20.00-30.00 SEC 1.95 Gbites/S ec[ 5] 30.00-40.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 40.00-50.00 sec 1.94 GBytes 1.67 Gbits/sec
[ 5] 50.00-60.00 sec 1.94 GBytes 1.67 Gbits/sec
- - - - - - - - - - - - - - - - - - - ------
[id] interval transfer bitrate retr
[5] 0.00-60.04 SEC 11.5 GBITS 1.64 GBITS /SEC 0 SENDER[5] 0.00-60.00 SEC 11.5 GBYTES 1.64 GBITS/SEC Receiver
IPERF DONE.
This looks much better than openwrt (1.12Gbps).
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
0
1
2
3
4
5
6devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.3.1 -i10
Connecting to host 192.168.3.1, port 5201
[ 5] local 192.168.3.100 port 37794 connected to 192.168.3.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 10.00-20.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 20.00-30.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 30.00-40.00 sec 2.73 GBytes 2.34 Gbits/sec 0 2.90 MBytes
[ 5] 40.00-50.00 sec 2.73 GBytes 2.35 Gbits/sec 0 4.37 MBytes
[ 5] 50.00-60.00 sec 2.73 GBytes 2.35 Gbits/sec 0 4.37 MBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.00 sec 16.4 GBytes 2.35 Gbits /sec 0 sender
[ 5] 0.00-60.00 sec 16.4 GBytes 2.35 Gbits/sec receiver
iperf Done.
Oh, great! This is the first time I've gotten 2.35 Gbps transfer speeds, so it looks promising!
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
Unless I'm mistaken, 25% of the IRQ means a core is fully utilized to handle this.Now let's try:
0
1
2
3
4
5
6
7devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.3.1 -i 10 -R
Connecting to host 192.168.3.1, port 5201
Reverse mode, remote host 192.168.3.1 is sending
[ 5] local 192.168.3.100 port 37800 connected to 192.168.3.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.92 GBytes 1.65 Gbits/sec
[ 5] 10.00-20.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 20.00-30.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 30.00-40.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 40.00-50.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 50.00-60.00 sec 1.84 GBytes 1.58 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.01 sec 11.1 GBytes 1.59 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.1 GBytes 1.59 Gbits/sec receiver
iperf Done.
The result is 1.59 Gbps, not quite perfect, but still better than OpenWrt.
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
^C
The CPU is running at full speed again, and it is far from 100% utilization, so I think the problem should lie elsewhere. Now we can use ethtool again to check the eth2 information and statistics.
0
1$ sudo ethtool -i eth2
driver: r8125
version: 9.008.00-NAPI
firmware-version:
expansion-rom-version:
bus-info: 0001:11:00.0
supports-statistics: yes
supports-test: no
supports-eeprom-access: no
supports-register-dump: yes
supports-priv-flags: no
0
1
2
3
4
5
6
7
8
9
sudo ethtool -S eth2
NIC statistics:
tx_packets: 8506609
rx_packets: 12553353
tx_errors: 0
rx_errors: 0
rx_missed: 0
align_errors: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 12553209
broadcast: 144
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 12543719502
rx_octets: 18471602900
rx_multicast64: 0
tx_unicast64: 8503557
tx_broadcast64: 3035
tx_multicast64: 17
tx_pause_on: 35
tx_pause_off: 35
tx_pause_all: 70
tx_deferred: 0
tx_late_collision: 0
tx_all_collision: 0
tx_aborted32: 0
align_errors32: 0
rx_frame_too_long: 0
rx_runt: 0
rx_pause_on: 0
rx_pause_off: 0
rx_pause_all: 0
rx_unknown_opcode: 0
rx_mac_error: 0
tx_underrun32: 0
rx_mac_missed: 335
rx_tcam_dropped: 0
tdu: 0
rdu: 35
In the test results here, I found that there are more rx_mac_missed. So I guess there should be some tweaks to improve performance.But according to my previous experience of adjusting the settings of RTL8156B, adjusting the settings is really tricky, and people with experience can't seem to agree on what setting options to adjust. I mainly refer to the RTL8156/8125 driver. Realtek engineers, as well as some networking experts among our readers.Configuring NAT between two 2.5GbE interfaces
Since the 2.5GbE interface doesn't work optimally with iperf3, I didn't bother testing the router performance in FriendlyWrt, but a few people asked me anyway. So next I will show how I configured NAT in Ubuntu 20.04 and will continue to test NAT performance, keeping in mind that it will definitely improve a lot in a few weeks or months.
Here we need to enable IP forwarding and NAT. The instructions I used were adapted from a post on networkreverse.
Edit /etc/sysctl.conf to enable IP forwarding (uncomment the following lines):
net.ipv4.ip_forward=1
Apply changes:
sudo sysctl -p
Now we enable NAT:
sudo iptables ! -o lo -t nat -A POSTROUTING -j MASQUERADE
We can now ping Xtreme on the 192.168.3.0 subnet from the laptop on the 192.168.2.0 subnet i11:
jaufranc@cnx-laptop-4:~$ ping 192.168.3.100
PING 192.168.3.100 (192.168.3.100) 56(84) bytes of data.
If you want the changes to be permanent, you can execute the following:
sudo apt install iptables-persistent
sudo sh -c 'iptables-save /etc/iptables/rules.v4'
I tried iperf3 between a UP Xtreme i11 and my laptop, the data was routed through a NanoPi R5S router.@cnx -laptop -4: ~ $ iPerf3 -T 60 -C 192.168.3.100 -i 10Connecting to host 192.168.3.100, PORT 5201
[5] LOCAL 192.168.2.130 Port 59430 Connect to 192.168.3.100 Port 5201
Bitrate Retr CWND[5] 0.00-10.00 SEC 914 MBYTES 767 MBITS/SEC 355 1011 Kbytes
[5] 10.00-20.00 SE C 912 MBYTES 765 Mbits/SEC 324 1.23 MBYTES[5] 20.00-30.00 SEC 917 MBITS/SEC 124 1.09 MBYTES
[5] 30.00-40.00 SEC 915 MBYTES 767 MBITS/SEC 150 942 Kbytes
[5] 40.00-50.00 SEC 915 MB 915 MB 915 YTES 767 MBITS L6- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.00 sec 5.36 GBytes 768 Mbits/sec 1095 sender
[ 5] 0.00-60.06 SEC 5.36 GBYTES 767 MBITS/SEC ReceiverTML3
IPERF : ~ $ iPerf3 -T 60 -C 192.168.3.100 -i 10 -R
Connecting to Host 192.168.3.100, Port 5201
Reverse Mode, remote host 192.168 .3.100 is sendingml7
[5] LOCAL 192.168.2.130 Port 59434 Connect to 192.168.3.100 Port 5201
[id] in Terval Transfer Bitrate
[5] 0.00-10.00 SEC 1.09 GBYTES 935 MBITS/SEC
[5] 10.00-20.00 SEC 1.09 GBYT ES 938 mbits /sec
[5] 20.00-30.00 SEC 1.09 GBYTES 938 mbits/sec
[5] 30.00-40.00 SEC 1.09 gbytes 938 mbits/ sec[5] 40.00-50.00 SEC 1.09 GBYTES 939 Mbits/Sec
[5] 50.09 Gbytes 1.09 Gbytes 937 MBITS/SEC
- - - -----------------
[ID] internsfer bitrate
[5] 0.00-60.05 SEC 6.55 GBYTES 937 MBITS/SEC 973 SENDER
[5] 0.00-60.00 SEC 6.55 GBYTES 937 MBITS/Security
IPERF 5
The transmission speed is 768 Mbps, and the other direction is 937 Mbps.@Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % cpu % sys % usr % nice % IRQ TEMP
1: 48: 1992MHz 0.90 25% 0.90 25% 0 % 0% 0% 0% 0% 25% 54.4 ° C
H6
If you use SBC-Bench.sh, the running frequency will be 1992 (Actually 1845 MHz), and 25%IRQ should be just It means that a Core has been completely used to deal with IRQ.is displayed according to the MPSTAT command, which should be processed by Core #0.
l3
TML3H0H0
$ MPSTAT -P All -I Sum
Linux 5.10.66 (F Riendlyelec) 06/05/22 _aarch64_ (4 cpu)
can be confirmed by using TOP and HTOP.
NANOPI
I just have a wall power meter. 6 Free -5.1 W
IPERF3 to ETH1 - 6.3 to 6.7w NAT test -6.2Wht between laptop and mini computers Ml7 It was tested after installing Ubuntu 20.04 and NVMe SSD on this product.
I am still using OpenWrt and without SSD NANOPI R5s for testing:
free L32IPERF3 - 6.0 to 6.2 W Note that because I use wall -mounted power meters, this value will include the power adapter (Khadas Vim4 usb. -C power adapter) efficiency loss. And this value may be higher than the value of using USB-C power timing. This should also be optimized by setting.
The final conclusion
's evaluation today is here. The optimization part should actually include "changing the firmware to make the RockChip kernel run in 1992 MHz", "Adjustment of various settings related to PCIe and Ethernet settings" and so on. Most of them are not very familiar.
(3), THT Linglong M NUC brush X86 soft route evaluation
foreword
many friends to eat X86 soft routes, or they are ready to step into X86 soft routing pit! First of all, it is necessary to clarify whether you are willing to spend time to toss. You can consider buying the finished OPNENWRT router to experience it. You can also buy a router that brushes the Merlin firmware or brush the PANDORABOX (Pandora) firmware to play.prepares related network devices
a suitable 2.5G port switch and X86 soft route with a 2.5G power port, a Type-C end of a 2.5G mesh or external wired network card at USB-A end.
TSMC Linglong M NUC is a mini machine that is very suitable for X86 soft routing. It has two 2.5GE power outlets, two SO-DIMM memory slots and a PCIE 3.0 NVME SSD M.2 slot. It is recommended to buy bare metal, bring up memory and NVMe SSD. Here are two 4GB Micron DDR4 3200MHz memory bar and 1 Intel Optane M10 16GB (Ao Teng Memory).
is installed with memory and NVME SSD.
can be used to use it.
Xike is a new network equipment brand. This is a two -layer management switch with 8 2.5G electric ports and 4 million -dimensional light ports to support port aggregation and VLAN division.
Friends bought this Xike switch to help me upgrade the network of my home, intending to be a 2.5G internal network environment in the whole house.
begins to adjust the device, which is still very simple for experienced people.
can see that the 10G optical port of the XIKE SKS7300-8GPY4XGS switch has been turned up, and the connection status is 10G Full rate. For friends who have 10G SFP modules, they can use 10G LAN. However, at this stage, 2.5G LAN is more suitable for the actual experience needs of the public. If you upgrade to a 10G optical port, the overall hardware cost will be significantly higher several times. For small and medium-sized enterprises, this switch is very suitable for use as an access switch connected to a 10G core switch. It is still more recommended for home users who are in need of 2.5G electrical ports and friends who like to mess with 2.5G LAN to buy this XIKE switch.
For users who have a GPON Stick ONU module, plugging it directly into the SFP interface of this SKS7300-8GPY4XGS switch can save an optical modem. It effectively solves the problem of insufficient accumulation of equipment in the weak current box, and can also improve the space inside the weak current box and reduce the problem of heat accumulation between devices.
What should I do if my computer (desktop and laptop) does not have a 2.5G network port? You can buy a superior USB to RJ45 2.5G wired network card. Connect to the computer through the Type-C end, and the RJ45 network port on the other end has a 2.5GbE (2500Mbps) rate.
Take a USB flash drive to make a boot disk
I generally like to use the tool BalenaEtcher to make a USB flash drive. It is simple and easy to operate! Supports ISO, DMG, IMG, IMG.GZ and other image file formats to create boot disks.
makes the Sea Spider V9 boot disk (flash)
. Go to the Sea Spider official website and find the download area in the service and support. Scroll to the upgrade address at the bottom, find V9.0, and select Download Now next to it to download the Sea Spider V9 firmware. Note that the ISO image file is downloaded.
Sea Spider V9 system
default login address: 192.168.0.1. The account and password are set by yourself after flashing the machine.
You can view the current information overview on the overview. The instrument icon shows CPU usage, memory usage, and disk usage. Below is the monitoring of the real-time total traffic of the intranet, and on the right is some key information. The LAN port already in use by
can identify the network port model: Realtek RTL8125 2.5GbE. The network port icon also identifies 2.5Gb.
Nigou can view the current memory usage in the overview. After flashing and initializing the system, you can see that 10.5% is used by default and 835.86 MB of storage space has been used. Total memory size: 7.6GB (8GB total memory), remaining 6.8GB available.
You can also view the CPU load in the overview. The
interface load can also view the data of the LAN port. The LAN port here is running in full-duplex mode, 2500Mb/s (that is, the 2.5GbE Gigabit port rate).
You can see the motherboard/CP information in the hardware information. Basically, the relevant information displayed is complete.
Here is the 16GB Optane memory, available capacity: 13.41GB, you can check the partitioning of the hard disk storage.
Both 2.5GbE Gigabit network ports can be recognized normally.
Haispider has module management, including Docker container modules.
Basically these modules are very complete for soft routing, including VPN IPsec, PHP suite, IGMP multicast proxy, etc.
FTP service.
file storage and sharing.
Comment: Sea Spider V9 requires purchasing relevant authorizations to fully use all functions. Fortunately, there is a long trial period. Relatively many functions are more suitable for network workers. For ordinary novice users, the operation may be very difficult, and some professional terms are convoluted and difficult to understand.
Make iStore OS boot disk (flash)
Go to KoolCenter official website to download iStore OS firmware. Find the iStore OS directory in the firmware list and enter it.
Considering that the newer X86 soft routers are all UEFI bootable, you need to select the x86_64_efi directory to download the iStore OS firmware.
Generally, there is Date (date) in the bottom list. Just select the iStore OS firmware with the latest date.
iStore OS (based on openwrt in-depth customization)
It is recommended to read the firmware usage precautions for the first time to understand the login IP (management port address) and login password.
In fact, iStore OS is deeply customized for ease of use by Chinese people. Compared with the original openwrt, it has more application stores (plug-ins) and Docker gameplay. The homepage of
will have legends showing traffic statistics, IP address, and network interface status, which looks clear at a glance.
Scroll down to see the disk information, storage service, Docker, download service and remote domain name. The playability is still very high.
Scroll to the bottom and you can see the system information. Nigou has identified the CPU temperature, CPU usage, and memory usage.
network wizard, this is the best wizard function for setting up the connection method for novice users. There is system information in the
overview. You can see the computer model, processor architecture and other related information.
You can also see the memory usage in the overview.
One of the more difficult to master and use soft routers is disk management. Here iStore OS directly implements a graphical DiskMan disk management, which reduces the difficulty of operation for novices and improves the ease of use.
iStore OS has built-in Docker
. For many friends who want to play soft routing, Docker is definitely a must-try. However, many openwrts do not have built-in Docker and require manual installation of plug-ins, which stumps many beginners or novice users who are just getting started. iStore OS has built-in Docker components directly to facilitate containerized deployment. Moreover, it supports CPU and memory configuration settings in the advanced settings, which is very user-friendly.
iStore OS exclusive app store
If you want to research smart home experience, you can consider Home Assistant.
Yiyouyun is suitable for micro home deployment data service center. There are many tutorials on the Internet, so I won’t go into it in depth.
If you can’t afford a dedicated NAS device, soft routers can actually serve as simple NAS, but most X86 soft routers are not easy to use as a NAS. There are two tools directly available in iStore OS to solve this problem. For friends who want to build a video and audio center, they are two must-install plug-ins: Jellyfin and Nas Tools. If you know how to use it, you can definitely easily turn an X86 soft router into a NAS.
Aria2 downloader is also a very practical download tool if you know how to use it. If you have two broadbands at home, in order to prevent the broadband from being idle during work or business trips, you might as well try deploying Netcenter Cloud to earn some income.
I actually prefer the HomeBox intranet speed test, which has a better experience than the self-built SpeedTest speed test server. Later I will show you the experience of using HomeBox intranet speed test.
DiskMan disk management is definitely a plug-in worth installing. Friends who know how to use qBittorrent downloader say it is easy to use. There are more detailed and professional introductions on the specific usage methods and tutorials online, so if you are interested, go and research it.
KMS server can solve the problem of Windows and Office not being activated.
Installing the plug-in is as simple and convenient as installing the app. This is also the greatest ease of use of iStore OS.
These four plug-ins are what I use most. Among them, multi-line multicast is suitable for areas with broadband overlay. In a word, 300M overlay can run up to 600M, and 500M overlay can run up to 1000M, but there is no need to upgrade the broadband package and spend extra money. It’s worth thinking about.
KMS server is enabled. Basically, local Windows and Office have no longer encountered unauthorized situations.
This is the HomeBox intranet speed test, which can directly measure the speed of the intranet. I use Teclast Linglong M NUC’s 2.5GbE Gigabit network port, CAT6 (Category 6e) network cable, and Type-C to RJ45 2.5GbE external wired network card, which can achieve a 2500Mbps full-link LAN (intranet) experience .
Summary
Users with basic network engineering skills or a relatively good understanding of the Internet can consider starting X86 soft routing. Sea Spider V9 actually prefers traditional network equipment, combining functions such as switches, routers, and AC controllers. iStore OS is suitable for the general public. It is more playable and easier to get started. I personally recommend flashing the iStore OS firmware to use it, which is basically the Router OS in openwrt that I have used and experienced well.
The full output of the sbc-bench.sh script can be found at at pastebin We see that the "1992" MHz advertised frequency is tested in reality as 1845 MHz, so I feel like some optimization could be done here.(1), Electric Rhino R68S Soft Route Evaluation
I. Appearance and key configuration Analysis of
.1 R68S Key ConfigurationCPU is the memory and html. 6 flash memory chip .Electric Rhino R68S RK-3568 ARM Processor, 22nm process, 4 -core Cortex A55 Architecture, main frequency 2GHz, supports the maximum 8G memory;
RK3568 built -in 1 QSGMII, 2 PCIE 3.0x1 channels, also with 2 Gigabit Ether interfaces, so you can use 42.5GB+2x1000MB 4 4 Gigabit Ether interface. Net port combination.
R68S core configuration is shown in the figure below:
HTM core configuration
LPDDR4 memory chip is from Rayson, the model is RS1G32LF4D2BDS- 53BT, the capacity is 4GB (2GB version is 70 yuan); Resee (under Shenzhen Jiang Bolong), the model is NCEMASKGG-16G, the capacity is 16GB;
There is a Gigabit Ethernet exchange chip in the upper left corner of the processor with a model of RTL8211F. When you see this crab logo, you know that from Riyu ;
does n’t there be 4 mesh ports? What about the remaining 3 exchanges?
is on the back of the motherboard.
Two 2.5GB exchange chips also comes from Riyu, the model is RTL8125BG:
motherboard
Ruyu's 8125B, Intel 's I225 B3 and Qualcomm QCA8081 are currently the most common.2 R68s Appearance.1 encryption and decryption performanceengineering machine is sent, there is only one piece of bare plate.
is because I want to do the temperature test in weak electric box , so the boss sent a shell after a few days (for test model, non -final version).
I received is made of plastic material (also CNC aluminum shell, the price will be about 100 yuan), and there are a lot of heat dissipation holes.
official version shell will be made into a frosted effect
HT There is no dust -proof cotton, so if you use R68S to put the desktop, you still need to consider the problem of dust. After all, dust is the first of electronic equipment. One killer:
front
HTM has a large amount of heat dissipation holes in the two sides of the 2 side:
machine side HT
If you will put R68s in a weak electric box like me, then the hollow design of the shell is actually very beneficial to heat dissipation.
But if you put R68S on the desktop, I also recommend buying a CNC aluminum shell, that is, dustproof and waterproof, heat dissipation is okay.
cnc shell is about length. Single version.
I actually tested it, and the 12V 1A light cat power adapter can also work normally.
Of course, if you have the need to connect to external hard disks, I suggest that it is best to put a 12V 1.5A power adapter.
When it comes to external hard disks, the USB 3.0 performance of R68S is still very strong. Reading and writing can run to about 120MB. If you want to watch movies on the R68S, you don't have to worry about the original disk high -code movie will be stuck.
2. Performance and stability Test
For a soft routing , performance and stability are equally important.
directly look at the table:
Zhihu@仙鱼
CPU CoreMARK
AES...
CHA...
AES hard decoding
R68S (RK3568)
03101
is
N1(S905D)
3
8704
is
2 R4S (RK3399)
36089
is
J1900
no
J4125
51102is
The encryption and decryption performance of R68S is much stronger than the N1 of ARM architecture , but it lags behind the 6-core R4S (RK3399) by about 20%. The CPU CoreMARK of
R68S is about 6000 points behind J1900. However, J1900 does not support hardware AES decoding after all, and R68S still has some advantages in this regard. If
is compared with J4125, which also supports AES hard decoding, R68S lags behind by more than half.
Although the R68S is not excellent in terms of scores, in actual use, the performance is more than sufficient.
.2 NAT performanceNAT performance mainly tests the broadband speed. The tested topology is as follows:
NAT test
The test rate is as follows:
It can be seen that if the transmission loss is removed, the R68S has reached the limit of Gigabit broadband.
At this time, the CPU load is around 45%.
.3 LAN forwarding performanceLAN forwarding capability mainly tests the forwarding rate between the two 2.5G LAN ports of the R68S. I did a total of 3 different tests. The topologies of
R68S and iPerf3 client installed on the computer respectively, ThinkPAD X13 (2.5G USB 3.0 network card) is directly connected to the 2.5G network port of R68S. The test results are:
iPerf3
The downlink rate is close to 2.1Gb, which may be limited by the performance of the USB network card. The uplink rate is slightly worse at 1.94Gb.
At this time, the CPU load is about 30%.
LAN TO The downlink rate of LAN
can reach around 2.3Gb, which is close to the limit, but the uplink rate still lags behind. In multiple tests, the fastest rate only reached 1.85Gb.
At the same time, I also tried large file transfer: the reading rate of
can be stable at 210MB, and the peak value can reach 242MB.
However, the writing speed is relatively unstable and fluctuates around 150MB. This may be due to a bottleneck in the hard disk speed, or the network card uplink is inherently slow.
LAN file transfer speed
Thinkpad The X13's AX200 network card can achieve a 2402Mb negotiation rate under the Redmi AX5400 e-sports router, and the laptop uses WIFI most of the time, so the LAN TO WIFI6 test is also very necessary:
LAN TO WIFI6
AX5400 is directly connected to the NAS. The maximum rate that can be reached by using WIFI6 is 1850/624. The rate drops slightly after being forwarded by R68S, but the drop is only 5%, which is acceptable.
Overall, the LAN forwarding capability of R68S is quite good. If you don’t have many 2.5G devices, you can still save a 2.5G switch by using R68S.
However, if you install docker on R68S, the LAN bridge forwarding performance will drop significantly. You must add an switch :
.4 Temperature and StabilityThe firmware of R68S is designed by the master lean, and it only lasts 10 days. Judging from the situation, the stability is still very good.
running time
R68S I have been using it in a weak current box.
The current indoor temperature in Zhejiang is about 27 degrees, and the standby temperature of the R68S bare board is about 47 degrees. After the
case arrived, I observed that the standby temperature soared to about 53 degrees.
However, it’s not just the R68S that’s rising in temperature. In the past two days, the casing of my Redmi AX5400 router has been very hot even if it is placed on the TV cabinet.
When the temperature is high, the indoor temperature may be around 31 degrees, and the temperature of the weak current box will be even higher. I will continue to update the temperature monitoring of R68S.
Zhejiang has been experiencing high temperatures for more than 40 days, and the indoor temperature is basically around 32 degrees.
The temperature of R68S in the weak current box is basically around 58 degrees, and it is very stable in use, with no crashes or interruptions.
3. Tutorial on playing with the machine
Here we mainly talk about the tutorial on flashing the machine and installing some plug-ins
.1 Resetting and flashing the machineR68S has a reset button. If the system is confused, or you cannot enter the router settings page, then press and hold the reset button. It is very convenient to restore the factory settings (when plugged in) by pressing the button for 5 seconds.
In addition, you need to flash the R68S, which can be completed in a few seconds. Compared with the N1 flash machine, it is really a world of difference.
However, convenience is convenience, and you still need to learn the tutorial.
) Install Rockchip micro-driver and flash toollink: https://pan.baidu.com/s/1pTxQ1yfjYTccClHcy068aw?pwd=6tnl Extraction code: 6tnl
firmware selects
Flashing sequence
You will hear a prompt on the computer in about two seconds, and the development tool will also prompt that the device has been found.
Erase Except Flash
. Flashing
.2 WEB upgradeOf course, only when a major version is upgraded or the internal partition is changed, you need to use to brush . For ordinary updates, just use WEB upgrade:
WEB upgrade
WEB upgrade requires the "sysupgrade" image, and the file suffix name is general It's all tar.
.2 plug-in installationmachine comes with a pure version of firmware built into the factory. If you need to use certain firmware, you can upload the IPK or install it with the SSH command.
For example, the Alibaba Cloud Disk WebDAV plug-in that is very needed by audio and video players is not built-in when it leaves the factory. We can use the SSH tool or the TYYD copy command in OpenWRT to install it (copy one by one and press Enter):
wget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/aliyundrive-webdav_1.3.2-1_aarch64_generic.ipkwget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/luci-app-aliyundrive-webdav_1.3.2_all.ipkwget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/luci-i18n-aliyundrive-webdav-zh-cn_1.3.2-1_all.ipkopkg install aliyundrive-webdav_1.3.2-1_aarch64_generic.ipkopkg install luci-app-aliyundrive-webdav_1.3.2_all.ipkopkg install luci-i18n-aliyundrive-webdav-zh-cn_1.3.2-1_all.ipkAfter the installation is completed, refresh it and there will be a plug-in in the service:
Alibaba Cloud Disk
4. Summary and purchase suggestions
Electric Rhino R68S soft router, for me Personally it's very suitable.
Because I already have a finished NAS with a 2.5G network port, and the router usually uses AP mode, so for me, I only need a soft router with a 2.5G network port as a gateway.
The current soft routing market is very large. The price of J4125's 2.5G soft routing has also been reduced to about 600 yuan. If you have more needs, such as the standard all-in-one gameplay of Aikuai split + OpenWRT bypass + unraid, then R68S may Unable to meet your request.
But if you are like me and only need a pure soft router, then R68S is still very suitable for you.
(2), NanoPi R5S soft routing evaluation
NanoPi R5S unboxing
NanoPi with metal case SSD and cooling pad After unboxing R5S
, I found that the router has been assembled. They also came with 6 rubber feet and a piece of
NanoPi R5S
NanoPi R5S has a microSD card slot on one side, and the rear panel has a USB-C port for power supply, a WiFi antenna hole (this antenna hole can also be inserted into GPIO, UART console and other cables), two 2.5 GbERJ45 LAN port, a Gigabit Ethernet WAN port and HDMI video output .
NanoPi R5S mask key
On the other side of the NanoPi R5S we will find a mask button for firmware upgrades, while on the front panel there are four LED lights for the "System" and Ethernet ports, and two USB 3.0 port.
NanoPi R5S disassembly
Generally, the main reasons for wanting to disassemble are as follows: out of curiosity, to install M.2 NVMe SSD, to solder SPI flash memory, to connect some GPIO, RTC battery or to debug UART to TTL plate. It's easy to take it apart, just loosen the four screws that need to be loosened.
After taking it apart, you can see the bottom of the circuit board with the M.2 Key M slot, the space occupied by the SPI flash memory (right), and the Samsung KLM8G1GETF-B041 eMMC 5.1 flash memory 8GB motherboard.
SPI flash and M.2 socket on the NanoPi R5S board
Before removing the circuit board from the case, four more screws need to be loosened.
NanoPi R5S SBC
Rayson RS512M32LM4 D2BDS is a 2GB LPDDR4X memory chip. I found the RTL8211F (GbE) and
NanoPi R5S with a metal casing that acts as a CPU cooler.
Test setup using 2.5GbE USB dongle and UP Xtreme i11 mini PC
You may remember that I had some performance issues with not too long ago when using the RTL8156B USB dongle , but now this issue has been resolved. Because Realtek sent me another RTL8156BG, I was previously using iperf3 on my laptop from UP Xtreme When i11 mini PC transmits data and performs full duplex test, its test result is 2.34Gbps/2.29Gbps.
I am now using the same settings for testing, but this time the middle stuff is replaced with NanoPi R5S.
NanoPi R5S
TP-Link switch here is only used as a Gigabit Ethernet switch , and the NanoPi is connected through Xiaomi AX6000 router The R5S router's WAN port connects to the internet, which prevents me from having to install some software packages on the device. While placing the NanoPi R5S on top of the TP-Link switch resulted in better photos, both devices get very hot and I don't recommend this. Since my Ethernet cable was very short, I had to move the router to a table to test it. The
OpenWrt and iperf3 benchmarks
FriendlyWrt come pre-installed on the router, so they are ready to use right out of the box. It is also possible to immediately access the LuCI interface or SSH using "root" as user and "password" as password. This is indeed very convenient, but it is not very safe, and may be against the law in some countries. Anyway, it is best to change the password at least the first time you use it. The status of
FriendlyWrt
FriendlyWrt is based on the kernel of OpenWrt 22.03.0-rc1 and Linux 5.10.66. It uses less than 250MB of RAM when idle and using default settings, which is more than enough since the system comes with 2GB of RAM.
FriendlyWrt network
I haven't connected SSD yet, so I only mounted the root partition. The available memory is 6.7 GB, and the root partition uses 920 KB. All interfaces can correctly obtain IP addresses through DHCP when starting up. Devices on the LAN can also be accessed through hostname.lan and obtain some IPv6 addresses.
Starting with the iperf3 benchmark, I first ran "iperf3 -s" on the NanoPi R5S and ran the following commands on the laptop:
- Download: (View the results of the Rx test on the NanoPiR5S)
iperf3 -t 60 -c 192.168.2.1 -i 10Connecting to host 192.168.2.1, port 5201[ 5] local 192.168.2.130 port 48782 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate Retr Cwnd[ 5] 0.00-10.00 sec 2.23 GBytes 1.92 Gbits/sec 0 1.69 MBytes[ 5] 10.00-20.00 sec 2.02 GBytes 1.74 Gbits/sec 0 1.69 MBytes[ 5] 20.00-30.00 sec 2.33 GBytes 2.00 Gbits/sec 0 2.64 MBytes[ 5] 30.00-40.00 sec 1.66 GBytes 1.42 Gbits/sec 0 2.64 MBytes[ 5] 40.00-50.00 sec 2.62 GBytes 2.25 Gbits/sec 0 2.64 MBytes[ 5] 50.00-60.00 sec 2.01 GBytes 1.73 Gbits/sec 0 2.64 MBytes- - - - - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr[ 5] 0.00-60.00 sec 12.9 GBytes 1.84 Gbits/sec 0 sender[5] 0.00-60.05 sec 12.9 GBytes 1.84 Gbits/sec receiveriperf Done.
- Upload (View the results of Tx test on NanoPiR5S):
$ iperf3 -t 60 -c 192.168.2.1 -i 10 -RConnecting to host 192.168.2.1, port 5201Reverse mode, remote host 192.168.2.1 is sending[ 5] local 192.168.2.130 port 48786 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate[ 5] 0.00-10.00 sec 1.29 GBytes 1.11 Gbits/sec[ 5] 10.00-20.00 sec 1.31 GBytes 1.12 Gbits/sec[ 5] 20.00-30.00 sec 1.33 GBytes 1.14 Gbits/sec[ 5] 30.00-40.00 sec 1.27 GBytes 1.09 Gbits/sec[ 5] 40.00-50.00 sec 1.30 GBytes 1.12 Gbits/sec[ 5] 50.00-60.00 sec 1.30 GBytes 1.12 Gbits/sec- - - - - - - - - - - - - - - - - - - - - - - - - -[ID] Interval Transfer Bitrate Retr[5] 0.00-60.06 sec 7.80 GBytes 1.12 Gbits/sec 0 sender[ 5] 0.00-60.00 sec 7.80 GBytes 1.12 Gbits/sec receiveriperf Done.
It can be seen from the above that this is not exactly the 2.35 advertised by FriendlyElec. Gbps and 1.85 Mbps because in this configuration I'm seeing 1.84 Gbps and 1.12 Gbps. There are also some changes on the Rx side when looking at the 10 second transfer test report. Fortunately, no retransmission issues were found.Now we try the same with another WAN port on NanoPi R5S and run the command from UP Xtreme i11:
- Download (Rx):
devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.2.1 -i 10Connecting to host 192.168.2.1, port 5201[ 5] local 192.168.2.207 port 52052 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate Retr Cwnd[ 5] 0.00-10.00 sec 2.59 GBytes 2.22 Gbits/sec 0 1.67 MBytes[ 5] 10.00-20.00 sec 2.62 GBytes 2.25 Gbits/sec 0 1.93 MBytes[ 5] 20.00-30.00 sec 2.60 GBytes 2.24 Gbits/sec 0 1.93 MBytes[ 5] 30.00-40.00 sec 2.47 GBytes[ 5] 50.00-60.00 sec 2.45 GBytes 2.10 Gbits/sec 0 4.90 MBytes- - - - - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr[ 5] 0.00-60.00 sec 15.2 GBytes 2.17 Gbits/sec 0 sender[5] 0.00-60.00 sec 15.1 GBytes 2.17 Gbits/sec receiveriperf Done.
- upload (Tx);
devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.2.1 -i 10 -RConnecting to host 192.168.2.1, port 5201Reverse mode, remote host 192.168.2.1 is sending[ 5] local 192.168.2.207 port 52056 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate[ 5] 0.00-10.00 sec 1.31 GBytes 1.13 Gbits/sec[ 5] 10.00-20.00 sec 1.29 GBytes 1.11 Gbits/sec[ 5] 20.00-30.00 sec 1.32 GBytes 1.14 Gbits/sec[ 5] 30.00-40.00 sec 1.30 GBytes 1.11 Gbits/sec[ 5] 40.00-50.00 sec 1.33 GBytes 1.14 Gbits/sec[ 5] 50.00-60.00 sec 1.27 GBytes 1.09 Gbits/sec- - - - - - - - - - - - - - - - - - - - - - - - - -[ID] Interval Transfer Bitrate Retr[5] 0.00-60.00 sec 7.82 GBytes 1.12 Gbits/sec 0 sender[5] 0.00-60.00 sec 7.82 GBytes 1.12 Gbits/sec receiveriperf Done.when downloading (Rx) at 2.17 The Gbps operation seems to be better; but when uploading (Tx), it is still running very slow at 1.12 Gbps.
Let’s take a look at the speed when using two WAN ports at the same time? One is running "iperf3 on UP Xtreme i11 -s", the other port is to run the following command on the notebook:
0
1
2
3
4
5
6jaufranc@cnx-laptop-4:~$ iperf3 -t 60 -c 192.168.2.207 -i 10
Connecting to host 192.168.2.207, port 5201
[ 5] local 192.168.2.130 port 49762 connected to 192.168.2.207 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 1.39 GBytes 1.19 Gbits/sec 0 3.15 MBytes
[ 5] 10.00-20.00 sec 1.42 GBytes 1.22 Gbits/sec 0 3.15 MBytes
[ 5] 20.00-30.00 sec 1.44 GBytes 1.24 Gbits/sec 0 3.15 MBytes
[ 5] 30.00-40.00 sec 1.42 GBytes 1.22 Gbits/sec 0 3.15 MBytes
[ 5] 40.00-50.00 sec 1.42 GBytes 1.22 Gbits/sec 0 3.15 MBytes
[ 5] 50.00-60.00 sec 1.39 GBytes 1.19 Gbits/sec 0 3.15 MByte /sec 0 sender
[ 5] 0.00-60.05 sec 8.48 GBytes 1.21 Gbits/sec receiver
iperf Done.
This data should be within the expected range, because the upload (Tx) data we obtained before was relatively low; at 1.21 Gbps, it is still a little higher than the 1.12 Gbps when we only used upload (Tx) .Let’s try again in reverse:
$ iperf3 -t 60 -c 192.168.2.207 -i 10 -RConnecting to host 192.168.2.207, port 5201Reverse mode, remote host 192.168.2.207 is sending[ 5] local 192.168.2.130 port 49766 connected to 192.168.2.207 port 5201[ ID] Interval Transfer Bitrate[ 5] 0.00-10.00 sec 2.04 GBytes 1.75 Gbits/sec[ 5] 10.00-20.00 sec 2.04 GBytes 1.75 Gbits/sec[ 5] 20.00-30.00 sec 2.04 GBytes 1.75 Gbits/sec[ 5] 30.00-40.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 40.00-50.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 50.00-60.00 sec 2.05 GBytes 1.76 Gbits/sec- - - - - - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr[ 5] 0.00-60.05 sec 12.3 GBytes 1.75 Gbits/sec 0 sender[ 5] 0.00-60.00 sec 12.3 GBytes 1.76 Gbits/sec receiveriperf Done.
This time it was 1.75 Gbps, I have no idea what happened or how this happened.I then ran the 600 second test again using full duplex and checked the CPU, memory and temperatures reported in LuCI.
The CPU load of NanoPi R5S
can be seen from the above that its CPU load is between 2.0 and 2.5.
Temperature of NanoPi R5S
The utilization of CPU #0 is relatively low, and CPU #1 to #3 occupy most of the resources.
never went above 60°C in a room with an ambient temperature of 28°C.
FriendlyWrt Memory Usage
For some unknown reason, the available memory usage of 1.8GB has hardly changed. Performance during full-duplex testing was: 1.66 Gbps and 736 Mbps.The result of
is a bit disappointing. I am now intending to switch to Friendlycore (Ubuntu Core) to check whether similar results will be obtained, and then perform further testing. If you want me to evaluate other contents of OpenWrt, you can tell me in the comment area.
is further tested to switch to the FriendlyCore image based on Ubuntu 20.04, because I am more familiar with the Debian -based operating system, and some tools cannot run on OpenWRT. Note that the performance of the test results is still not ideal. This is why I call the prediction test. With the use and adjustment software of more and more people, the situation should be improved in the next few months.
OPENWRT optimization?
Now we do not discuss the mirror -based image based on Ubuntu. Because Friendelec (Guangzhou Friendship Electronic Technology Co., Ltd.) informed me that they had added some optimization, so I upgraded the updated version of Friendlywrt and tested this image: TML3. TML6 we did some of the new mirror image Optimized, such as network card interrupt settings, uninstalled support, etc. " So I downloaded the" RK3568888 found on Google Drive on Google Drive -eflasher-Friendlywrt-20220526.img.gz ", and then record it with USBImager to , Then Boot to the router. After
, it will automatically record the image into the EMMC flash. If you connect to the display, you can operate according to the results. After the operation is completed, remove the Microsd card and restart the router.
can be checked by connecting the HDMI display (as shown above) or viewing the LED on the device. This process is very fast, just a few seconds installed on the EMMC flash memory.
This new version of the image is mainly on 40-net-smpinity files.pre-installed before In FriendlyWrt, it looks like this:
0
1
2
3friendlyelec,nanopi-r5s)
set_interface_core 8 "eth0"
echo 7 /sys/class/net/eth0/queues/rx-0/rps_cpus
set_interface_core 2 "eth1-0"
set_interface_core 4 "eth1-16"
set_interface_core 4 "eth1-18"
echo b /sys/class/net/eth1/queues/rx-0/rps_cpus
set_interface_core 4 "eth2-0"
set_interface_core 2 "eth2-16"
set_interface_core 2 "eth2-18"
echo 9 /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
and the new ht The ml640-net-smp-affinity files are indeed different:
0
1
2
3friendlyelec,nanopi-r5s)
set_interface_core 8 "eth0"
echo f /sys/class/net/eth0/queues/rx-0/rps_cpus
set_interface_core 4 "eth1-0"
set_interface_core 4 "eth1-16"
set_interface_core 4 "eth1-18"
echo b /sys/class/net/eth1/queues/rx-0/rps_cpus
set_interface_core 2 "eth2-0"
set_interface_core 2 "eth2-16"
set_interface_core 2 "eth2-18"
echo d /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
Willy Tarreau also explained the reason for the changes made to the eth1 interface:
It involves RPS. They receive this IRQ (interrupt request) on core 2 and redistribute the incoming traffic into core 0, core 1, core 3. This is the correct way to use RPS. However, to achieve this, you must manually assign a network performance testing tool iperf and observe the first saturated core. If you use ksoftirqd to saturate core 2 first, then make sure iperf is runnable on any of the other 3 cores. If core2 is slightly idle, then try setting iperf on it. If putting iperf on top of it causes ksoftirqd to pop up, then they will hinder each other, and the user will be more willing to change the RPS settings to help free up another core and use it for iperf.
I actually didn't try this method before testing and switching to Ubuntu.When I try to use the new FriendlyWrt image, I get even worse results:
0
1
2
3
4
5
6
7
8
9
$ iperf3 -t 60 -c 192.168.2.1 -i 10
Connecting to host 192.168.2.1, port 5201
[ 5] local 192.168.2.130 port 49590 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 1.92 GBytes 1.65 Gbits/sec 0 1.62 MBytes
[ 5] 10.00-20.00 sec 1.91 GBytes 1.64 Gbits/sec 10 2.34 MBytes
[ 5] 20.00-30.00 sec 1.90 GBytes 1.63 Gbits/sec 0 2.61 MBytes
[ 5] 30.00-40.00 sec 1.85 GBytes 1.59 Gbits/sec 4 1.30 MBytes
[ 5] 40.00-50.00 sec 1.88 GBytes 1.61 Gbits/sec 1 1.06 MBytes
[ 5] 50.00-60.00 sec 1.76 GBytes 1.51 Gbits/sec 2 868 KByte /sec 17 sender
[ 5] 0.00-60.05 sec 11.2 GBytes 1.60 Gbits/sec receiver
iperf Done.
$ iperf3 -t 60 -c 192.168.2.1 -i 10 -R
Connecting to host 192.168.2.1, port 5201
Reverse mode, remote host 192.168.2.1 is sending
[ 5] local 192.168.2.130 port 49594 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.22 GBytes 1.05 Gbits/sec
[ 5] 10.00-20.00 sec 1.36 GBytes 1.17 Gbits/sec
[ 5] 20.00-30.00 sec 1.31 GBytes 1.12 Gbits/sec
[ 5] 30.00-40.00 sec 1.46 GBytes 1.26 Gbits/sec
[ 5] 40.00-50.00 sec 1.47 GBytes 1.26 Gbits/sec
[ 5] 50.00-60.00 sec 1.46 GBytes 1.26 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.05 sec 8.29 GBytes 1.19 Gbits/sec 1 sender
[ 5] 0.00-60.00 sec 8.29 GBytes 1.19 Gbits/sec receiver
iperf Done.
Therefore, this issue has to be re-examined.M.2 NVMe SSD installation in NanoPi R5S
I purchased an APACER AS2280 (AP256GAS2280P4-1) PCIe Gen 3.0 x4 SSD a while ago and it can achieve sequential read speeds of up to 1,800 MB/s and up to 1,100 on the right hardware Sequential write speed in MB/s.
Apacer M.2 2280 PCIe SSD
The installation process was easy as I only had to loosen four screws to remove the bottom cover, install the SSD and then secure it in place with the screws they provided.
NANOPI R5s Installation
Riendlycore
I first tried to install Friendlycore with EFLASHER mirror.
use EFLASHER mirror to install FriendlyCore
I started the router, but later I noticed that the WAN interface link was not displayed on the TP-Link switch. Only the power supply LED lit up, and the power supply LED light up Mirror is normal. I tried again and settled through the EFLASHER UI settings by clicking "to complete", but still not.
eflasher Install F Riendlycore
Start directly from the Microsd card and run the operating system from there. After doing so, it is normal to run. However, if you intend to use the NANOPI R5s for a variety of uses and look forward to using the desktop environment in Ubuntu 20.04 mirrors, it may be disappointed because HDMI output can only be used to access the terminal.
UBuntu 20.04 FRIENDLELEC login TML3
Friendlycore system information
You can find the startup log on CNX Software Pastebin.我使用pi/pi凭据（用户名/密码）通过过了SSH登录，并将系统升级到了最新软件包：
sudo apt update
sudo apt dist-upgrade
现在，我们运行一些命令来获取系统信息：@FriendlyELEC:~$ cat /etc/lsb-releaseDistRib_id=ubuntu
Distrib_release=20.04
h6h6distrib_codename = Focal
DistRib_Descripting = "Ubuntu 20.04.4 LTS"
PI@Friendlye LEC: ~ $ Uname -ALinux Friendlyelec 5.10.66 #219 SMP PRIMPT FRI APR 22 18:20:21 cst 2022 AARCH64 aarch64 aarch64 GNU/Linux
pi@FriendlyELEC:~$ free -mh
total used free shared buff/cache available
Mem: 1.9Gi 150Mi 1.7Gi 3.0Mi 114Mi 1.7Gi
Swap: 0B 0B 0B
pi@FriendlyELEC:~$ df -mh
Filesystem Size Used Avail Use% Mounted on
udev 969M 0 969M 0% /dev
tmpfs 197M 480K 196M 1% /run
overlay 27G 1013M 26G 4% /
tmpfs 981M 0 981M 0% /dev/shm
tmpfs 5.0M 4.0K 5.0M 1% /run/lock
tmpfs 981M 0 981M 0% /sys/ FS/CGROUP
TMPFS 197m 0 197m 0%/RUN/User/1000
Except for the NVMe driver without automatic mounting, everything looks pretty good.Now we use inxi to find more details:
0
1
2
3
4
5
6
7
8
9
24
pi@FriendlyELEC:~$ inxi -Fc0
System:
Host: FriendlyELEC Kernel: 5.10.66 aarch64 bits: 64 Console: tty 0
Distro: Ubuntu 20.04.4 LTS (Focal Fossa)
Machine:
Type: ARM Device System: FriendlyElec NanoPi R5S details: N/A
serial: 8cbfe79e107c459c
Battery:
ID-1: test_battery charge: 100% condition: N/A
CPU:
Topology: Quad Core model: N/A variant: cortex-a55 bits: 64 type: MCP
Speed: 408 MHz min/max: 408/1992 MHz Core speeds (MHz): 1: 1992 2: 1992
Graphics:
Device-1: display-subsystem driver: rockchip_drm v: N/A
Device-2: mali-bifrost driver: mali v: N/A
Device-3: rk3568-dw-hdmi driver: dwhdmi_rockchip v: N/A
Display: server: -G --display
Audio:
Device-1: rk3568-dw-hdmi driver: dwhdmi_rockchip
Device-2: simple-audio-card driver: asoc_simple_card
Device-3: simple-audio-card driver: N/A
Device-4: simple-audio-card driver: asoc_simple_card
Sound Server: ALSA v: k5.10.66
Network:
Device-1: Realtek RTL8125 2.5GbE driver: r8125
IF: eth1 state: down mac: e2:1d:62:a1:1a:ca
Device-2: Realtek RTL8125 2.5GbE driver: r8125
IF: eth1 state: down mac: e2:1d:62:a1:1a:ca
Device-3: rk3568-gmac driver: rk_gmac_dwmac
IF-ID-1: eth0 state: up speed: 1000 Mbps duplex: full
mac: de:1d:62:a1:1a:ca
IF-ID-2: eth2 state: down mac: 12:bf:2b:d6:4b:e0
Drives:
Local Storage: total: 274.88 GiB used: 1012.3 MiB (0.4%)
ID-1: /dev/mmcblk0 model: SD16G size: 29.12 GiB
ID-2: /dev/mmcblk2 model: 8GTF4R size: 7.28 GiB
ID-3: /dev/nvme0n1 vendor: Apacer model: AS2280P4 256GB size: 238.47 GiB
Partition:
ID-1: / size: 26.48 GiB used: 1012.3 MiB (3.7%) fs: overlay
source: ERR-102
Sensors:
System Temperatures: cpu: 46.1 C mobo: N/A
Fan Speeds (RPM): N/A
Info:
Processes: 130 Uptime: 7m Memory: 1.92 GiB used: 211.4 MiB (10.8%)
Init: systemd Shell: bash inxi: 3.0.38
Only the eth0 WAN port is open, the eth1/eth2 2.5GbE port is closed, and no configuration is displayed at all.FriendlyElec seems to focus on Friendlywrt mirror. They told me that it has not been optimized on Friendlycore, so most people may use Friendlywrt because it is easier to configure network and router settings. I saw that Apacer AS2280P4 SSD was actually detected, but it was formatted without the box, so I could only use mkfs.ext4 to format it.NanoPi R5S Benchmark
Now we do this by running SBC on the router Bench method to benchmark the CPU, hoping to find some problems as much as possible:
10
1
2
3
4
5
6
7
8
9
$ sudo /bin/bash ./sbc-bench.sh -c
[sudo] password for pi:
WARNING: dmesg output does not contain early boot messages which
help in identifying hardware details.
It is recommended to reboot now and then execute the benchmarks.
Press [ctrl]-[c] to stop or [enter] to continue.
Average load and/or CPU utilization too high (too much background activity). Waiting...
Too busy for benchmarking: 07:21:06 up 3 min, 1 user, load average: 0.41, 0.27, 0.11, cpu: 3%
Too busy for benchmarking: 07:21:11 up 3 min, 1 user, load average: 0.38, 0.26, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:16 up 3 min, 1 user, load average: 0.35 , 0.26, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:21 up 3 min, 1 user, load average: 0.32, 0.25, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:26 up 3 min, 1 user, load average: 0.29, 0.25, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:31 up 3 min, 1 user, load average: 0.27, 0.24, 0.10, cpu: 1%
sbc-bench v0.9.7
Installing needed tools. This may take some time. Done.
Checking cpufreq OPP. Done (results will be available in 20-28 minutes).
Executing tinymembench. Done.
Executing 7-zip benchmark. Done.
Executing cpuminer. 5 more minutes to wait. Done.
Checking cpufreq OPP. MB/s
memset: 6191.5 MB/s (0.2%)
Cpuminer total scores (5 minutes execution): 6.87,6.86,6.85,6.84,6.83,6.82,6.79 kH/s
OpenSSL results:
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-128-cbc 173609.37k 509936.75k 972013.31k 1264387.07k 1383497.73k 1392645.46k
aes-128-cbc 175451.26k 506569.66k 973690.71k 1264628.74k 1382845.10k 1393180.67k
aes-192-cbc 166539.51k 448796.48k 790104.58k 970846.55k 1040621.57k 1046462.46k
aes-192-cbc 168407.31k 451709.25k 792148.91k 970579.63k 1041061.21k 1046375.08k
aes-256-cbc 159430.38k 412822.74k 676804.10k 809129.64k 857347.41k 861137.58k
aes-256-cbc 162313.43k 412763.39k 677746.94k 809317.38k 857642.33k 861334.19k
Unable to upload full test results. Please copy&paste the below stuff to pastebin.com and
provide the URL. Check the output for throttling and swapping please.
I started it almost immediately after booting, so the dmesg output should be complete (see the boot loading earlier in this review for details), but some information is missing from the script.The full output of the sbc-bench.sh script can be found at at pastebin We see that the "1992" MHz advertised frequency is tested in reality as 1845 MHz, so I feel like some optimization could be done here.
NVMe benchmark
I use iozone 3 for NVMe The SSD was tested 3 times, 1 of which was a 100MB file:
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 100M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
Iozone: Performance Test of File I/O
Version $Revision: 3.489 $
Compiled for 64 bit mode.
Build: linux
Include fsync in write timing
O_DIRECT feature enabled
Auto Mode
File size set to 102400 iozone -e -I -a -s 100M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
Output is in kBytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 kBytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
02400 4 34994 53668 30431 30385 30136 59719
02400 16 102031 130543 80174 80125 79796 133162
02400 512 300692 296328 276975 291837 276681 313464
02400 1024 309822 340026 308900 326826 306102 339059
02400 16384 357975 392544 369753 391219 370336 390004iozone test complete.
and then a 500MB file:
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 500M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
and finally a 1GB file:
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 1000M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
024000 4 35105 58082 30395 30447 27458 60895
024000 16 102421 135279 80210 80314 74596 133579
024000 512 300759 314704 282743 283883 277911 313413
024000 1024 329840 337468 318228 319091 317641 337714
024000 16384 383289 385247 382642 382850 382870 381344
After a series of operations, the results are more or less consistent in all three tests without much change.In the end, I got a read-write speed of about 380MB/s, which was far lower than the original promotional writing and reading speed of SSD, and the results of ODROID-M1. I think this is because the PCIe 2.0 X1 interface is used in this design, not the PCIE Gen 3.0 X2 interface used in Hardkernel board.Below is the output of LSPCI, for reference only:
.5GBE interface configuration and benchmark test
TML3
0
1
2
3
4L6 15
PI@FriendlyElec:/Media/NVME0N1 $ Sudo LSPCI -VTML3
Subsystem: Phison Electronics Corporation Device 5013
FLAGS: Bus Master, Fast Devsel, Laten Cy 0, IRQ 87
MEML6Memory at 380900000 (64-bit, Non-Prefetchable) [SIZE = 16K]
capabilities: [80] e0] e0] XPress Endpoint, MSI 00CAPabilities: [ D0] MSI-X: ENABLE+ Count = 9 Masked-
Capabilities: [E0] ENable-Count = 1/8 Maskable+ capabilities: [F8] Power Management Version 3TML3
CAPABILIES: [100] Lanetcy Tolerance l6capabilities: [110] l1 PM SubstateS
CAPABILIES: [200] Advanced Error Reportingml7
CAPABILITIES: [300] Secondary PCI EXPRER ss
kernel driver in use: nvme
ML6NANOPI R5s router
is configured with ETH0 Gigabit Ethernet "WAN" interface. Therefore, we must manually configure two 2.5GBE ports. I used the same test platform with the same "first part of Friendlywrt Evaluation", namely Ubuntu 20.04 laptop. Then, I connected REALTEK RTL8156BG USB 3.0 to 2.5GBE crypto dog to ETH1, UP XTREME I11 mini PC is connected to ETH2. I did not use the bridge interface like in Friendlywrt, but configured two different subnets: ETH1 was 192.168.2.0, ETH2 was 192.168.3.0.
Now we create two new files in /etc/network/interfaces.d/:
- eth1
auto eth1
iface eth1 inet static
address 192.168.2.1
network 192.168.2.0
netmask 255.255.255.0
broadcast 192.168.2.255
- eth2
auto eth2
iface eth2 inet static
address 192.168.3.1
network 192.168.3.0
netmask 255.255.255.0
broadcast 192.168.3.255
Now install the DHCP server
sudo apt install isc-dhcp-server
Edit the /etc/dhcp/dhcpd.conf file using our two subnets:
subnet 192.168.2.0 netmask 255.255.255.0 {
range 192.168.2.100 192.168.2.200;
option routers 192.168.2.1;
}
subnet 192.168.3.0 netmask 255.255.255.0 {
range 192.168.3.100 192.168.3.200;
option routers 192.168.3.1;
}
Before restarting the dhcp server:
sudo systemctl restart isc-dhcp-server
At this point, the laptop and mini PC should be able to obtain their IP addresses from the NanoPi R5S on their respective subnets. Now we can start benchmarking the interface.Download iperf3 using eth1 connected to laptop and then from R5S perspective receive:
0
1
2
3
4
5
6$ iperf3 -t 60 -c 192.168.2.1 -i 10
Connecting to host 192.168.2.1, port 5201
[ 5] local 192.168.2.130 port 59822 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 2.28 GBytes 1.96 Gbits/sec 42 1.41 MBytes
[ 5] 10.00-20.00 sec 2.02 GBytes 1.74 Gbits/sec 0 1.61 MBytes
[ 5] 20.00-30.00 sec 1.72 GBytes 1.48 Gbits/sec 0 1.62 MBytes
[ 5] 30.00-40.00 sec 1.87 GBytes 1.61 Gbits/sec 0 1.62 MBytes
[ 5] 40.00-50.00 sec 1.89 GBytes 1.62 Gbits/sec 0 1.70 MBytes
[ 5] 50.00-60.00 sec 2.06 GBytes 1.77 Gbits/sec 21 1.66 MByte /sec 63 sender
[ 5] 0.00-60.04 sec 11.8 GBytes 1.69 Gbits/sec receiver
iperf Done.
This is better than the 1.85 I got in OpenWrt Gbps is slower, and there is some retransmission of content.During the transfer I also use l sbc-bench.sh monitoring system:
0
1
213
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
The system did run at its advertised maximum frequency during testing, and I didn't see any obvious issues here.I also checked some information and statistics using ethtool:
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ethtool -i eth1
driver: r8125
version: 9.008.00-NAPI
firmware-version:
expansion-rom-version:
bus-info: 0000:01:00.0
supports-statistics: yes
supports-test: no
supports-eeprom-access: no
supports-register-dump: yes
supports-priv-flags: no
pi@FriendlyELEC:~$ sudo ethtool -S eth1
NIC statistics:
tx_packets: 451228
rx_packets: 9569147
tx_errors: 0
rx_errors: 0
rx_missed: 0
align_errors: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 9569102
broadcast: 45
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 31676089
rx_octets: 14506385933
rx_multicast64: 0
tx_unicast64: 451214
tx_broadcast64: 2
tx_multicast64: 12
tx_pause_on: 570
tx_pause_off: 570
tx_pause_all: 1140
tx_deferred: 0
tx_late_collision: 0
tx_all_collision: 0
tx_aborted32: 0
align_errors32: 0
rx_frame_too_long: 0
rx_runt: 0
rx_pause_on: 0
rx_pause_off: 0
rx_pause_all: 0
rx_unknown_opcode: 0
rx_mac_error: 0
tx_underrun32: 0
rx_mac_missed: 31
rx_tcam_dropped: 0
tdu: 0
rdu: 570
I do get some rx_mac_missed.Now let's test it in reverse:
0
1
2
3
4
5
6
7$ iperf3 -t 60 -c 192.168.2.1 -i 10 -R
Connecting to host 192.168.2.1, port 5201
Reverse mode, remote host 192.168.2.1 is sending
[ 5] local 192.168.2.130 port 59826 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.75 GBytes 1.50 Gbits/sec
[ 5] 10.00-20.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 20.00-30.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 30.00-40.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 40.00-50.00 sec 1.94 GBytes 1.67 Gbits/sec
[ 5] 50.00-60.00 sec 1.94 GBytes 1.67 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.04 sec 11.5 GBytes 1.64 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.5 GBytes 1.64 Gbits/sec receiver
iperf Done.
This looks much better than OpenWrt (1.12Gbps).@Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % CPU % SYS % USR % Nice % IRQ TEMP
IRQ percentage is much lower than RX, but I think it is normal for TX.We switch to ETH2:
0
1
2
3
4TML3
6Devkit@UPX -I11: ~ $ iPerf3 -T 60 -C 192.168.3.1 -i10
Connecting T o host 192.168.3.1, port 5201[5] LOCAL 192.168.3.100 Port 37794 Connected to 192.16888 .3.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 10.00-20.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 20.00-30.00 SEC 2.73 GBYTES 2.35 Gbits/SEC 0 1.81 MBytes
[5] 30.00-40.00 SEC 2.73 GBITS/SEC 0 2.90 l3
[5] 40.00-50.00 SEC 2.73 GBYTES 2.35 GBITS/SEC 0 4.37 MBYTES
[5] 50.00-60.00 SEC 2.73 GBYTES 2.35 GBITS/SEC 0 4.37 MBYTES- - - - - - - - - - - --------
Tra on Retr
[5] 0.00-60.00 SEC 16.4 Gbytes 2.35 Gbits/SEC 0 SENDER
[5] 0.00-60.00 SEC 16.4 GBYTES 2.35 GBITS/Secret
IPERF .
Oh, great! This is the first time I have received a 2.35 Gbps transmission speed, so it still looks hopeful! @Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % CPU % SYS % USR % Nice % IRQ TEMP
Connecting to host 192.168.3.1, Port 5201
Unless I get wrong, 25%of IRQ means that a core can be used to handle these.Now we try:@upx -I11: ~ $ iPerf3 -T 60 -C 192.168.3.1 -i 10 -Rreverse Mode, Remote Host 192.168.3.1 is Sending
[5] Local 192.168.3.100 Port 37800 Connect to 192.168.3.1 Port 5201
[ID] Interval Transfer
[5] 0.00-10.00 SEC 1.92 GBYTES 1.65 GBITS/Sec
[5] 10.00-20.00 SEC 1.84 GBYTES 1.58 GBITS/Sec
[5] 20.00-30.00 SEC 1.84 Gbytes 1.58 Gbits/sec[ 5] 30.00-40.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 40.00-50.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 50.00-60.00 sec 1.84 GBytes 1.58 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.01 sec 11.1 GBytes 1.59 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.1 GBytes 1.59 GBITS/SEC Receiver
IPERF DONE.
The result is 1.59, not perfect , But still better than OpenWrt.
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
^C
The CPU is running at full speed again, and it is far from 100% utilization, so I think the problem should lie elsewhere. Now we can use ethtool again to check the eth2 information and statistics.
0
1$ sudo ethtool -i eth2
driver: r8125
version: 9.008.00-NAPI
firmware-version:
expansion-rom-version:
bus-info: 0001:11:00.0
supports-statistics: yes
supports-test: no
supports-eeprom-access: no
supports-register-dump: yes
supports-priv-flags: no
0
1
2
3
4
5
6
7
8
9
sudo ethtool -S eth2
NIC statistics:
tx_packets: 8506609
rx_packets: 12553353
tx_errors: 0
rx_errors: 0
rx_missed: 0
align_errors: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 12553209
broadcast: 144
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 12543719502
rx_octets: 18471602900
rx_multicast64: 0
tx_unicast64: 8503557
tx_broadcast64: 3035
tx_multicast64: 17
tx_pause_on: 35
tx_pause_off: 35
tx_pause_all: 70
tx_deferred: 0
tx_late_collision: 0
tx_all_collision: 0
tx_aborted32: 0
align_errors32: 0
rx_frame_too_long: 0
rx_runt: 0
rx_pause_on: 0
rx_pause_off: 0
rx_pause_all: 0
rx_unknown_opcode: 0
rx_mac_error: 0
tx_underrun32: 0
rx_mac_missed: 335
rx_tcam_dropped: 0
tdu: 0
rdu: 35
In the test results here, I found that there are more rx_mac_missed. So I guess there should be some tweaks to improve performance.But according to my previous experience of adjusting the settings of RTL8156B, adjusting the settings is really tricky, and people with experience can't seem to agree on what setting options to adjust. I mainly refer to the RTL8156/8125 driver. Realtek engineers, as well as some networking experts among our readers.Configuring NAT between two 2.5GbE interfaces
Since the 2.5GbE interface doesn't work optimally with iperf3, I didn't bother testing the router performance in FriendlyWrt, but a few people asked me anyway. So next I will show how I configured NAT in Ubuntu 20.04 and will continue to test NAT performance, keeping in mind that it will definitely improve a lot in a few weeks or months.
Here we need to enable IP forwarding and NAT. The instructions I used were adapted from a post on networkreverse.
Edit /etc/sysctl.conf to enable IP forwarding (uncomment the following lines):
net.ipv4.ip_forward=1
Apply changes:
sudo sysctl -p
Now we enable NAT:
sudo iptables ! -o lo -t nat -A POSTROUTING -j MASQUERADE
We can now ping Xtreme on the 192.168.3.0 subnet from the laptop on the 192.168.2.0 subnet i11:
jaufranc@cnx-laptop-4:~$ ping 192.168.3.100
PING 192.168.3.100 (192.168.3.100) 56(84) bytes of data.
If you want the changes to be permanent, you can execute the following:
sudo apt install iptables-persistent
sudo sh -c 'iptables-save /etc/iptables/rules.v4'
I tried iperf3 between a UP Xtreme i11 and my laptop, the data was routed through a NanoPi R5S router.
0
1
2
3
4
5
6
7
8
9
jaufranc@cnx-laptop-4:~$ iperf3 -t 60 -c 192.168.3.100 -i 10
Connecting to host 192.168.3.100, port 5201
[ 5] local 192.168.2.130 port 59430 connected to 192.168.3.100 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 914 MBytes 767 Mbits/sec 355 1011 KBytes
[ 5] 10.00-20.00 sec 912 MBytes 765 Mbits/sec 324 1.23 MBytes
[ 5] 20.00-30.00 sec 917 MBytes 769 Mbits/sec 124 1.09 MBytes
[ 5] 30.00-40.00 sec 915 MBytes 767 Mbits/sec 150 942 KBytes
[ 5] 40.00-50.00 sec 915 MBytes 767 Mbits/sec 78 1.22 MBytes
[ 5] 50.00-60.00 sec 919 MBytes 771 Mbits/sec 64 1.03 MBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.00 sec 5.36 GBytes 768 Mbits/sec 1095 sender
[ 5] 0.00-60.06 sec 5.36 GBytes 767 Mbits/sec receiver
iperf Done.
jaufranc@cnx-laptop-4:~$ iperf3 -t 60 -c 192.168.3.100 -i 10 -R
Connecting to host 192.168.3.100, port 5201
Reverse mode, remote host 192.168.3.100 is sending
[ 5] local 192.168.2.130 port 59434 connected to 192.168.3.100 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.09 GBytes 935 Mbits/sec
[ 5] 10.00-20.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 20.00-30.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 30.00-40.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 40.00-50.00 sec 1.09 GBytes 939 Mbits/sec
[ 5] 50.00-60.00 sec 1.09 GBytes 937 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.05 sec 6.55 GBytes 937 Mbits/sec 973 sender
[ 5] 0.00-60.00 sec 6.55 GBytes 937 Mbits/sec receiver
iperf Done.
has a transmission speed of 768 Mbps in one direction and 937 Mbps in the other direction.@Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % cpu % sys % usr % nice % IRQ TEMP
1: 48: 1992MHz 0.90 25% 0.90 25% 0 % 0% 0% 0% 0% 25% 54.4 ° C
H6
If you use SBC-Bench.sh, the running frequency will be 1992 (Actually 1845 MHz), and 25%IRQ should be just It means that a Core has been completely used to deal with IRQ.is displayed according to the MPSTAT command, which should be processed by Core #0.
.5GBE interface configuration and benchmark test
l3
TML3H0H0
$ MPSTAT -P All -I Sum
Linux 5.10.66 (F Riendlyelec) 06/05/22 _aarch64_ (4 cpu)
can be confirmed by using TOP and HTOP.
NANOPI
I just have a wall power meter. 6 Free -5.1 W
IPERF3 to ETH1 - 6.3 to 6.7w NAT test -6.2Wht between laptop and mini computers Ml7 It was tested after installing Ubuntu 20.04 and NVMe SSD on this product.
I am still using OpenWrt and without SSD NANOPI R5s for testing:
free L32IPERF3 - 6.0 to 6.2 W Note that because I use wall -mounted power meters, this value will include the power adapter (Khadas Vim4 usb. -C power adapter) efficiency loss. And this value may be higher than the value of using USB-C power timing. This should also be optimized by setting.
The final conclusion
's evaluation today is here. The optimization part should actually include "changing the firmware to make the RockChip kernel run in 1992 MHz", "Adjustment of various settings related to PCIe and Ethernet settings" and so on. Most of them are not very familiar.
(3), THT Linglong M NUC brush X86 soft route evaluation
foreword
many friends to eat X86 soft routes, or they are ready to step into X86 soft routing pit! First of all, it is necessary to clarify whether you are willing to spend time to toss. You can consider buying the finished OPNENWRT router to experience it. You can also buy a router that brushes the Merlin firmware or brush the PANDORABOX (Pandora) firmware to play.prepares related network devices
a suitable 2.5G port switch and X86 soft route with a 2.5G power port, a Type-C end of a 2.5G mesh or external wired network card at USB-A end.
TSMC Linglong M NUC is a mini machine that is very suitable for X86 soft routing. It has two 2.5GE power outlets, two SO-DIMM memory slots and a PCIE 3.0 NVME SSD M.2 slot. It is recommended to buy bare metal, bring up memory and NVMe SSD. Here are two 4GB Micron DDR4 3200MHz memory bar and 1 Intel Optane M10 16GB (Ao Teng Memory).
is installed with memory and NVME SSD.
can be used to use it.
Xike is a new network equipment brand. This is a two -layer management switch with 8 2.5G electric ports and 4 million -dimensional light ports to support port aggregation and VLAN division.
Friends bought this Xike switch to help me upgrade the network of my home, intending to be a 2.5G internal network environment in the whole house.
begins to adjust the device, which is still very simple for experienced people.
can see that the 10G optical port of the XIKE SKS7300-8GPY4XGS switch has been turned up, and the connection status is 10G Full rate. For friends who have 10G SFP modules, they can use 10G LAN. However, at this stage, 2.5G LAN is more suitable for the actual experience needs of the public. If you upgrade to a 10G optical port, the overall hardware cost will be significantly higher several times. For small and medium-sized enterprises, this switch is very suitable for use as an access switch connected to a 10G core switch. It is still more recommended for home users who are in need of 2.5G electrical ports and friends who like to mess with 2.5G LAN to buy this XIKE switch.
For users who have a GPON Stick ONU module, plugging it directly into the SFP interface of this SKS7300-8GPY4XGS switch can save an optical modem. It effectively solves the problem of insufficient accumulation of equipment in the weak current box, and can also improve the space inside the weak current box and reduce the problem of heat accumulation between devices.
What should I do if my computer (desktop and laptop) does not have a 2.5G network port? You can buy a superior USB to RJ45 2.5G wired network card. Connect to the computer through the Type-C end, and the RJ45 network port on the other end has a 2.5GbE (2500Mbps) rate.
Take a USB flash drive to make a boot disk
I generally like to use the tool BalenaEtcher to make a USB flash drive. It is simple and easy to operate! Supports ISO, DMG, IMG, IMG.GZ and other image file formats to create boot disks.
makes the Sea Spider V9 boot disk (flash)
. Go to the Sea Spider official website and find the download area in the service and support. Scroll to the upgrade address at the bottom, find V9.0, and select Download Now next to it to download the Sea Spider V9 firmware. Note that the ISO image file is downloaded.
Sea Spider V9 system
default login address: 192.168.0.1. The account and password are set by yourself after flashing the machine.
You can view the current information overview on the overview. The instrument icon shows CPU usage, memory usage, and disk usage. Below is the monitoring of the real-time total traffic of the intranet, and on the right is some key information. The LAN port already in use by
can identify the network port model: Realtek RTL8125 2.5GbE. The network port icon also identifies 2.5Gb.
Nigou can view the current memory usage in the overview. After flashing and initializing the system, you can see that 10.5% is used by default and 835.86 MB of storage space has been used. Total memory size: 7.6GB (8GB total memory), remaining 6.8GB available.
You can also view the CPU load in the overview. The
interface load can also view the data of the LAN port. The LAN port here is running in full-duplex mode, 2500Mb/s (that is, the 2.5GbE Gigabit port rate).
You can see the motherboard/CP information in the hardware information. Basically, the relevant information displayed is complete.
Here is the 16GB Optane memory, available capacity: 13.41GB, you can check the partitioning of the hard disk storage.
Both 2.5GbE Gigabit network ports can be recognized normally.
Haispider has module management, including Docker container modules.
Basically these modules are very complete for soft routing, including VPN IPsec, PHP suite, IGMP multicast proxy, etc.
FTP service.
file storage and sharing.
Comment: Sea Spider V9 requires purchasing relevant authorizations to fully use all functions. Fortunately, there is a long trial period. Relatively many functions are more suitable for network workers. For ordinary novice users, the operation may be very difficult, and some professional terms are convoluted and difficult to understand.
Make iStore OS boot disk (flash)
Go to KoolCenter official website to download iStore OS firmware. Find the iStore OS directory in the firmware list and enter it.
Considering that the current new X86 soft routing is booted by UEFI, so select X86_64_EFI to download the iStore OS firmware.
Generally has Date (Date) in the bottom list. Select the latest Istore OS firmware.
istore OS
In fact, ISTORE OS is the in -depth customization of Chinese people. Compared with the original OpenWrt, the gameplay (plug -in) and Docker gameplay.
homepage will have a legend with traffic statistics, IP addresses, and network interface status. It looks clear at a glance.
pulls down to see disk information, storage services, docker, download services and remote domain names. The playability is still very high.
can be seen at the bottom to see the system information. The dog recognizes the CPU temperature, CPU usage rate, and memory usage.
network wizard, this is the best direction function for Xiaobai users.
has system information. You can see the computer model, processor architecture and other related information.
can also see the use of memory at the overview.
Soft Route is difficult to master a disk management. Here ISTORE OS directly completes a graphical Diskman disk management, which reduces the difficulty of novice operation and enhances the simpleness of use.
istore OS built -in Docker
is definite It is difficult to install plug -ins. Many beginners or little white users who have just started. The iStore OS directly built -in Docker components for easy deployment of containerization. And supporting CPU and memory configuration settings in advanced settings, which is very user -friendly.
ISTORE OS Unique App Store
For wanting to study the smart home experience, you can consider Home Assistant.
is easy to be suitable for miniature home deployment data service centers. Many tutorials on the Internet will not be discussed in depth.
If you can't afford a dedicated NAS device, in fact, soft routes can also act as simple NAS, but most of the X86 soft routes are not easy to use. There are two tools that can be solved directly here to solve this problem. For friends who want to build an audio -visual center: Jellyfin and NAS Tools. If you can use it, it can definitely easily make the X86 soft routing into a NAS.
ARIA2 downloader. If you can use it, it is also a very practical download tool. For the two broadband at home, in order to avoid the broadband during the work or during the business trip, try to deploy the net Xinyun to earn some benefits.
I actually prefer HOMEBOX inner mesh speed measuring. This is better experienced by the Speedtest speed measuring server. I will show the experience of the HOMEBOX intranet speed measuring experience later.
DISKMAN disk management is definitely a plug -in worth installed. Friends who use QBITTORRENT downloaders will use it to use it. The specific use methods and tutorials have more detailed professional introduction on the tutorial network. If you are interested in the research, go to the study.
KMS server can solve the problem that Windows and Office prompts not activated.
installation plug -in is as simple and convenient as installing the app, which is also the biggest ease of use of Isstore OS.
I use most of these four plug -ins. Among them, multi -line multi -broadcast is suitable for broadband overlay areas. One sentence can run to 600m in one sentence, 500M superposition can run 1000m but do not need to upgrade broadband packages and extra money. It is worthwhile.
KMS server is enabled. Basically, local Windows and Office have never encountered unauthorized situations.
This is the HOMEBOX intranet speed measurement, which can directly measure the rate of internal network. I use Taipower Linglong M NUC's 2.5GBE Gigabit Gigabit Port, CAT6 (Super Six) network cables, Type-C to RJ45 2.5GBE external wired network card, which can achieve a 2500Mbps full-link local area network (internal network) experience Essence
summarizes
for users who have the foundation of the network or understand the network. Sea Spider V9 is actually more inclined to traditional network devices, combining switches, routers, AC controllers and other functions. ISTORE OS is suitable for the general public, has a higher playability, and it is more difficult to get started. I personally recommend brush IStore OS firmware to use, basically it is the personal OS person who has experienced a good experience in OpenWrt.
In the end, I got a read-write speed of about 380MB/s, which was far lower than the original promotional writing and reading speed of SSD, and the results of ODROID-M1. I think this is because the PCIe 2.0 X1 interface is used in this design, not the PCIE Gen 3.0 X2 interface used in Hardkernel board.Below is the output of LSPCI, for reference only:
TML3
0
1
2
3
4L6 15
PI@FriendlyElec:/Media/NVME0N1 $ Sudo LSPCI -VTML3
Subsystem: Phison Electronics Corporation Device 5013
FLAGS: Bus Master, Fast Devsel, Laten Cy 0, IRQ 87
MEML6Memory at 380900000 (64-bit, Non-Prefetchable) [SIZE = 16K]
capabilities: [80] e0] e0] XPress Endpoint, MSI 00CAPabilities: [ D0] MSI-X: ENABLE+ Count = 9 Masked-
Capabilities: [E0] ENable-Count = 1/8 Maskable+ capabilities: [F8] Power Management Version 3TML3
CAPABILIES: [100] Lanetcy Tolerance l6capabilities: [110] l1 PM SubstateS
CAPABILIES: [200] Advanced Error Reportingml7
CAPABILITIES: [300] Secondary PCI EXPRER ss
kernel driver in use: nvme
ML6NANOPI R5s router
is configured with ETH0 Gigabit Ethernet "WAN" interface. Therefore, we must manually configure two 2.5GBE ports. I used the same test platform with the same "first part of Friendlywrt Evaluation", namely Ubuntu 20.04 laptop. Then, I connected REALTEK RTL8156BG USB 3.0 to 2.5GBE crypto dog to ETH1, UP XTREME I11 mini PC is connected to ETH2. I did not use the bridge interface like in Friendlywrt, but configured two different subnets: ETH1 was 192.168.2.0, ETH2 was 192.168.3.0.
Now we create two new files in /etc/network/interfaces.d/:
- eth1
auto eth1
iface eth1 inet static
address 192.168.2.1
network 192.168.2.0
netmask 255.255.255.0
broadcast 192.168.2.255
- eth2
auto eth2
iface eth2 inet static
address 192.168.3.1
network 192.168.3.0
netmask 255.255.255.0
broadcast 192.168.3.255
Now install the DHCP server
sudo apt install isc-dhcp-server
Edit the /etc/dhcp/dhcpd.conf file using our two subnets:
subnet 192.168.2.0 netmask 255.255.255.0 {
range 192.168.2.100 192.168.2.200;
option routers 192.168.2.1;
}
subnet 192.168.3.0 netmask 255.255.255.0 {
range 192.168.3.100 192.168.3.200;
option routers 192.168.3.1;
}
Before restarting the dhcp server:
sudo systemctl restart isc-dhcp-server
At this point, the laptop and mini PC should be able to obtain their IP addresses from the NanoPi R5S on their respective subnets. Now we can start benchmarking the interface.Download iperf3 using eth1 connected to laptop and then from R5S perspective receive:
0
1
2
3
4
5
6$ iperf3 -t 60 -c 192.168.2.1 -i 10
Connecting to host 192.168.2.1, port 5201
[ 5] local 192.168.2.130 port 59822 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 2.28 GBytes 1.96 Gbits/sec 42 1.41 MBytes
[ 5] 10.00-20.00 sec 2.02 GBytes 1.74 Gbits/sec 0 1.61 MBytes
[ 5] 20.00-30.00 sec 1.72 GBytes 1.48 Gbits/sec 0 1.62 MBytes
[ 5] 30.00-40.00 sec 1.87 GBytes 1.61 Gbits/sec 0 1.62 MBytes
[ 5] 40.00-50.00 sec 1.89 GBytes 1.62 Gbits/sec 0 1.70 MBytes
[ 5] 50.00-60.00 sec 2.06 GBytes 1.77 Gbits/sec 21 1.66 MByte /sec 63 sender
[ 5] 0.00-60.04 sec 11.8 GBytes 1.69 Gbits/sec receiver
iperf Done.
This is better than the 1.85 I got in OpenWrt Gbps is slower, and there is some retransmission of content.During the transfer I also use l sbc-bench.sh monitoring system:
0
1
213
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
The system did run at its advertised maximum frequency during testing, and I didn't see any obvious issues here.I also used EthTool to check some information and statistical information: @friendlyelec: ~ $ Sudo ETHTOOL -I ETH1TML3DRIVER: R8125
Version: 9.008.00-napi
FIRMWARE-VERSION: TML3EML6Expans:
BUS-INFO: 0000: 01: 00.0
SUPPPPOTS -StATISTICS: YESTML7SUPPPOS-TEST: no
SUPPPROM-ACCESS: NO
supports-register-dump: YES
SUPPOPPORTS-FLAGS: no
PI@Friendlyelec: ~ Sudo ETHTOOL -S ETH1NIC Statistics:
TX_PACKETS: 451228
RX_PACKETS: 95 69147TX_ERRORS: 0
RX_ERRORS: 0RX_MISSED: 0l3align_error: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 9569102
broadcast: 45
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 31676089
rx_octets: 14506385933
rx_multicast64: 0
tx_unicast64: 451214
tx_broadcast64: 2
tx_multicast64: 12
tx_pause_on: 570
tx_pause_off: 570
tx_pause_all: 1140
tx_deferred: 0
tx_late_collision: 0
tx_all_collision: 0
tx_aborted32: 0
align_errors32: 0
rx_frame_too_long: 0
rx_runt: 0
rx_pause_on: 0
rx_pause_off: 0
rx_pause_all: 0
RX_UNKNOWN_OPCODE: 0
RX_MAC_ERROR: 0htmmmmm l7tx_underrun32: 0
RX_MAC_MISSED: 31
RX_TCAM_DROOPED: 0tdu: 0
rdu: 570
Really got some RX_MAC_MISSED.Now let's test it in reverse:
0
1
2
3
4
5
6
7$ iperf3 -t 60 -c 192.168.2.1 -i 10 -R
Connecting to host 192.168.2.1, port 5201
Reverse mode, remote host 192.168.2.1 is sending
[ 5] local 192.168.2.130 port 59826 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.75 GBytes 1.50 Gbits/sec
[ 5] 10.00-20.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 20.00-30.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 30.00-40.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 40.00-50.00 sec 1.94 GBytes 1.67 Gbits/sec
[ 5] 50.00-60.00 sec 1.94 GBytes 1.67 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.04 sec 11.5 GBytes 1.64 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.5 GBytes 1.64 Gbits/sec receiver
iperf Done.
This looks much better than OpenWrt (1.12Gbps).@Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % CPU % SYS % USR % Nice % IRQ TEMP
CPU is the memory and html. 6 flash memory chip .
IRQ percentage is much lower than RX, but I think it is normal for TX.(1), Electric Rhino R68S Soft Route Evaluation
I. Appearance and key configuration Analysis of
.1 R68S Key ConfigurationElectric Rhino R68S RK-3568 ARM Processor, 22nm process, 4 -core Cortex A55 Architecture, main frequency 2GHz, supports the maximum 8G memory;
RK3568 built -in 1 QSGMII, 2 PCIE 3.0x1 channels, also with 2 Gigabit Ether interfaces, so you can use 42.5GB+2x1000MB 4 4 Gigabit Ether interface. Net port combination.
R68S core configuration is shown in the figure below:
HTM core configuration
LPDDR4 memory chip is from Rayson, the model is RS1G32LF4D2BDS- 53BT, the capacity is 4GB (2GB version is 70 yuan); Resee (under Shenzhen Jiang Bolong), the model is NCEMASKGG-16G, the capacity is 16GB;
There is a Gigabit Ethernet exchange chip in the upper left corner of the processor with a model of RTL8211F. When you see this crab logo, you know that from Riyu ;
does n’t there be 4 mesh ports? What about the remaining 3 exchanges?
is on the back of the motherboard.
Two 2.5GB exchange chips also comes from Riyu, the model is RTL8125BG:
motherboard
Ruyu's 8125B, Intel 's I225 B3 and Qualcomm QCA8081 are currently the most common.2 R68s Appearance.1 encryption and decryption performanceengineering machine is sent, there is only one piece of bare plate.
is because I want to do the temperature test in weak electric box , so the boss sent a shell after a few days (for test model, non -final version).
I received is made of plastic material (also CNC aluminum shell, the price will be about 100 yuan), and there are a lot of heat dissipation holes.
official version shell will be made into a frosted effect
HT There is no dust -proof cotton, so if you use R68S to put the desktop, you still need to consider the problem of dust. After all, dust is the first of electronic equipment. One killer:
front
HTM has a large amount of heat dissipation holes in the two sides of the 2 side:
machine side HT
If you will put R68s in a weak electric box like me, then the hollow design of the shell is actually very beneficial to heat dissipation.
But if you put R68S on the desktop, I also recommend buying a CNC aluminum shell, that is, dustproof and waterproof, heat dissipation is okay.
cnc shell is about length. Single version.
I actually tested it, and the 12V 1A light cat power adapter can also work normally.
Of course, if you have the need to connect to external hard disks, I suggest that it is best to put a 12V 1.5A power adapter.
When it comes to external hard disks, the USB 3.0 performance of R68S is still very strong. Reading and writing can run to about 120MB. If you want to watch movies on the R68S, you don't have to worry about the original disk high -code movie will be stuck.
2. Performance and stability Test
For a soft routing , performance and stability are equally important.
directly look at the table:
Zhihu@仙鱼
CPU CoreMARK
AES...
CHA...
AES hard decoding
R68S (RK3568)
03101
is
N1(S905D)
3
8704
is
2 R4S (RK3399)
36089
is
J1900
no
J4125
51102is
The encryption and decryption performance of R68S is much stronger than the N1 of ARM architecture , but it lags behind the 6-core R4S (RK3399) by about 20%. The CPU CoreMARK of
R68S is about 6000 points behind J1900. However, J1900 does not support hardware AES decoding after all, and R68S still has some advantages in this regard. If
is compared with J4125, which also supports AES hard decoding, R68S lags behind by more than half.
Although the R68S is not excellent in terms of scores, in actual use, the performance is more than sufficient.
.2 NAT performanceNAT performance mainly tests the broadband speed. The tested topology is as follows:
NAT test
The test rate is as follows:
It can be seen that if the transmission loss is removed, the R68S has reached the limit of Gigabit broadband.
At this time, the CPU load is around 45%.
.3 LAN forwarding performanceLAN forwarding capability mainly tests the forwarding rate between the two 2.5G LAN ports of the R68S. I did a total of 3 different tests. The topologies of
R68S and iPerf3 client installed on the computer respectively, ThinkPAD X13 (2.5G USB 3.0 network card) is directly connected to the 2.5G network port of R68S. The test results are:
iPerf3
The downlink rate is close to 2.1Gb, which may be limited by the performance of the USB network card. The uplink rate is slightly worse at 1.94Gb.
At this time, the CPU load is about 30%.
LAN TO The downlink rate of LAN
can reach around 2.3Gb, which is close to the limit, but the uplink rate still lags behind. In multiple tests, the fastest rate only reached 1.85Gb.
At the same time, I also tried large file transfer: the reading rate of
can be stable at 210MB, and the peak value can reach 242MB.
However, the writing speed is relatively unstable and fluctuates around 150MB. This may be due to a bottleneck in the hard disk speed, or the network card uplink is inherently slow.
LAN file transmission speed
HT The AX200 network card of ml0thinkpad X13 can achieve 2402MB negotiation rate under Redmi AX5400 e -version router, and the laptop computer uses wifi most of the time, so LAN TO WIFI6 test is also necessary:
lan to wifi6
AX5400 direct connecting NAS, the maximum rate of WIFI6 is 1850/6 24. After the R68S repost, the rate decreases slightly, but the decrease is 5%, which is acceptable.
Overall, R68S's LAN forwarding ability is still quite possible. 2.5G switch.
, if it is used in R68S DOCKER, the resumption of the local area network bridge will be greatly reduced. The firmware of the TML6R68S is fooled by the LEAN God, and it is used for 10 days Judging from the situation, stability is still very good.
running time
R68s I have been thrown in a weak electric box.
At present, Zhejiang's indoor temperature is almost about 27 degrees, and the R68S nude standby temperature is about 47 degrees. After the
shell arrived, I observed that the down temperature was soaring to about 53 degrees.
But the temperature has risen not only R68S. In the past two days, my Redmi AX5400 router, even if it is placed on the TV cabinet, the shell is already very hot. When
, when high temperature, the indoor temperature may be about 31 degrees, the temperature of the weak electric box will be higher, and I will continue to update the temperature monitoring of the R68S.
Zhejiang has been in high temperature for more than 40 days, and the indoor temperature is basically about 32 degrees. The R68S temperature of
is basically about 58 degrees in the weak electric box, and it is still very stable. It has not occurred and has not occurred.
三、玩机教程
这里主要讲一下刷机和部分插件安装的教程
.1 重置与刷机R68S带有重置按钮，如果系统混乱了，或者进不去路由器设置页面了，那么按住重The set button can restore the factory settings (under plug -in), which is still very convenient.
has to blow the R68S flashing machine, which can be completed in a few seconds. Compared with the N1 flashing, it is really different.
, but convenient and convenient, the tutorial still has to be learned.
) Install Risin Microcarmament and flashing toollink: https: //pan.baidu.com/s/1ptxq1yfjytccclhcy068aw? PWD = 6tnl extraction code: 6t NL
firmware Select
flashing order
can hear the prompt sound on the computer after about two seconds, and the device will be prompted on the development tool.
flashing machine
.2 upgrade. Tong's update is enough to use web upgrade:
web upgrade
web The upgrade requires the "Sysupgrade" image. The file suffix name is generally TAR.
.2 plug -in installationmachine is built -in pure version of the firmware. If you need to use certain firmware, you can upload IPK or SH command installation.
. For example, the Alibaba Cloud Plate WEBDAV plug -in that the audio and video player is very needed is not built in the factory. We can use the SSH tool or OPENWRT Tyyd to enter the command installation (replicate the car one by one):
wget https://github.com/messense /Aliyundrive-BDAV/releases/download/v1.3.2/aliyundrive-webdav_1.3.2-1_Aarch64_gneric.ipkwget https://github.com/messense /Aliyundrive-Webdav/releases/download/v1.3.2/luci- app-aliyundrive- webdav_1.3.2_all.ipkwget https://github.com/messense/aliyundrive-webdav/releases/download/v1.3.2/luci-i18n -webdav-zh- cn_1.3.2-1_all.ipkopkg Install Aliyyundrive-Webdav_1. 3.2-1_AARCH64_GENERIC.IPKOPKGG Install Luci-ALIYUNDrive-Webdav_1.3.2_ALL.IPKOPKG Install Luci-I18n-WEBDAV-ZH-CN_1.3.2-1_ All.ipkis refreshed after installation is complete, and the service has plug -in:
Alibaba Cloud Plate4. Summary and purchase suggestions
Electric Rhino R68S soft routing, which is very suitable for me.
is because I already have the finished NAS with a 2.5G network port. The router usually uses an AP mode, so for me, you only need to have a soft route with 2.5G mesh.
The current soft routing market is very rolled, and the 2.5G soft route of J4125 has also reduced the price to about 600 yuan. If you have more needs, such as love fast diversion+OpenWRT bypass+UNRAID's standard All in one gameplay, then R68S may be possible Can't meet your requirements.
But if you just need a pure soft route, then R68S is still very suitable for you.
(2), NANOPI R5s soft routing evaluation
NANOPI R5s
with metal shell ssd and sanoPi R R 5S
I found that the router has been assembled. They also accompanied 6 rubber foot pads and a section of
NANOPI R5STML12
NANOPI is there. With a USB-C port for power supply, a WIFI antenna hole (this antenna hole can also insert GPIO, UART CONSOLE and other cables), two 2.5 2.5 GBE RJ45 LAN port, a Gigabit Ethernet WAN port and HDMI video output .
NANOPI R5s cover key
We will find one for solidarity The cover button for upgrade, four on the front panel for LED lights for "system" and Ethernet ports, and two USB 3.0 port.NANOPI R5s Disassemble
. Generally, the main reason for disassembling is the following:: Out of curiosity, to install M.2 NVMe, to welded SPI flash memory, you need to connect some GPIO, RTC battery, or adjust the UART to TTL to TTL Test plate. If you want to open it, it is easy to open. You can easily disassemble the four screws that need to be released.
can see the bottom of the circuit board with the M.2 KEY M slot, the SPI flash memory occupation space (right), and the Samsung KLM8G1GETF-B041 EMMC 5.1 flash memory 8GB.
NANOPI R5S SPI flash memory and M.2 socket
need to remove the four screws before taking the circuit board from the shell.
NANOPI R5S Sbc
R6RAYSON RS512LM4 D2BDS is a 2GB LPDDR4X memory chip. RTL8211F (GBE) and
NanoPi R5S with a metal casing that acts as a CPU cooler.
Test setup using 2.5GbE USB dongle and UP Xtreme i11 mini PC
You may remember that I had some performance issues with not too long ago when using the RTL8156B USB dongle , but now this issue has been resolved. Because Realtek sent me another RTL8156BG, I was previously using iperf3 on my laptop from UP Xtreme When i11 mini PC transmits data and performs full duplex test, its test result is 2.34Gbps/2.29Gbps.
I am now using the same settings for testing, but this time the middle stuff is replaced with NanoPi R5S.
NanoPi R5S
TP-Link switch here is only used as a Gigabit Ethernet switch , and the NanoPi is connected through Xiaomi AX6000 router The R5S router's WAN port connects to the internet, which prevents me from having to install some software packages on the device. While placing the NanoPi R5S on top of the TP-Link switch resulted in better photos, both devices get very hot and I don't recommend this. Since my Ethernet cable was very short, I had to move the router to a table to test it. The
OpenWrt and iperf3 benchmarks
FriendlyWrt come pre-installed on the router, so they are ready to use right out of the box. It is also possible to immediately access the LuCI interface or SSH using "root" as user and "password" as password. This is indeed very convenient, but it is not very safe, and may be against the law in some countries. Anyway, it is best to change the password at least the first time you use it. The status of
FriendlyWrt
FriendlyWrt is based on the kernel of OpenWrt 22.03.0-rc1 and Linux 5.10.66. It uses less than 250MB of RAM when idle and using default settings, which is more than enough since the system comes with 2GB of RAM.
FriendlyWrt network
I haven't connected SSD yet, so I only mounted the root partition. The available memory is 6.7 GB, and the root partition uses 920 KB. All interfaces can correctly obtain IP addresses through DHCP when starting up. Devices on the LAN can also be accessed through hostname.lan and obtain some IPv6 addresses.
Start from the IPERF3 benchmark test. I first run the "iPerf3 -S" on the NANOPI R5S and run the following commands on the laptop:
download: (NANO NANO NANO View the results of the RX test on PIR5S) iperf3 -T 60 -C 192.168.2.1 -i 10Connecting to host 192.168.2.1, port 5201[ 5] local 192.168.2.130 port 48782 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate Retr Cwnd[ 5] 0.00-10.00 sec 2.23 GBytes 1.92 Gbits/sec 0 1.69 MBYTES [5] 10.00-20.00 SEC 2.02 GBYTES 1.74 Gbites/SEC 0 1.69 MBYTES [5] 20.00-30.00 SEC 2.33 GBITS/SEC 0 2.64 MBYTES [5] 30.00-40.00 SEC 1.66 ES 1.42 GBITS/SEC 0 2.64 MBYTES [ 5] 40.00-50.00 sec 2.62 GBytes 2.25 Gbits/sec 0 2.64 MBytes[ 5] 50.00-60.00 sec 2.01 GBytes 1.73 Gbits/sec 0 2.64 MBytes- - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr [5] 0.00-60.00 SEC 12.9 GBITES 1.84 GBITS/SEC 0 Sender [5] 0.00-60.05 SEC 12.9 Gbites/Secits/Secits E.
- (on NANOPIR5S Check the results of the TX test):
$ iPerf3 -T 60 -C 192.168.2.1 -RConnecting to host 192.168.2.1, Port 5201Reverse Mode, Remote Host 192. 168.2.1 is sending [5] LOCAL 192.168.2.130 Port 48786 Connect TO 192.168.2.1 port 5201 [ID] Interval Transfer Bitrate [5] 0.00-10.00 SEC 1.29 GBYTES 1.11 GBITS/Secain [5] 10.00-20.00 SEC 1.31 GBITS/SEC [5] 20.00-30 .00 SEC 1.33 GBYTES 1.14 GBITS/ sec[ 5] 30.00-40.00 sec 1.27 GBytes 1.09 Gbits/sec[ 5] 40.00-50.00 sec 1.30 GBytes 1.12 Gbits/sec[ 5] 50.00-60.00 sec 1.30 GBytes 1.12 Gbits/sec- - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr [5] 0.00-60.06 SEC 7.80 GBYTES 1.12 GBITS/Secits [5] 0.00-60.00 SEC 7.80 GBITS/SEC Receiverf Done.
is not exactly the above 2.35 GBPS and 1.85 Mbps when it is propagated by Friendlyelec, because in this configuration, I see 1.84 Gbps and 1.12 Gbps. When viewing the 10 -second transmission test report, RX has also changed. Fortunately, no problem was found.Now we try the same with another WAN port on NanoPi R5S and run the command from UP Xtreme i11:
- Download (Rx):
devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.2.1 -i 10Connecting to host 192.168.2.1, port 5201[ 5] local 192.168.2.207 port 52052 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate Retr Cwnd[ 5] 0.00-10.00 sec 2.59 GBytes 2.22 Gbits/sec 0 1.67 MBytes[ 5] 10.00-20.00 sec 2.62 GBytes 2.25 Gbits/sec 0 1.93 MBytes[ 5] 20.00-30.00 sec 2.60 GBytes 2.24 Gbits/sec 0 1.93 MBytes[ 5] 30.00-40.00 sec 2.47 GBytes[ 5] 50.00-60.00 sec 2.45 GBytes 2.10 Gbits/sec 0 4.90 MBytes- - - - - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr[ 5] 0.00-60.00 sec 15.2 GBytes 2.17 Gbits/sec 0 sender[5] 0.00-60.00 sec 15.1 GBytes 2.17 Gbits/sec receiveriperf Done.
- upload (Tx);
devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.2.1 -i 10 -RConnecting to host 192.168.2.1, port 5201Reverse mode, remote host 192.168.2.1 is sending[ 5] local 192.168.2.207 port 52056 connected to 192.168.2.1 port 5201[ ID] Interval Transfer Bitrate[ 5] 0.00-10.00 sec 1.31 GBytes 1.13 Gbits/sec[ 5] 10.00-20.00 sec 1.29 GBytes 1.11 Gbits/sec[ 5] 20.00-30.00 sec 1.32 GBytes 1.14 Gbits/sec[ 5] 30.00-40.00 sec 1.30 GBytes 1.11 Gbits/sec[ 5] 40.00-50.00 sec 1.33 GBytes 1.14 Gbits/sec[ 5] 50.00-60.00 sec 1.27 GBytes 1.09 Gbits/sec- - - - - - - - - - - - - - - - - - - - - - - - - -[ID] Interval Transfer Bitrate Retr[5] 0.00-60.00 sec 7.82 GBytes 1.12 Gbits/sec 0 sender[5] 0.00-60.00 sec 7.82 GBytes 1.12 Gbits/sec receiveriperf Done.when downloading (Rx) at 2.17 The Gbps operation seems to be better; but when uploading (Tx), it is still running very slow at 1.12 Gbps.
Let’s take a look at the speed when using two WAN ports at the same time? One is running "iperf3 on UP Xtreme i11 -s", the other port is to run the following command on the notebook:
0
1
2
3
4
5
6jaufranc@cnx-laptop-4:~$ iperf3 -t 60 -c 192.168.2.207 -i 10
Connecting to host 192.168.2.207, port 5201
[ 5] local 192.168.2.130 port 49762 connected to 192.168.2.207 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 1.39 GBytes 1.19 Gbits/sec 0 3.15 MBytes
[ 5] 10.00-20.00 sec 1.42 GBytes 1.22 Gbits/sec 0 3.15 MBytes
[ 5] 20.00-30.00 sec 1.44 GBytes 1.24 Gbits/sec 0 3.15 MBytes
[ 5] 30.00-40.00 sec 1.42 GBytes 1.22 Gbits/sec 0 3.15 MBytes
[ 5] 40.00-50.00 sec 1.42 GBytes 1.22 Gbits/sec 0 3.15 MBytes
[ 5] 50.00-60.00 sec 1.39 GBytes 1.19 Gbits/sec 0 3.15 MByte /sec 0 sender
[ 5] 0.00-60.05 sec 8.48 GBytes 1.21 Gbits/sec receiver
iperf Done.
This data should be within the expected range, because the upload (Tx) data we obtained before was relatively low; at 1.21 Gbps, it is still a little higher than the 1.12 Gbps when we only used upload (Tx) .Let’s try again in reverse:
$ iperf3 -t 60 -c 192.168.2.207 -i 10 -RConnecting to host 192.168.2.207, port 5201Reverse mode, remote host 192.168.2.207 is sending[ 5] local 192.168.2.130 port 49766 connected to 192.168.2.207 port 5201[ ID] Interval Transfer Bitrate[ 5] 0.00-10.00 sec 2.04 GBytes 1.75 Gbits/sec[ 5] 10.00-20.00 sec 2.04 GBytes 1.75 Gbits/sec[ 5] 20.00-30.00 sec 2.04 GBytes 1.75 Gbits/sec[ 5] 30.00-40.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 40.00-50.00 sec 2.05 GBytes 1.76 Gbits/sec[ 5] 50.00-60.00 sec 2.05 GBytes 1.76 Gbits/sec- - - - - - - - - - - - - - - - - - - - - - - - - - - - [ID] Interval Transfer Bitrate Retr[ 5] 0.00-60.05 sec 12.3 GBytes 1.75 Gbits/sec 0 sender[ 5] 0.00-60.00 sec 12.3 GBytes 1.76 Gbits/sec receiveriperf Done.
This time it was 1.75 Gbps, I have no idea what happened or how this happened.I then ran the 600 second test again using full duplex and checked the CPU, memory and temperatures reported in LuCI.
The CPU load of NanoPi R5S
can be seen from the above that its CPU load is between 2.0 and 2.5.
Temperature of NanoPi R5S
The utilization of CPU #0 is relatively low, and CPU #1 to #3 occupy most of the resources.
never went above 60°C in a room with an ambient temperature of 28°C.
FriendlyWrt Memory Usage
For some unknown reason, the available memory usage of 1.8GB has hardly changed. Performance during full-duplex testing was: 1.66 Gbps and 736 Mbps.The result of
is a bit disappointing. I am now intending to switch to Friendlycore (Ubuntu Core) to check whether similar results will be obtained, and then perform further testing. If you want me to evaluate other contents of OpenWrt, you can tell me in the comment area.
is further tested to switch to the FriendlyCore image based on Ubuntu 20.04, because I am more familiar with the Debian -based operating system, and some tools cannot run on OpenWRT. Note that the performance of the test results is still not ideal. This is why I call the prediction test. With the use and adjustment software of more and more people, the situation should be improved in the next few months.
OPENWRT optimization?
Now we do not discuss the mirror -based image based on Ubuntu. Because Friendelec (Guangzhou Friendship Electronic Technology Co., Ltd.) informed me that they had added some optimization, so I upgraded the updated version of Friendlywrt and tested this image: TML3. TML6 we did some of the new mirror image Optimized, such as network card interrupt settings, uninstalled support, etc. " So I downloaded the" RK3568888 found on Google Drive on Google Drive -eflasher-Friendlywrt-20220526.img.gz ", and then record it with USBImager to , Then Boot to the router. After
, it will automatically record the image into the EMMC flash. If you connect to the display, you can operate according to the results. After the operation is completed, remove the Microsd card and restart the router.
can be checked by connecting the HDMI display (as shown above) or viewing the LED on the device. This process is very fast, just a few seconds installed on the EMMC flash memory.
This new version of the image is mainly on 40-net-smpinity files.In the previously pre -installed Friendly, it looks like this:
0
1H7H7
2
3FriendlyElec, Nanopi-R5s)
set_interface_core 8 "ETH0"
ECHO 7/SYS/Class/NET/ETH0/Queues/RPS_CPS
SET_IN Terface_core 2 "ETH1-0"Set_interface_core 4 "ETH1-16"
Set_interface_core 4 "ETH1-18"
EML6E Cho B/SYS/Class/NET/ETH1/Queues/RX-0/RPS_CPushml7
SET_INTERFACE_CORE 4 "ETH2-0"
Set_interf ACE_CORE 2 "ETH2-16"
SET_INTERFACE_CORE 2 "ETH2- 18"
echo 9 /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
而新的
0
1
2
3friendlyelec,nanopi-r5s)
set_interface_core 8 "eth0"
echo f /SYS/Class/Net/Eth0/Queues/RX-0/RPS_CPUS
SET_INTERFACE_COR Core 4 "ETH1-16" Hilml7
SET_INTERFACE_COR -0/rps_cpus
set_interface_core 2 "eth2-0"
set_interface_core 2 "eth2-16"
set_interface_core 2 "eth2-18"
echo d /sys/class/net/eth2/queues/rx-0/rps_cpus
;;
esac
Willy Tarreau也解释了Reasons for changing the ETH1 interface:
It involves rps. They received this IRQ (interrupt request) on Core 2 and re -assigned the flow to Core 0, Core 1, and Core 3. This is the correct method of using RPS. However, to reach this point, you must manually distribute a network performance testing tool iPerf and observe the first saturated Core. If you first use KSOFTIRQD to saturate Core 2, then make sure that iPerf can run on any of the other three cores. If Core2 is slightly idle, try to set an iPerf on it. If you put the iPerf on it will pop up KSoftirqd, then they will hinder each other, and users will be more willing to change the RPS settings to help release another Core and use it for IPERF.
Before testing and switching to Ubuntu, I actually didn't try this method.When I tried to use the new Friendlywrt mirror, the result was worse: $ iPerf3 -T 60 -C 192.168.2.1 -i 10
Connecting to host 192.168.2.1, Port 5201
[5] LOCAL 192.168.2.130 Port 49590 Connect to 192.168. 2.1 Port 5201
[ID] interval transfer bitrate Retr CWND
[5] 0.00-10.00 SEC 1.92 GBYTES 1.65 GBITS/ SEC 0 1.62 MByteml7
[5] 10.00-20.00 SEC 1.91 GBYTES 1.64 GBITS/SEC 10 2.34 MBYTES
[5] 20.00-30.00 s EC 1.90 GBytes 1.63 Gbits/sec 0 2.61 MBytes
[ 5] 30.00-40.00 sec 1.85 GBytes 1.59 Gbits/sec 4 1.30 MBytes
[ 5] 40.00-50.00 sec 1.88 GBytes 1.61 Gbits/sec 1 1.06 MBytes
[ 5] 50.00-60.00 sec 1.76 GBytes 1.51 GBITS/SEC 2 868 KBYTES
- ------------------
[ID] intermfer bitrate TR[5] 0.00-60.00 SEC 11.2 Gbytes 1.61 Gbits/Secret 17 SENDER
[5] 0.00-60.05 SEC 11.2 GBYTES 1.60 GBITS/Secire
IPERF ml7$ iPerf3 -T 60 -C 192.168.2.1 -i 10 -RTML3
Connecting to host 192.168.2.1, Port 5201
reverse mode, remote host 192.168 .2.1 is sending
[5] LOCAL 192.168.2.130 Port 49594 Connected to 192.168.2.1 Port 5201
[ID] Internet Val Transfer Bitraate[5] 0.00-10.00 SEC 1.22 GBYTES 1.05 GBITS/SEC
[5] 10.00-20.00 SEC 1.36 s 1.17 GBITS /sec
[5] 20.00-30.00 SEC 1.31 GBYTES 1.12 Gbits/Sec
[5] 30.00-40.00 SEC 1.26 gbites 1.26 gbites 1.26 gbites S/SEC[5] 40.00-50.00 SEC 1.47 GBYTES 1.26 GBITS/Sec
[5] 50.00-60.00 SEC 1.46 Gbytes 1.26 Gbits/Sec- - - - ---------------------
[id] interval 7[5] 0.00-60.05 SEC 8.29 GBITES 1.19 GBITS/SEC 1 SENDER
[5] 0.00-60.00 SEC 8.29 GBYTES 1.19 GBITS/Sec itiver
IPERF Done.
So, this problem has to be re -examined.NANOPI R5s M.2 NVME SSD Install
I bought Apacer AS2280 (AP256GAS2280P4-1) PCIE Gen 3.0 X4 SSD not long ago. It can achieve the correct hardware on the correct hardware. Reading speed and as high as 1,100 as high as 1,800 mb/s The sequence of MB/s is written.
APACER M.2 2280 PCIE SSD
L6 installation process is simple because I only need to loosen four four four four. The screws can be removed, installed SSD, and then fixed it with the screws they provided.
NANOPI R5s Installation
Riendlycore
I first tried to install Friendlycore with EFLASHER mirror.
use EFLASHER mirror to install FriendlyCore
I started the router, but later I noticed that the WAN interface link was not displayed on the TP-Link switch. Only the power supply LED lit up, and the power supply LED light up Mirror is normal. I tried again and settled through the EFLASHER UI settings by clicking "to complete", but still not.
eflasher Install F Riendlycore
Start directly from the Microsd card and run the operating system from there. After doing so, it is normal to run. However, if you intend to use the NANOPI R5s for a variety of uses and look forward to using the desktop environment in Ubuntu 20.04 mirrors, it may be disappointed because HDMI output can only be used to access the terminal.
UBuntu 20.04 FRIENDLELEC login TML3
Friendlycore system information
You can find the startup log on CNX Software Pastebin.我使用pi/pi凭据（用户名/密码）通过过了SSH登录，并将系统升级到了最新软件包：
sudo apt update
sudo apt dist-upgrade
现在，我们运行一些命令来获取系统信息：@FriendlyELEC:~$ cat /etc/lsb-releaseDistRib_id=ubuntu
Distrib_release=20.04
h6h6distrib_codename = Focal
DistRib_Descripting = "Ubuntu 20.04.4 LTS"
PI@Friendlye LEC: ~ $ Uname -ALinux Friendlyelec 5.10.66 #219 SMP PRIMPT FRI APR 22 18:20:21 cst 2022 AARCH64 aarch64 aarch64 GNU/Linux
pi@FriendlyELEC:~$ free -mh
total used free shared buff/cache available
Mem: 1.9Gi 150Mi 1.7Gi 3.0Mi 114Mi 1.7Gi
Swap: 0B 0B 0B
pi@FriendlyELEC:~$ df -mh
Filesystem Size Used Avail Use% Mounted on
udev 969M 0 969M 0% /dev
tmpfs 197M 480K 196M 1% /run
overlay 27G 1013M 26G 4% /
tmpfs 981M 0 981M 0% /dev/shm
tmpfs 5.0M 4.0K 5.0M 1% /run/lock
tmpfs 981M 0 981M 0% /sys/ FS/CGROUP
TMPFS 197m 0 197m 0%/RUN/User/1000
Except for the NVMe driver without automatic mounting, everything looks pretty good.Now we use inxi to find more details:
0
1
2
3
4
5
6
7
8
9
24
pi@FriendlyELEC:~$ inxi -Fc0
System:
Host: FriendlyELEC Kernel: 5.10.66 aarch64 bits: 64 Console: tty 0
Distro: Ubuntu 20.04.4 LTS (Focal Fossa)
Machine:
Type: ARM Device System: FriendlyElec NanoPi R5S details: N/A
serial: 8cbfe79e107c459c
Battery:
ID-1: test_battery charge: 100% condition: N/A
CPU:
Topology: Quad Core model: N/A variant: cortex-a55 bits: 64 type: MCP
Speed: 408 MHz min/max: 408/1992 MHz Core speeds (MHz): 1: 1992 2: 1992
Graphics:
Device-1: display-subsystem driver: rockchip_drm v: N/A
Device-2: mali-bifrost driver: mali v: N/A
Device-3: rk3568-dw-hdmi driver: dwhdmi_rockchip v: N/A
Display: server: -G --display
Audio:
Device-1: rk3568-dw-hdmi driver: dwhdmi_rockchip
Device-2: simple-audio-card driver: asoc_simple_card
Device-3: simple-audio-card driver: N/A
Device-4: simple-audio-card driver: asoc_simple_card
Sound Server: ALSA v: k5.10.66
Network:
Device-1: Realtek RTL8125 2.5GbE driver: r8125
IF: eth1 state: down mac: e2:1d:62:a1:1a:ca
Device-2: Realtek RTL8125 2.5GbE driver: r8125
IF: eth1 state: down mac: e2:1d:62:a1:1a:ca
Device-3: rk3568-gmac driver: rk_gmac_dwmac
IF-ID-1: eth0 state: up speed: 1000 Mbps duplex: full
mac: de:1d:62:a1:1a:ca
IF-ID-2: eth2 state: down mac: 12:bf:2b:d6:4b:e0
Drives:
Local Storage: total: 274.88 GiB used: 1012.3 MiB (0.4%)
ID-1: /dev/mmcblk0 model: SD16G size: 29.12 GiB
ID-2: /dev/mmcblk2 model: 8GTF4R size: 7.28 GiB
ID-3: /dev/nvme0n1 vendor: Apacer model: AS2280P4 256GB size: 238.47 GiB
Partition:
ID-1: / size: 26.48 GiB used: 1012.3 MiB (3.7%) fs: overlay
source: ERR-102
Sensors:
System Temperatures: cpu: 46.1 C mobo: N/A
Fan Speeds (RPM): N/A
Info:
Processes: 130 Uptime: 7m Memory: 1.92 GiB used: 211.4 MiB (10.8%)
Init: systemd Shell: bash inxi: 3.0.38
Only the eth0 WAN port is open, the eth1/eth2 2.5GbE port is closed, and no configuration is displayed at all.FriendlyElec seems to focus on Friendlywrt mirror. They told me that it has not been optimized on Friendlycore, so most people may use Friendlywrt because it is easier to configure network and router settings. I saw that Apacer AS2280P4 SSD was actually detected, but it was formatted without the box, so I could only use mkfs.ext4 to format it.NanoPi R5S Benchmark
Now we do this by running SBC on the router Bench method to benchmark the CPU, hoping to find some problems as much as possible:
10
1
2
3
4
5
6
7
8
9
$ sudo /bin/bash ./sbc-bench.sh -c
[sudo] password for pi:
WARNING: dmesg output does not contain early boot messages which
help in identifying hardware details.
It is recommended to reboot now and then execute the benchmarks.
Press [ctrl]-[c] to stop or [enter] to continue.
Average load and/or CPU utilization too high (too much background activity). Waiting...
Too busy for benchmarking: 07:21:06 up 3 min, 1 user, load average: 0.41, 0.27, 0.11, cpu: 3%
Too busy for benchmarking: 07:21:11 up 3 min, 1 user, load average: 0.38, 0.26, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:16 up 3 min, 1 user, load average: 0.35 , 0.26, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:21 up 3 min, 1 user, load average: 0.32, 0.25, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:26 up 3 min, 1 user, load average: 0.29, 0.25, 0.11, cpu: 1%
Too busy for benchmarking: 07:21:31 up 3 min, 1 user, load average: 0.27, 0.24, 0.10, cpu: 1%
sbc-bench v0.9.7
Installing needed tools. This may take some time. Done.
Checking cpufreq OPP. Done (results will be available in 20-28 minutes).
Executing tinymembench. Done.
Executing 7-zip benchmark. Done.
Executing cpuminer. 5 more minutes to wait. Done.
Checking cpufreq OPP. MB/s
memset: 6191.5 MB/s (0.2%)
Cpuminer total scores (5 minutes execution): 6.87,6.86,6.85,6.84,6.83,6.82,6.79 kH/s
OpenSSL results:
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-128-cbc 173609.37k 509936.75k 972013.31k 1264387.07k 1383497.73k 1392645.46k
aes-128-cbc 175451.26k 506569.66k 973690.71k 1264628.74k 1382845.10k 1393180.67k
aes-192-cbc 166539.51k 448796.48k 790104.58k 970846.55k 1040621.57k 1046462.46k
aes-192-cbc 168407.31k 451709.25k 792148.91k 970579.63k 1041061.21k 1046375.08k
aes-256-cbc 159430.38k 412822.74k 676804.10k 809129.64k 857347.41k 861137.58k
aes-256-cbc 162313.43k 412763.39k 677746.94k 809317.38k 857642.33k 861334.19k
Unable to upload full test results. Please copy&paste the below stuff to pastebin.com and
provide the URL. Check the output for throttling and swapping please.
I started it almost immediately after booting, so the dmesg output should be complete (see the boot loading earlier in this review for details), but some information is missing from the script.The full output of the sbc-bench.sh script can be found at at pastebin We see that the "1992" MHz advertised frequency is tested in reality as 1845 MHz, so I feel like some optimization could be done here.
NVMe benchmark
I use iozone 3 for NVMe The SSD was tested 3 times, 1 of which was a 100MB file:
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 100M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
Iozone: Performance Test of File I/O
Version $Revision: 3.489 $
Compiled for 64 bit mode.
Build: linux
Include fsync in write timing
O_DIRECT feature enabled
Auto Mode
File size set to 102400 iozone -e -I -a -s 100M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
Output is in kBytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 kBytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
02400 4 34994 53668 30431 30385 30136 59719
02400 16 102031 130543 80174 80125 79796 133162
02400 512 300692 296328 276975 291837 276681 313464
02400 1024 309822 340026 308900 326826 306102 339059
02400 16384 357975 392544 369753 391219 370336 390004iozone test complete.
and then a 500MB file:
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 500M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
and finally a 1GB file:
pi@FriendlyELEC:/media/nvme0n1$ sudo iozone -e -I -a -s 1000M -r 4k -r 16k -r 512k -r 1024k -r 16384k -i 0 -i 1 -i 2
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
024000 4 35105 58082 30395 30447 27458 60895
024000 16 102421 135279 80210 80314 74596 133579
024000 512 300759 314704 282743 283883 277911 313413
024000 1024 329840 337468 318228 319091 317641 337714
024000 16384 383289 385247 382642 382850 382870 381344
After a series of operations, the results are more or less consistent in all three tests without much change.In the end, I got a read-write speed of about 380MB/s, which was far lower than the original promotional writing and reading speed of SSD, and the results of ODROID-M1. I think this is because the PCIe 2.0 X1 interface is used in this design, not the PCIE Gen 3.0 X2 interface used in Hardkernel board.Below is the output of LSPCI, for reference only:
.5GBE interface configuration and benchmark test
TML3
0
1
2
3
4L6 15
PI@FriendlyElec:/Media/NVME0N1 $ Sudo LSPCI -VTML3
Subsystem: Phison Electronics Corporation Device 5013
FLAGS: Bus Master, Fast Devsel, Laten Cy 0, IRQ 87
MEML6Memory at 380900000 (64-bit, Non-Prefetchable) [SIZE = 16K]
capabilities: [80] e0] e0] XPress Endpoint, MSI 00CAPabilities: [ D0] MSI-X: ENABLE+ Count = 9 Masked-
Capabilities: [E0] ENable-Count = 1/8 Maskable+ capabilities: [F8] Power Management Version 3TML3
CAPABILIES: [100] Lanetcy Tolerance l6capabilities: [110] l1 PM SubstateS
CAPABILIES: [200] Advanced Error Reportingml7
CAPABILITIES: [300] Secondary PCI EXPRER ss
kernel driver in use: nvme
ML6NANOPI R5s router
is configured with ETH0 Gigabit Ethernet "WAN" interface. Therefore, we must manually configure two 2.5GBE ports. I used the same test platform with the same "first part of Friendlywrt Evaluation", namely Ubuntu 20.04 laptop. Then, I connected REALTEK RTL8156BG USB 3.0 to 2.5GBE crypto dog to ETH1, UP XTREME I11 mini PC is connected to ETH2. I did not use the bridge interface like in Friendlywrt, but configured two different subnets: ETH1 was 192.168.2.0, ETH2 was 192.168.3.0.
Now we create two new files in /etc/network/interfaces.d/:
- eth1
auto eth1
iface eth1 inet static
address 192.168.2.1
network 192.168.2.0
netmask 255.255.255.0
broadcast 192.168.2.255
- eth2
auto eth2
iface eth2 inet static
address 192.168.3.1
network 192.168.3.0
netmask 255.255.255.0
broadcast 192.168.3.255
Now install the DHCP server
sudo apt install isc-dhcp-server
Edit the /etc/dhcp/dhcpd.conf file using our two subnets:
subnet 192.168.2.0 netmask 255.255.255.0 {
range 192.168.2.100 192.168.2.200;
option routers 192.168.2.1;
}
subnet 192.168.3.0 netmask 255.255.255.0 {
range 192.168.3.100 192.168.3.200;
option routers 192.168.3.1;
}
Before restarting the dhcp server:
sudo systemctl restart isc-dhcp-server
At this point, the laptop and mini PC should be able to obtain their IP addresses from the NanoPi R5S on their respective subnets. Now we can start benchmarking the interface.使用连接到笔记本电脑的eth1下载iperf3，然后从R5S的角度接收：
0
1
2
3
4
5
6$ iperf3 -t 60 -c 192.168.2.1 -i 10
Connecting to host 192.168.2.1, port 5201
[ 5] local 192.168.2.130 port 59822 connected to 192.168.2.1 Port 5201
[ID] Interval Transfer RetR CWND
[5] 0.00-10.00 SEC 2.28 GBYTES 1.9 1.9 6 GBITS/SEC 42 1.41 MBYTES
[5] 10.00-20.00 SEC 2.02 GBITS/SEC 0 1.61 MBYTES
[5] 20 .00- 30.00 sec 1.72 GBytes 1.48 Gbits/sec 0 1.62 MBytes
[ 5] 30.00-40.00 sec 1.87 GBytes 1.61 Gbits/sec 0 1.62 MBytes
[ 5] 40.00-50.00 sec 1.89 GBytes 1.62 Gbits/sec 0 1.70 MBytes
[ 5] 50.00-60.00 sec 2.06 GBYTES 1.77 GBITS/SEC 21 1.66 MBYTES
- - - - - - - - - - - -----
sofer bitrate retr
[5] 0.00-60.00 SEC 11.8 GBYTES 1.70 GBITS/SEC 63 SENDER
[5] 0.00-60.04 SEC 11.8 GBYTES 1.69 GBITS/SEC iPerf Done.
This is slower than the 1.85 gbps I got in OpenWrt, and there are also transmission content.During the transfer I also use l sbc-bench.sh monitoring system:
0
1
213
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
The system did run at its advertised maximum frequency during testing, and I didn't see any obvious issues here.I also checked some information and statistics using ethtool:
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ethtool -i eth1
driver: r8125
version: 9.008.00-NAPI
firmware-version:
expansion-rom-version:
bus-info: 0000:01:00.0
supports-statistics: yes
supports-test: no
supports-eeprom-access: no
supports-register-dump: yes
supports-priv-flags: no
pi@FriendlyELEC:~$ sudo ethtool -S eth1
NIC statistics:
tx_packets: 451228
rx_packets: 9569147
tx_errors: 0
rx_errors: 0
rx_missed: 0
align_errors: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 9569102
broadcast: 45
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 31676089
rx_octets: 14506385933
rx_multicast64: 0
tx_unicast64: 451214
tx_broadcast64: 2
tx_multicast64: 12
tx_pause_on: 570
tx_pause_off: 570
tx_pause_all: 1140
tx_deferred: 0
tx_late_collision: 0
tx_all_collision: 0
tx_aborted32: 0
align_errors32: 0
rx_frame_too_long: 0
rx_runt: 0
rx_pause_on: 0
rx_pause_off: 0
rx_pause_all: 0
rx_unknown_opcode: 0
rx_mac_error: 0
tx_underrun32: 0
rx_mac_missed: 31
rx_tcam_dropped: 0
tdu: 0
rdu: 570
I do get some rx_mac_missed.Now let's test it in reverse:
0
1
2
3
4
5
6
7$ iperf3 -t 60 -c 192.168.2.1 -i 10 -R
Connecting to host 192.168.2.1, port 5201
Reverse mode, remote host 192.168.2.1 is sending
[ 5] local 192.168.2.130 port 59826 connected to 192.168.2.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.75 GBytes 1.50 Gbits/sec
[ 5] 10.00-20.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 20.00-30.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 30.00-40.00 sec 1.95 GBytes 1.67 Gbits/sec
[ 5] 40.00-50.00 sec 1.94 GBytes 1.67 Gbits/sec
[ 5] 50.00-60.00 sec 1.94 GBytes 1.67 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.04 sec 11.5 GBytes 1.64 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.5 GBytes 1.64 Gbits/sec receiver
iperf Done.
This looks much better than OpenWrt (1.12Gbps).
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
0
1
2
3
4TML3
6Devkit@UPX -I11: ~ $ iPerf3 -T 60 -C 192.168.3.1 -i10
Connecting T o host 192.168.3.1, port 5201[5] LOCAL 192.168.3.100 Port 37794 Connected to 192.16888 .3.1 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 10.00-20.00 sec 2.73 GBytes 2.35 Gbits/sec 0 1.81 MBytes
[ 5] 20.00-30.00 SEC 2.73 GBYTES 2.35 Gbits/SEC 0 1.81 MBytes
[5] 30.00-40.00 SEC 2.73 GBITS/SEC 0 2.90 l3
[5] 40.00-50.00 SEC 2.73 GBYTES 2.35 GBITS/SEC 0 4.37 MBYTES
[5] 50.00-60.00 SEC 2.73 GBYTES 2.35 GBITS/SEC 0 4.37 MBYTES- - - - - - - - - - - --------
Tra on Retr
[5] 0.00-60.00 SEC 16.4 Gbytes 2.35 Gbits/SEC 0 SENDER
[5] 0.00-60.00 SEC 16.4 GBYTES 2.35 GBITS/Secret
IPERF .
Oh, great! This is the first time I have received a 2.35 Gbps transmission speed, so it still looks hopeful! @Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % CPU % SYS % USR % Nice % IRQ TEMP
Unless I get wrong, 25%of IRQ means that a core can be used to handle these.Now let's try:
0
1
2
3
4
5
6
7devkit@UPX-i11:~$ iperf3 -t 60 -c 192.168.3.1 -i 10 -R
Connecting to host 192.168.3.1, port 5201
Reverse mode, remote host 192.168.3.1 is sending
[ 5] local 192.168.3.100 port 37800 connected to 192.168.3.1 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.92 GBytes 1.65 Gbits/sec
[ 5] 10.00-20.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 20.00-30.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 30.00-40.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 40.00-50.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 50.00-60.00 sec 1.84 GBytes 1.58 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.01 sec 11.1 GBytes 1.59 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.1 GBytes 1.59 Gbits/sec receiver
iperf Done.
The result is 1.59 Gbps, not quite perfect, but still better than OpenWrt.@Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % cpu % sys % usr % nice % IRQ TEMP
^C
CPU is running again at full speed, and the distance is 100%different, so I think the problem should be Out of it elsewhere. Now we can use EthTool to check ETH2 information and statistical information again.
TML3H0H0
TML3
0TML3
L6 $ SUDO ETHTOL -i ETH2
DRIVER: R8125
Version: 9.008.00 -napi
FI6Fi rmware-version:expans-Rom-version:
bus-info: 0001: 11: 00.0
Supports-Statistics: YES
SUPPPOSTS-TEST: noTML3
SUPPPPROM-ACCESS: no
supports-register -dump: yes
supports-priv-flags: no
0
1
2
3
4
5
6
7
8
9
sudo ethtool -S eth2
NIC statistics:
tx_packets: 8506609
rx_packets: 12553353
tx_errors: 0
rx_errors: 0
rx_missed: 0
align_errors: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 12553209
broadcast: 144
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 12543719502
rx_octets: 18471602900
rx_multicast64: 0
tx_unicast64: 8503557
TX_BROADCAST64: 3035
TX_Multicast64: 17
tx_PAUSE _ON: 35
TX_PAUSE_OFF: 35
TX_PAUSE_ALL: 70_Deferred: 0Htmmmm l7tx_late_collision: 0
tx_all_collision: 0
tml6_aborted32: 0ALIGN_ERRORS32: 0
RX_FRAME_TOO_LONG: 0
RX_RUNT: 0
HTMMML3HTMM l0rx_pause_on: 0RX_PAUSE_OFF: 0RX_PAUSE_ALL: 0TML3rx_unknownown_opcode: 0
rx_mac_error: 0
tx_underrun32: 0r6r6r6r6r6r6r6r6r6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6, x_mac_missed: 335
rx_tcam_dropped: 0
TDU: 0
RDU: 35RDU: 35RDML6RDML6RDML6RDML6RDML6RDML6RML0R6RDU: 35htmlml 7
test results here Among them, I found more rx_mac_missed. So I guess there should be some adjustments to improve performance.But according to my previous experience of adjusting the settings of RTL8156B, adjusting the settings is really tricky, and people with experience can't seem to agree on what setting options to adjust. I mainly refer to the RTL8156/8125 driver. Realtek engineers, as well as some networking experts among our readers.Configuring NAT between two 2.5GbE interfaces
Since the 2.5GbE interface doesn't work optimally with iperf3, I didn't bother testing the router performance in FriendlyWrt, but a few people asked me anyway. So next I will show how I configured NAT in Ubuntu 20.04 and will continue to test NAT performance, keeping in mind that it will definitely improve a lot in a few weeks or months.
Here we need to enable IP forwarding and NAT. The instructions I used were adapted from a post on networkreverse.
Edit /etc/sysctl.conf to enable IP forwarding (uncomment the following lines):
net.ipv4.ip_forward=1
Apply changes:
sudo sysctl -p
Now we enable NAT:
sudo iptables ! -o lo -t nat -A POSTROUTING -j MASQUERADE
We can now ping Xtreme on the 192.168.3.0 subnet from the laptop on the 192.168.2.0 subnet i11:
jaufranc@cnx-laptop-4:~$ ping 192.168.3.100
PING 192.168.3.100 (192.168.3.100) 56(84) bytes of data.
If you want the changes to be permanent, you can execute the following:
sudo apt install iptables-persistent
sudo sh -c 'iptables-save /etc/iptables/rules.v4'
I tried iperf3 between a UP Xtreme i11 and my laptop, the data was routed through a NanoPi R5S router.@cnx -laptop -4: ~ $ iPerf3 -T 60 -C 192.168.3.100 -i 10Connecting to host 192.168.3.100, PORT 5201
[5] LOCAL 192.168.2.130 Port 59430 Connect to 192.168.3.100 Port 5201
Bitrate Retr CWND[5] 0.00-10.00 SEC 914 MBYTES 767 MBITS/SEC 355 1011 Kbytes
[5] 10.00-20.00 SE C 912 MBYTES 765 Mbits/SEC 324 1.23 MBYTES[5] 20.00-30.00 SEC 917 MBITS/SEC 124 1.09 MBYTES
[5] 30.00-40.00 SEC 915 MBYTES 767 MBITS/SEC 150 942 Kbytes
[5] 40.00-50.00 SEC 915 MB 915 MB 915 YTES 767 MBITS L6- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.00 sec 5.36 GBytes 768 Mbits/sec 1095 sender
[ 5] 0.00-60.06 SEC 5.36 GBYTES 767 MBITS/SEC ReceiverTML3
IPERF : ~ $ iPerf3 -T 60 -C 192.168.3.100 -i 10 -R
Connecting to Host 192.168.3.100, Port 5201
Reverse Mode, remote host 192.168 .3.100 is sendingml7
[5] LOCAL 192.168.2.130 Port 59434 Connect to 192.168.3.100 Port 5201
[id] in Terval Transfer Bitrate
[5] 0.00-10.00 SEC 1.09 GBYTES 935 MBITS/SEC
[5] 10.00-20.00 SEC 1.09 GBYT ES 938 mbits /sec
[5] 20.00-30.00 SEC 1.09 GBYTES 938 mbits/sec
[5] 30.00-40.00 SEC 1.09 gbytes 938 mbits/ sec[5] 40.00-50.00 SEC 1.09 GBYTES 939 Mbits/Sec
[5] 50.09 Gbytes 1.09 Gbytes 937 MBITS/SEC
- - - -----------------
[ID] internsfer bitrate
[5] 0.00-60.05 SEC 6.55 GBYTES 937 MBITS/SEC 973 SENDER
[5] 0.00-60.00 SEC 6.55 GBYTES 937 MBITS/Security
IPERF 5
The transmission speed is 768 Mbps, and the other direction is 937 Mbps.
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
According to the mpstat command display, this should be processed by core #0.
$ mpstat -P ALL -I SUM
Linux 5.10.66 (FriendlyELEC) 06/05/22 _aarch64_ (4 CPU)
The above point can be confirmed by using top and htop.
NanoPi R5S power consumption
I happen to have a wall power meter and can use it to check power consumption:
- Idle - 5.1 W
- Iperf3 to eth1 - 6.3 to 6.7W
- NAT test between laptop and mini – 6.2W
The above data was tested after installing Ubuntu 20.04 and NVMe SSD on this product.
I also tested on a NanoPi R5S using OpenWrt and no SSD:
- Idle - 4.6W
- Iperf3 - 6.0 to 6.2 W
Note that since I am using a wall-mounted power meter, this number will include efficiency losses in the power adapter (Khadas VIM4 USB-C Power Adapter). And the value may be higher than using a USB-C power meter. This should also be possible through settings to optimize power consumption.
final conclusion
That’s it for today’s evaluation. The optimization part should actually include "changing the firmware so that the Rockchip core runs at 1992 MHz", "adjusting various settings related to PCIe and Ethernet settings" and so on. Most of them I'm not familiar with yet.
(3), Taipower Linglong M NUC Flash X86 Soft Router Review
Foreword
Many friends have bought X86 soft routers and most of them have been idle and dusty, or they are preparing to step into the pit of X86 soft routers! First of all, you need to make it clear whether you are willing to spend time to fiddle with it. If you are not willing to spend time to study, I advise you to give up and consider buying X86 soft routing. You can consider buying a finished opnenwrt router to experience it, or you can buy a router with Merlin firmware or a router with PandoraBox firmware to play with.
Prepare relevant network equipment
A suitable 2.5G port switch and an X86 soft router with a 2.5G electrical port, and an external wired network card on the Type-C side of the 2.5G network port or the USB-A side.
Teclast Linglong M NUC is a mini machine that is very suitable for use as an X86 soft router. It has two 2.5GE electrical ports, two SO-DIMM memory slots and a PCIe 3.0 NVMe SSD M.2 slot. It is recommended to buy bare metal and bring your own memory modules and NVMe SSD. Two 4GB Micron DDR4 3200MHz memory sticks and one Intel Optane M10 16GB (Optane memory) are prepared here.
After installing the memory and NVMe SSD, take the prepared Router OS U disk to install the soft routing system.
can be used after powering on.
Xike is a new network equipment brand. It is a layer 2 management switch that comes with 8 2.5G electrical ports and 4 10G optical ports. It supports port aggregation and VLAN division.
A friend bought this XIKE switch and asked me to help upgrade the network at home, planning to create a 2.5G intranet environment for the whole house.
Starting to debug the equipment is very simple for experienced people.
can see the 10G light port of Xike SKS7300-8GPY4XGS switch. The connection status 10G Full rate. For friends with 10G SFP modules in the hand, it can toss the Magazine LAN. However, at this stage, 2.5G LAN is more suitable for the actual needs of the public. If you upgrade to the 10G (10G) light port, the overall hardware cost will be several times higher. For small and medium -sized enterprises, this switch is very suitable for connecting 10,000 -dimensional core switches for access switches. It is still more recommended for home users who need 2.5G electric ports and friends who like to toss 2.5G LAN to start this Xike switch.
has a cat stick (GPON Stick onu module) in the opponent, and it can save a light cat on this SKS7300-8GPy4xgs switch. Effectively solve the disadvantages of the equipment accumulation in the weak electric box, and can also improve the problem of the heat accumulation between the space inside the weak electric box to reduce the heat accumulation between the equipment.
Computer (desktop and laptop) What should I do if there is no 2.5G network port? You can buy a superior USB to RJ45 2.5G wired network card. Enter the computer through the Type-C end, and the RJ45 network port on the other end is 2.5GBE (2500Mbps) rate.
takes a U disk to make guidance disk
. I generally like to use Balenaetcher to make a U disk. The simple fool -type operation is very easy! Support ISO, DMG, IMG, IMG.GZ and other mirror file format formats to make guidance disks.
Make a sea spider V9 guide disk (flashing machine)
to the sea spider web web to find the download area in the service and support. Pull to the bottom upgrade address, find V9.0, select the next download next to you can download the sea spider V9 firmware. Note that the download is the ISO mirror file.
Sea Spider V9 System
default login address: 192.168.0.1. Accounts and passwords are set by themselves after flashing.
can view the current information profile on the overall view. The instrument icon shows the CPU usage rate, memory usage, and disk usage. Below is the monitoring of total internal network in real network, and there are some key information on the right side.
is already using the LAN port. It can identify the network port model: Realtek RTL8125 2.5GBE, and the mesh icon also identifies 2.5GB.
In the overall view, the dog checked the current memory usage. After the flashing machine, you can see 10.5%of the default use, and 835.86 MB storage space has been used. Total memory: 7.6GB (8GB of total memory), and the remaining 6.8GB is available.
can also check the CPU load in the overall view.
interface can also view the data of the LAN port. I am running in the full dual -work mode, 2500MB/s (that is, 2.5GBE Gigabit mouth rate).
can see the motherboard/CP information in the hardware information, which basically shows the relevant information.
Here is 16GB of Ao Teng Memory, available capacity: 13.41GB, you can check the partition of the hard disk storage.
Two 2.5GBE Gigabit Gear Networks can be identified normally.
Sea Spider has module management, including the Docker container module.
基本上这些模块对于软路由来说都算是很齐全了，VPN IPsec、PHP套件、IGMP组播代理等都一应俱全。
FTP service.
file storage sharing.
Comment: Sea Spider V9 needs to buy related authorization to use all the functions, fortunately, there is a longer trial time. Relatively many functions are more biased towards the use of web workers. For ordinary white users, it may be difficult to operate, and some professional terms are also very difficult to understand.
Make astore OS guide disk (flashing)
to Koolcenter official website to download isstore OS firmware. Find the IStore OS directory in the firmware list.
Considering that the current new X86 soft routing is booted by UEFI, so select X86_64_EFI to download the iStore OS firmware.
Generally has Date (Date) in the bottom list. Select the latest Istore OS firmware.
istore OS
In fact, ISTORE OS is the in -depth customization of Chinese people. Compared with the original OpenWrt, the gameplay (plug -in) and Docker gameplay.
homepage will have a legend with traffic statistics, IP addresses, and network interface status. It looks clear at a glance.
pulls down to see disk information, storage services, docker, download services and remote domain names. The playability is still very high.
can be seen at the bottom to see the system information. The dog recognizes the CPU temperature, CPU usage rate, and memory usage.
network wizard, this is the best direction function for Xiaobai users.
has system information. You can see the computer model, processor architecture and other related information.
can also see the use of memory at the overview.
Soft Route is difficult to master a disk management. Here ISTORE OS directly completes a graphical Diskman disk management, which reduces the difficulty of novice operation and enhances the simpleness of use.
istore OS built -in Docker
is definite It is difficult to install plug -ins. Many beginners or little white users who have just started. The iStore OS directly built -in Docker components for easy deployment of containerization. And supporting CPU and memory configuration settings in advanced settings, which is very user -friendly.
ISTORE OS Unique App Store
For wanting to study the smart home experience, you can consider Home Assistant.
is easy to be suitable for miniature home deployment data service centers. Many tutorials on the Internet will not be discussed in depth.
If you can't afford a dedicated NAS device, in fact, soft routes can also act as simple NAS, but most of the X86 soft routes are not easy to use. There are two tools that can be solved directly here to solve this problem. For friends who want to build an audio -visual center: Jellyfin and NAS Tools. If you can use it, it can definitely easily make the X86 soft routing into a NAS.
ARIA2 downloader. If you can use it, it is also a very practical download tool. For the two broadband at home, in order to avoid the broadband during the work or during the business trip, try to deploy the net Xinyun to earn some benefits.
I actually prefer HOMEBOX inner mesh speed measuring. This is better experienced by the Speedtest speed measuring server. I will show the experience of the HOMEBOX intranet speed measuring experience later.
DISKMAN disk management is definitely a plug -in worth installed. Friends who use QBITTORRENT downloaders will use it to use it. The specific use methods and tutorials have more detailed professional introduction on the tutorial network. If you are interested in the research, go to the study.
KMS server can solve the problem that Windows and Office prompts not activated.
installation plug -in is as simple and convenient as installing the app, which is also the biggest ease of use of Isstore OS.
I use most of these four plug -ins. Among them, multi -line multi -broadcast is suitable for broadband overlay areas. One sentence can run to 600m in one sentence, 500M superposition can run 1000m but do not need to upgrade broadband packages and extra money. It is worthwhile.
KMS server is enabled. Basically, local Windows and Office have no longer encountered unauthorized situations.
This is the HomeBox intranet speed test, which can directly measure the speed of the intranet. I use Teclast Linglong M NUC’s 2.5GbE Gigabit network port, CAT6 (Category 6e) network cable, and Type-C to RJ45 2.5GbE external wired network card, which can achieve a 2500Mbps full-link LAN (intranet) experience .
Summary
Users with basic network engineering skills or a relatively good understanding of the Internet can consider starting X86 soft routing. Sea Spider V9 actually prefers traditional network equipment, combining functions such as switches, routers, and AC controllers. iStore OS is suitable for the general public. It is more playable and easier to get started. I personally recommend flashing the iStore OS firmware to use it, which is basically the Router OS in openwrt that I have used and experienced well.
Now we try:@upx -I11: ~ $ iPerf3 -T 60 -C 192.168.3.1 -i 10 -RConnecting to host 192.168.3.1, Port 5201reverse Mode, Remote Host 192.168.3.1 is Sending
[5] Local 192.168.3.100 Port 37800 Connect to 192.168.3.1 Port 5201
[ID] Interval Transfer
[5] 0.00-10.00 SEC 1.92 GBYTES 1.65 GBITS/Sec
[5] 10.00-20.00 SEC 1.84 GBYTES 1.58 GBITS/Sec
[5] 20.00-30.00 SEC 1.84 Gbytes 1.58 Gbits/sec[ 5] 30.00-40.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 40.00-50.00 sec 1.84 GBytes 1.58 Gbits/sec
[ 5] 50.00-60.00 sec 1.84 GBytes 1.58 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.01 sec 11.1 GBytes 1.59 Gbits/sec 0 sender
[ 5] 0.00-60.00 sec 11.1 GBytes 1.59 GBITS/SEC Receiver
IPERF DONE.
The result is 1.59, not perfect , But still better than OpenWrt.
0
1
2
3
4
5
6
7
8
9
pi@FriendlyELEC:~$ sudo ./sbc-bench.sh -m
Rockchip RK3568 (35682000), Kernel: aarch64, Userland: arm64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 408 1992 Cortex-A55 / r2p0
Thermal source: /sys/devices/virtual/thermal/thermal_zone0/ (soc-thermal)
Time CPU load %cpu %sys %usr %nice %io %irq Temp
^C
The CPU is running at full speed again, and it is far from 100% utilization, so I think the problem should lie elsewhere. Now we can use ethtool again to check the eth2 information and statistics.
TML3H0H0
TML3
0TML3
L6 $ SUDO ETHTOL -i ETH2
DRIVER: R8125
Version: 9.008.00 -napi
FI6Fi rmware-version:expans-Rom-version:
bus-info: 0001: 11: 00.0
Supports-Statistics: YES
SUPPPOSTS-TEST: noTML3
SUPPPPROM-ACCESS: no
supports-register -dump: yes
supports-priv-flags: no
0
1
2
3
4
5
6
7
8
9
sudo ethtool -S eth2
NIC statistics:
tx_packets: 8506609
rx_packets: 12553353
tx_errors: 0
rx_errors: 0
rx_missed: 0
align_errors: 0
tx_single_collisions: 0
tx_multi_collisions: 0
unicast: 12553209
broadcast: 144
multicast: 0
tx_aborted: 0
tx_underrun: 0
tx_octets: 12543719502
rx_octets: 18471602900
rx_multicast64: 0
tx_unicast64: 8503557
TX_BROADCAST64: 3035
TX_Multicast64: 17
tx_PAUSE _ON: 35
TX_PAUSE_OFF: 35
TX_PAUSE_ALL: 70_Deferred: 0Htmmmm l7tx_late_collision: 0
tx_all_collision: 0
tml6_aborted32: 0ALIGN_ERRORS32: 0
RX_FRAME_TOO_LONG: 0
RX_RUNT: 0
HTMMML3HTMM l0rx_pause_on: 0RX_PAUSE_OFF: 0RX_PAUSE_ALL: 0TML3rx_unknownown_opcode: 0
rx_mac_error: 0
tx_underrun32: 0r6r6r6r6r6r6r6r6r6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6R6, x_mac_missed: 335
rx_tcam_dropped: 0
TDU: 0
RDU: 35RDU: 35RDML6RDML6RDML6RDML6RDML6RDML6RML0R6RDU: 35htmlml 7
test results here Among them, I found more rx_mac_missed. So I guess there should be some adjustments to improve performance.But according to my previous pair of RTL8156B adjustment settings, adjustment settings are really tricky, and experienced people seem to be unanimous on the specific adjustment option. I mainly refer RTL8156/8125 driver Realtek engineers, as well as some online experts in readers.is configured between the two 2.5GBE interfaces nat
Since the 2.5GBE interface cannot reach the best cooperation with IPERF3, I don't worry about testing the performance of the router in Friendlywrt, but there are still a few people asking me. So next I will show how I configure NAT in Ubuntu 20.04, and I will continue to test NAT performance. Remember that it will definitely improve it in a few weeks or months.
is here, we need to enable IP forwarding and NAT. The instruction I use is adapted from a post on NETWORKREVERSE .
编辑/etc/sysctl.conf以启用IP转发（取消注释以下行）：
net.ipv4.ip_forward=1
应用更改：
sudo sysctl -p
现在我们启用NAT：
sudo iptables ! -o lo -t nat -A Postrouting -j Masquerade
We can PING on the laptop of 192.168.2.0 subnet. Treme i11:
jaufranc@CNX-LAPTOP-4: ~ Ping 192.168.3.100
ping 192.168.3.100 (192.168.3.100 ) 56 (84) bytes of data.
SUDO APT Install iPtables-Persistent
SUDO SH -C 'iPtables-save /ETABLES.v4'
I tried IPERF3 between UP XTreme I11 and my laptop.
0
1
2
3
4
5
6
7
8
9
jaufranc@cnx-laptop-4:~$ iperf3 -t 60 -c 192.168.3.100 -i 10
Connecting to host 192.168.3.100, port 5201
[ 5] local 192.168.2.130 port 59430 connected to 192.168.3.100 port 5201
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-10.00 sec 914 MBytes 767 Mbits/sec 355 1011 KBytes
[ 5] 10.00-20.00 sec 912 MBytes 765 Mbits/sec 324 1.23 MBytes
[ 5] 20.00-30.00 sec 917 MBytes 769 Mbits/sec 124 1.09 MBytes
[ 5] 30.00-40.00 sec 915 MBytes 767 Mbits/sec 150 942 KBytes
[ 5] 40.00-50.00 sec 915 MBytes 767 Mbits/sec 78 1.22 MBytes
[ 5] 50.00-60.00 sec 919 MBytes 771 Mbits/sec 64 1.03 MBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.00 sec 5.36 GBytes 768 Mbits/sec 1095 sender
[ 5] 0.00-60.06 sec 5.36 GBytes 767 Mbits/sec receiver
iperf Done.
jaufranc@cnx-laptop-4:~$ iperf3 -t 60 -c 192.168.3.100 -i 10 -R
Connecting to host 192.168.3.100, port 5201
Reverse mode, remote host 192.168.3.100 is sending
[ 5] local 192.168.2.130 port 59434 connected to 192.168.3.100 port 5201
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.09 GBytes 935 Mbits/sec
[ 5] 10.00-20.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 20.00-30.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 30.00-40.00 sec 1.09 GBytes 938 Mbits/sec
[ 5] 40.00-50.00 sec 1.09 GBytes 939 Mbits/sec
[ 5] 50.00-60.00 sec 1.09 GBytes 937 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-60.05 sec 6.55 GBytes 937 Mbits/sec 973 sender
[ 5] 0.00-60.00 sec 6.55 GBytes 937 Mbits/sec receiver
iperf Done.
has a transmission speed of 768 Mbps in one direction and 937 Mbps in the other direction.@Friendlyelec: ~ $ suno ./sbc-bench.sh -mrockChip RK3568 (35682000), kernel: AARCH64, userland: ARM64
CPSFS TOPOLOGY (Clusters, CPUFREQ MEMBERS, CLOCKSPEEDS)
CPUFREQ Minl7
CPU Cluster Policy Speed Speed Core Type
0 0 408 1992 Cortex-A55 / R2P0
Thermal Source:/SYS/DEVICES/VIRTUAL/Thermal_zone0/(SOC-thermal)
time CPU load % cpu % sys % usr % nice % IRQ TEMP
1: 48: 1992MHz 0.90 25% 0.90 25% 0 % 0% 0% 0% 0% 25% 54.4 ° C
H6
If you use SBC-Bench.sh, the running frequency will be 1992 (Actually 1845 MHz), and 25%IRQ should be just It means that a Core has been completely used to deal with IRQ.According to the mpstat command display, this should be processed by core #0.
$ mpstat -P ALL -I SUM
Linux 5.10.66 (FriendlyELEC) 06/05/22 _aarch64_ (4 CPU)
The above point can be confirmed by using top and htop.
NanoPi R5S power consumption
I happen to have a wall power meter and can use it to check power consumption:
- Idle - 5.1 W
- Iperf3 to eth1 - 6.3 to 6.7W
- NAT test between laptop and mini – 6.2W
The above data was tested after installing Ubuntu 20.04 and NVMe SSD on this product.
I also tested on a NanoPi R5S using OpenWrt and no SSD:
- Idle - 4.6W
- Iperf3 - 6.0 to 6.2 W
Note that since I am using a wall-mounted power meter, this number will include efficiency losses in the power adapter (Khadas VIM4 USB-C Power Adapter). And the value may be higher than using a USB-C power meter. This should also be possible through settings to optimize power consumption.
final conclusion
That’s it for today’s evaluation. The optimization part should actually include "changing the firmware so that the Rockchip core runs at 1992 MHz", "adjusting various settings related to PCIe and Ethernet settings" and so on. Most of them I'm not familiar with yet.
(3), Taipower Linglong M NUC Flash X86 Soft Router Review
Foreword
Many friends have bought X86 soft routers and most of them have been idle and dusty, or they are preparing to step into the pit of X86 soft routers! First of all, you need to make it clear whether you are willing to spend time to fiddle with it. If you are not willing to spend time to study, I advise you to give up and consider buying X86 soft routing. You can consider buying a finished opnenwrt router to experience it, or you can buy a router with Merlin firmware or a router with PandoraBox firmware to play with.
Prepare relevant network equipment
A suitable 2.5G port switch and an X86 soft router with a 2.5G electrical port, and an external wired network card on the Type-C side of the 2.5G network port or the USB-A side.
Teclast Linglong M NUC is a mini machine that is very suitable for use as an X86 soft router. It has two 2.5GE electrical ports, two SO-DIMM memory slots and a PCIe 3.0 NVMe SSD M.2 slot. It is recommended to buy bare metal and bring your own memory modules and NVMe SSD. Two 4GB Micron DDR4 3200MHz memory sticks and one Intel Optane M10 16GB (Optane memory) are prepared here.
After installing the memory and NVMe SSD, take the prepared Router OS U disk to install the soft routing system.
can be used after powering on.
Xike is a new network equipment brand. It is a layer 2 management switch that comes with 8 2.5G electrical ports and 4 10G optical ports. It supports port aggregation and VLAN division.
A friend bought this XIKE switch and asked me to help upgrade the network at home, planning to create a 2.5G intranet environment for the whole house.
Starting to debug the equipment is very simple for experienced people.
can see that the 10G optical port of the XIKE SKS7300-8GPY4XGS switch has been turned up, and the connection status is 10G Full rate. For friends who have 10G SFP modules, they can use 10G LAN. However, at this stage, 2.5G LAN is more suitable for the actual experience needs of the public. If you upgrade to a 10G optical port, the overall hardware cost will be significantly higher several times. For small and medium-sized enterprises, this switch is very suitable for use as an access switch connected to a 10G core switch. It is still more recommended for home users who are in need of 2.5G electrical ports and friends who like to mess with 2.5G LAN to buy this XIKE switch.
For users who have a GPON Stick ONU module, plugging it directly into the SFP interface of this SKS7300-8GPY4XGS switch can save an optical modem. It effectively solves the problem of insufficient accumulation of equipment in the weak current box, and can also improve the space inside the weak current box and reduce the problem of heat accumulation between devices.
What should I do if my computer (desktop and laptop) does not have a 2.5G network port? You can buy a superior USB to RJ45 2.5G wired network card. Connect to the computer through the Type-C end, and the RJ45 network port on the other end has a 2.5GbE (2500Mbps) rate.
Take a USB flash drive to make a boot disk
I generally like to use the tool BalenaEtcher to make a USB flash drive. It is simple and easy to operate! Supports ISO, DMG, IMG, IMG.GZ and other image file formats to create boot disks.
makes the Sea Spider V9 boot disk (flash)
. Go to the Sea Spider official website and find the download area in the service and support. Scroll to the upgrade address at the bottom, find V9.0, and select Download Now next to it to download the Sea Spider V9 firmware. Note that the ISO image file is downloaded.
Sea Spider V9 system
default login address: 192.168.0.1. The account and password are set by yourself after flashing the machine.
You can view the current information overview on the overview. The instrument icon shows CPU usage, memory usage, and disk usage. Below is the monitoring of the real-time total traffic of the intranet, and on the right is some key information. The LAN port already in use by
can identify the network port model: Realtek RTL8125 2.5GbE. The network port icon also identifies 2.5Gb.
Nigou can view the current memory usage in the overview. After flashing and initializing the system, you can see that 10.5% is used by default and 835.86 MB of storage space has been used. Total memory size: 7.6GB (8GB total memory), remaining 6.8GB available.
You can also view the CPU load in the overview. The
interface load can also view the data of the LAN port. The LAN port here is running in full-duplex mode, 2500Mb/s (that is, the 2.5GbE Gigabit port rate).
You can see the motherboard/CP information in the hardware information. Basically, the relevant information displayed is complete.
Here is the 16GB Optane memory, available capacity: 13.41GB, you can check the partitioning of the hard disk storage.
Both 2.5GbE Gigabit network ports can be recognized normally.
Haispider has module management, including Docker container modules.
Basically these modules are very complete for soft routing, including VPN IPsec, PHP suite, IGMP multicast proxy, etc.
FTP service.
file storage and sharing.
Comment: Sea Spider V9 requires purchasing relevant authorizations to fully use all functions. Fortunately, there is a long trial period. Relatively many functions are more suitable for network workers. For ordinary novice users, the operation may be very difficult, and some professional terms are convoluted and difficult to understand.
Make iStore OS boot disk (flash)
Go to KoolCenter official website to download iStore OS firmware. Find the iStore OS directory in the firmware list and enter it.
Considering that the current new X86 soft routing is booted by UEFI, so select X86_64_EFI to download the iStore OS firmware.
Generally has Date (Date) in the bottom list. Select the latest Istore OS firmware.
istore OS
In fact, ISTORE OS is the in -depth customization of Chinese people. Compared with the original OpenWrt, the gameplay (plug -in) and Docker gameplay.
homepage will have a legend with traffic statistics, IP addresses, and network interface status. It looks clear at a glance.
pulls down to see disk information, storage services, docker, download services and remote domain names. The playability is still very high.
can be seen at the bottom to see the system information. The dog recognizes the CPU temperature, CPU usage rate, and memory usage.
network wizard, this is the best direction function for Xiaobai users.
has system information. You can see the computer model, processor architecture and other related information.
can also see the use of memory at the overview.
Soft Route is difficult to master a disk management. Here ISTORE OS directly completes a graphical Diskman disk management, which reduces the difficulty of novice operation and enhances the simpleness of use.
istore OS built -in Docker
is definite It is difficult to install plug -ins. Many beginners or little white users who have just started. The iStore OS directly built -in Docker components for easy deployment of containerization. And supporting CPU and memory configuration settings in advanced settings, which is very user -friendly.
ISTORE OS Unique App Store
For wanting to study the smart home experience, you can consider Home Assistant.
is easy to be suitable for miniature home deployment data service centers. Many tutorials on the Internet will not be discussed in depth.
If you can't afford a dedicated NAS device, in fact, soft routes can also act as simple NAS, but most of the X86 soft routes are not easy to use. There are two tools that can be solved directly here to solve this problem. For friends who want to build an audio -visual center: Jellyfin and NAS Tools. If you can use it, it can definitely easily make the X86 soft routing into a NAS.
ARIA2 downloader. If you can use it, it is also a very practical download tool. For the two broadband at home, in order to avoid the broadband during the work or during the business trip, try to deploy the net Xinyun to earn some benefits.
I actually prefer HOMEBOX inner mesh speed measuring. This is better experienced by the Speedtest speed measuring server. I will show the experience of the HOMEBOX intranet speed measuring experience later.
DISKMAN disk management is definitely a plug -in worth installed. Friends who use QBITTORRENT downloaders will use it to use it. The specific use methods and tutorials have more detailed professional introduction on the tutorial network. If you are interested in the research, go to the study.
KMS server can solve the problem that Windows and Office prompts not activated.
installation plug -in is as simple and convenient as installing the app, which is also the biggest ease of use of Isstore OS.
I use most of these four plug -ins. Among them, multi -line multi -broadcast is suitable for broadband overlay areas. One sentence can run to 600m in one sentence, 500M superposition can run 1000m but do not need to upgrade broadband packages and extra money. It is worthwhile.
KMS server is enabled. Basically, local Windows and Office have no longer encountered unauthorized situations.
This is the HomeBox intranet speed test, which can directly measure the speed of the intranet. I use Teclast Linglong M NUC’s 2.5GbE Gigabit network port, CAT6 (Category 6e) network cable, and Type-C to RJ45 2.5GbE external wired network card, which can achieve a 2500Mbps full-link LAN (intranet) experience .
Summary
Users with basic network engineering skills or a relatively good understanding of the Internet can consider starting X86 soft routing. Sea Spider V9 actually prefers traditional network equipment, combining functions such as switches, routers, and AC controllers. iStore OS is suitable for the general public. It is more playable and easier to get started. I personally recommend flashing the iStore OS firmware to use it, which is basically the Router OS in openwrt that I have used and experienced well.
