Foreword
Intel announced the launch of the Xeon E-2100 processor in 2018 for entry-level server platforms. Before the Xeon E was renamed, it was the E3-1200 v series. v2/v3 was once known as "the god of foreign garbage generation." U".
▲Xeon E-2100 family, up to 6 cores and 12 threads, acceleration 4.7GHz, LGA1151 interface (compatible with C246 motherboard), up to 40 PCIe 3.0 channels 3.1 Gen2, Thunderbolt 3, Gigabit Wi-Fi, advanced features support including 128GB DDR4-2666 ECC memory, Intel SGX software protection technology, etc.
▲ There are 10 models of XEON E-2100 series, among which the suffix "G" represents the integrated UHD 630 nuclear display, while E-2136/2124 does not.
▲ The standard chipset of the Xeon E-2100 and 220 series was originally the C246 chipset, but the C246 motherboard is very expensive.
▲ But ASRock's motherboard can be cracked by BIOS ME FW to support Xeon E-2100 and 2200 series on the H310 motherboard, such as this ASRock H310CM-HDV. Unpack below.
box
▲ front of the package
▲ back of the pack
▲ full accessory
▲PCB front
▲PCB back
_pDVIp
▲ one HDMI can display three Among them, HDMI supports 4K. The PS2 keyboard and mouse interface is very practical and thoughtful. It has 4 USB2.0 ports and 2 USB3.1 Gen1 onboard, and an RTL8111H Gigabit network card. In most cases, it is enough. The key point is the size of 18x19, which is only a little larger than miniITX. , The price is much cheaper.
▲Two memory DIMMs support DDR4-2666. Don't think about overclocking. The highest memory support in BIOS is 2666. The expansion slot contains one PCIe 3.0 x1 and one PCIe 3.0 x16.
▲There are 4 SATA6G ports on board, all of which are PCH native ports.
▲Because of the test, I borrowed a XEON E-2124 the most beggar version of XEON E.
▲The main frequency is 3.3 quad-core four-wire, the all-core turbo frequency is 3.9GHz, and the single-core turbo frequency is up to 4.3GHz.
▲CPU IMC supports memory DDR-2666, the maximum support capacity is 128GB
▲The radiator uses two ultra-short American CSC DDR4-3200 ECC UDIMM 32GB of memory total 64GB
test
test configuration
_span3 span_span5 p124 ▲ Span_span2 ：ASRock H310CM-HDV (BIOS P4.2C)
Memory: Mircon DDR4-3200 ECC-UDIMM 32GB x2
SSD: Intel DC-S3520 800GB SATA pProp2 Span _p7 970
cooling: Cryorig C7
graphics card: MSI Ventus 2X Geforce RTX3070 8GB
power: Co RSA ir SF600
operating system:
MircoSoft Windows 10 X64 1909 Workstation Edition
Ubuntu 18.04 X64 Desktop Edition
BIOS
▲For this motherboard to support XEON E, you need L4.24C version of BIOS support _p7 img_
p2 ▲ think,DDR4-2666 to the top, and there is no ECC setting
▲There are basically no movable options in advanced CPU settings
img22p pimg
▲There is basically no span5span
img22p pimg
_p7
pimg
_p7
For moving things, the overclocking setting is invalid for XEON E.
CPUZ test
▲From the results of the CPUZ test, the single-core performance is close to I7-7700K, but because XEON is unable to crack the E-2124, the single-core performance is only close to I7-7700K. Multi-core performance still lags behind I7 7700K.
▲If you have to find a comparison position in the quad-core CPU, XEON E-2124 single-core and multi-core performance are closer to the default performance of I5 7600K, let’s take a look below. .
baking machine temperature test
▲ FPU using AIDA64 single copy test 8 minutes _span5 p5 p2 p2
5
2 _p img p img p5 p2
near 61 degrees for 8 minutes Use HWINFO to detect that the main frequency at the moment is quad-core 3.9GHz.
▲Using HWINFO to monitor the main frequency power consumption and temperature list at this time, it can be seen that the CPU power consumption at this time cannot reach 71W, but can only reach around 51W. The CPU temperature can be stabilized at around 63 degrees, and the main frequency is still at 3.9GHz.
The conclusion is that even with the AVX grill test, XEON E-2124 cannot reach 71W, and only has a power consumption of around 50W. If the temperature is used, the full load is within 63 degrees with Cryori C7. So I feel that this CPU is still more suitable for home servers such as NAS.
ECC validity diagnosis
Xeon E-2100 has always supported ECC UDIMM memory, but it needs C246 motherboard support, then does ASRock H310CM-HDV support ECC UDIMM this time?
ECC will take effect, There seem to be three necessary conditions:
1. memory controller (CPU) support.
Obviously, Xeon E-2124 supports ECC.
2. Must be ECC memory.
This is obvious. The memory I use is standard ECC UDIMM memory.
ECC UDIMM memory has one more storage module than ordinary memory to store redundant bits. The RDIMM memory has one more REG chip.
3. Motherboard/chip/BIOS support.
This is the most controversial. Do you need motherboard support? Some motherboards/BIOS have the option to turn on/off ECC, some do not. ASRock H310CM-HDV BIOS P4.2C does not have ECC support options, so it is likely that ECC is not supported. There
Ubuntu 18.04 next few commands can detect ECC:
first command:
sudo dmidecode -t memory
▲ valid here The message shows that ECC is not supported.
The second command:
dmesg |grep -i edac
_blockquote img30
_blockquote 351_blockquote
img30
As explained above, although ASRock H310CM-HDV can support XEON E-21XX series CPU, it does not support XEON E's ECC support feature.
phoronix-test-suite
Phoronix test suite is currently the most comprehensive test and benchmark test platform available under LINUX,It provides an extensible framework to easily add new tests. The software is designed to efficiently perform qualitative and quantitative benchmarks in a clean, reproducible, and easy-to-use way. The Phoronix test suite can be used to simply compare computer performance, hardware verification, and continuous integration/performance management. So the following test was performed under Ubuntu 18.04 x64.
▲Hardware testing of phoronix-test-suite
CPU benchmark
strong401 strong 401 strong 401 strong 401 Sysbench Sysbench 1.0.0 _span strong _ Sysbench 1.0.0 is another widely used test It was originally created in 2004 by Peter Zaitsev. Soon thereafter, Alexey Kopytov took over its development. It reached version 0.4.12, and development stopped. After a long break, Alexey started working on SysBench again in 2016. Soon, version 0.5 was released, and the OLTP benchmark was rewritten to use LUA based scripts. Then SysBench 1.0 was released in 2017. Compared with the old 0.4.12 version, the most important thing about 1.0 is that you can use LUA to customize the benchmark instead of using hard-coded scripts. This time I use the interim CPU test, and the results are calculated based on transactions processed per second. The higher the result, the better.
▲ traceability test as evidence
▲ polynuclear efficacy
▲ mononuclear efficacy
Stockfish 9
▲Stockfish is an advanced C++11 chess benchmark,It can be expanded to 128 CPU cores, and the result is based on the average amount of data processed per second. The higher the result, the better. Traceability: https://stockfishchess.org/
▲Test traceability certificate
p5p img single-core 3span img
38 Nuclear performance
asmFish 1.1.1
▲asmFish is an advanced measurement of chess prepared in the Assembly and the result is an average chess benchmark test per second. The higher the better.
▲ test traceability as evidence
▲ multicore performance
▲ single-core performance
CPU rendering
Blender 2.81
test conditions: CPU-only rendering operations
BMW27
▲ traceability test as evidence
▲ multicore performance
Classroom
▲ test traceability as evidence
▲Multi-core performance
Fishy Cat
p 2p
▲ test traceability as evidence
▲ multicore performance
Barbershop
▲ traceability test as evidence
▲ polynuclear efficacy
Barcelona
▲ traceability test as evidence
▲Multi-core performance
POV-Ray 3.7.0.7
The test is executed using the Benchmark script that comes with POV-Ray V3.7,This is a complete CPU rendering test, which tests multi-threaded and single-threaded results separately. The unit is expressed by the time to complete the rendering task. The lower the result, the better.
▲ traceability test as evidence
▲ polynuclear efficacy
▲ mononuclear efficacy
C-Ray 1.1
C-Ray 1.1 is a ray tracing benchmark, which can show the difference of processors under multi-threaded workload, similar to Windows Cinebench, the lower the better.
▲ traceability test as evidence
▲ polynuclear efficacy
▲ mononuclear efficacy
molecular dynamics
NAMD 2.13
▲ NAMD is a parallel molecular dynamics code,Designed to perform high-performance simulation of large-scale biomolecular systems. NAMD was developed by the Theoretical and Computational Biophysics Group of the Beckman School of Advanced Science and Technology at University of Illinois at Urbana-Champaign . More information about the benchmark can be found here.
▲ traceability test as evidence
▲ polynuclear efficacy
video decoder
x264 v2018-09-25
use A simple test of the sample 1080p video file on the x264 encoder running on the CPU, measured by the average number of frames processed per second, the higher the result, the better.
▲ test traceability as evidence
▲ multicore performance
▲ single-core performance
x265 v3. 1.2
A simple test of the x265 encoder running on the CPU using a sample 1080p video file,Measured by the average number of frames processed per second, the higher the result, the better.
▲ test traceability as evidence
▲ multicore performance
▲ single-core performance
compiled
Timed Linux Kernel Compilation 5.4
The time required to compile the Linux kernel in the default configuration: linux-5.4-rc3.tar.gz for performance comparison. The lower the result, the better.
▲Test traceability certificate
▲Multi-core performance
compile with the same default VM strong401
404 strong
strong 404
p The time required for the task to perform performance comparisons. The lower the result, the better.
▲ traceability test as evidence
▲ polynuclear efficacy
Timed PHP Compilation v7.1.9
same default configuration The time required for the PHP 7 build task,In order to perform performance comparisons, the lower the result, the better.
▲ test traceability as evidence
▲ multicore performance
▲ single-core performance
Build2 v0. 12
This test profile measures the time to boot/install build2 C++ build toolchain from source code. Build2 is a cross-platform build tool chain for C/C++ code and has features similar to Cargo, the lower the result, the better.
▲ test traceability as evidence
▲ multicore performance
depth study
oneDNN MKL-DNN 1.1
oneDNN MKL -DNN 1.1 is a test of Intel oneDNN (previously known as DNNL/Deep Neural Network Library/MKL-DNN), which is an Intel optimized library for deep neural networks,And using its built-in Benchnnn function, the test model uses recurrent neural network inference-data type: f32, the result is the total performance time reported. It is worth noting that this is a test that INTEL has optimized for its own CPU.
is optimized for the following CPUs:
Intel Atom processor with Intel SSE4.1 support
4th, 5th, 6th, 7th and 8th generations Intel(R)Core(TM) processor
Intel® Xeon® processor E3, E5 and E7 family (formerly known as Sandy Bridge, Ivy Bridge, Haswell and Broadwell)
Intel(R) Xeon Phi (TM) processors (previously Knights Landing and Knights Mill)
Intel Xeon Scalable Processors (previously known as Skylake, Cascade Lake and Cooper Lake)
Future Intel Xeon Scalable Processor (code-named Sapphire Rapids)
On CPUs based on Intel 64 or AMD 64 architecture, oneDNN detects the instruction set architecture (ISA) at runtime and uses just-in-time (JIT) code generation to deploy the latest support ISA optimized code. Future ISAs may disable the initial support in the library by default and need to use runtime controls to enable them.
▲ test traceability as evidence
▲ multicore performance
Intel Open Image Denoise 1.0.0
Intel ®Open Image Denoise The purpose is to provide an opening,A high-quality, efficient and easy-to-use denoising library that can significantly reduce rendering time in ray tracing-based rendering applications. It can filter out the Monte Carlo noise inherent in random ray tracing methods (such as path tracing), thereby reducing the number of samples required for each pixel by even multiple orders of magnitude (depending on the required closeness to the ground truth) . A simple but flexible C/C++ API ensures that the library can be easily integrated into most existing or new rendering solutions.
The core of the Intel® Open Image Denoise library is a set of high-efficiency noise reduction filters based on deep learning. These filters are trained to handle every pixel (spp) from 1 spp to almost complete convergence. Various samples. Therefore, it is suitable for preview and final frame rendering. The filter can use only noisy color (aesthetic) buffers to denoise the image, or in order to preserve as much detail as possible, you can also choose to use auxiliary feature buffers (such as albedo, normal). Most renderers support this type of buffer as an arbitrary output variable (AOV), or can usually be easily implemented.
Although the library comes with a set of pre-trained filter models, it is not necessary to use these models. In order to optimize the filter for a specific renderer, sample number, content type, scene, etc., you can use the attached training toolkit and the image data set provided by the user to train the model.
Intel® Open Image Denoise supports Intel® 64-based CPUs and compatible architectures, and can run on any device from laptops, workstations to computing nodes in HPC systems. It is efficient enough not only for offline rendering, but also for interactive ray tracing depending on the hardware used.
Intel Open Image Denoise is built internally on the Intel oneAPI deep neural network library (oneDNN),And automatically utilize modern instruction sets such as Intel SSE4, AVX2 and AVX-512 to achieve high denoising performance. To run Intel Open Image Denoise, a CPU supporting at least SSE4.1 is required.
▲16 spp to render Moana Island Scene, and use Intel® Open Image Denoise to denoise. The publicly available dataset is provided by Walt Disney Animation Studios.
▲ test traceability as evidence
▲ multicore performance
decryption
John The Ripper v1.9.0-jumbo-1
The benchmark test ofJohn The Ripper is a password cracking program, and Blowfish is one of the models.
▲ traceability test as evidence
▲ polynuclear efficacy
▲ mononuclear
Python effectiveness Effectiveness Evaluation
pybench1.1.3
▲This test reports the total time of different average timing test results from PyBench.PyBench reports the average test time of different functions (such as BuiltinFunctionCalls and NestedForLoops), and the total results provide a rough estimate of the average performance of Python on a given system. The test profile runs PyBench 20 times each time.
▲ traceability test as evidence
▲ polynuclear efficacy
▲ mononuclear efficacy
solution compression
7-zip Cpmpression 16.02
7-zip is a widely used compression/decompression program for cross-platform work.
▲ traceability test as evidence
▲ polynuclear efficacy
▲ mononuclear efficacy protocol performance
OpenSSL 1.1.1
OpenSSL is an open source toolkit,It can implement SSL (Secure Socket Layer) and TLS (Transport Layer Security) protocols. This test measures OpenSSL's RSA 4096-bit performance.
▲ test traceability as evidence
▲ multicore performance
▲ single-core performance
▲I tested a total of 133 items in this test, of which 8 items were not compared, so only 125 items passed and were valid.
Xeon E-2124 itself 4C4T in comparison with the 8C16T and even 16C32T CPUs fell in the wind is a normal thing, but we mainly look at the following single-core IPC performance, but in fact, the single-core performance is also inferior. Downwind.
I don't think that the conclusions drawn by a large number of tests are 100% valid, but it can reduce the probability of error, and the results of statistical regression can restore the truth as much as possible.
Summary
The conclusion is that although the Xeon E-2124 CPU is now outdated to the ranks of foreign garbage, the single-core performance is indeed 10-30% behind the current CPU, and the multi-core performance gap is even greater, the only one The highlight is that the power consumption and temperature are not bad, but because there is no built-in HD 630 that requires a separate display to support, Xeon E-2124G will be more suitable for NAS CPU. For ASRock H310CM-HDV this motherboard,It only supports XEON E-2100 series, but does not support ECC features, and there are several issues that need attention.
1. The H310CM-HDV purchased by JD or Taobao Tmall does not support XEON E by default, and a special version of BIOS is required.
, but you need a 6-8th generation CPU to light up the motherboard when you brush the BIOS yourself. This will cause a lot of trouble. In fact, Taobao also has stores that provide pre-flashing BIOS services. After flashing the 4.2C BIOS, there is a point It is worth noting: The driver of INTEL ME cannot be installed. It should be that this option is blocked inside the BIOS, because it is ME FW that locks XEON, which is easy to understand. In addition, this motherboard also supports WIN7, which is currently one of the few examples.
XEON E-2124 official version is around 500-620 yuan, 2124G is around 850 yuan, indeed XEON E-2124+ASRock H310CM- HDV can achieve within 1K. This evaluation only provides an objective and true evaluation of the performance of the board U kit for the garbage guys. As for whether this kit can become a new generation of garbage guys artifacts, it is not the focus of my evaluation. I am only responsible for the evaluation and reports that seek truth from facts.
This article is reprinted with permission from the author of the "What is Worth Buy" community author @gaojie20. The content of the article only represents the author's point of view. Please do not reprint without authorization.
.
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-