RTX 3090 What is the improvement of Ti
Preface
526 days after the release of RTX 3090, NVIDIA officially released the consumer-grade top flagship product - NVIDIA GEFORCE RTX 3090 Ti 24GB. Lao Huang’s operation actually made many people not understand. His own RTX 3090 firmly sits at the top of flagship products, and Su Ma’s RX 6950 has not come yet. What’s the rush? The complete GA102 core makes it hot and bright in the industry. Can’t money make more money?
And in this generation of RTX 30 series graphics cards, the number of SKUs for mid-to-high-end graphics cards is really large. I gave up the Super suffix of the RTX 20 series, but picked up the suffix of Ti. At the same time, the graphics cards with the graphics memory capacity as the suffix have also come out. Including the RTX 3090 Ti, the number of products based on the core of GA102 has reached 5, and the wind is indeed a bit messy.
At the same time, according to the entire RTX 30 series product life cycle, the ones that should be focused on mid-range and entry-level graphics cards, the focus is to increase the core production capacity of GA106/107 (or GA117?). But even so, the established fact is: NVIDIA GEFORCE RTX 3090 Ti 24GB is really here.
Is it really that GA102 core production capacity is too much? Hurry up to clear the price? Or is it another undescribable reason? We can't control this matter anymore. Let's take care of it. Is the release of RTX 3090 Ti more satisfying game users or content creators?
What are the improvements of ?
Improvement 1: Complete GA102 core
As a top flagship product, this time the RTX 3090 Ti finally used the full-spec GA102 core, which greatly satisfied many coercive high-end players . After all, the GA102 core has always been used on the RTX A6000 graphics card in the industry.
original image click to enlarge
From the specification table, the complete GA102 core RTX 3090 Ti has higher specifications. Compared with RTX 3090, it has 10,752 CUDA stream processors, 84 RT Cores, 336 Tensors Cores, and 112 rasterization processing units. Theoretically, the RTX 3090 Ti has better performance, but the small gap in the core is indeed difficult to get a clear gap with the RTX 3090. Let’s take a look at what direction the RTX 3090 Ti has to strengthen this time.
Improvement 2: higher core frequency, higher power consumption
Original image Click to enlarge
RTX 3090 Ti core basic frequency reaches 1560MHz, and Boost frequency reaches 1860MHz (FE public version). However, the RTX 3090 Ti FE public version is really not much, so NVIDIA directly opened it to AIC and started to build it. Most of the RTX 3090 Ti non-public graphics cards Boost frequency reaches 1890MHz, and the core frequency during actual gaming is naturally higher.
The disadvantage of higher core frequency is the increase in power consumption. The default TDP of RTX 3090 is 350W, and the recommended power supply requirement is 750W. The default TDP of the RTX 3090 Ti has reached 450W, and most non-public MAX TDPs have reached 480W, which has increased the demand for power supply. NVIDIA's official recommendation is to use an 850W power supply, but the author here recommends that you use a 1000W or 1250W power supply (preferably a power supply with a 12VHPWR interface native module cable), which is better compatible with future RTX 40 series graphics cards.
Improve 3: video memory particles with higher density and frequency
RTX 3090 Ti has also upgraded its video memory this time, and for the first time, it uses 512Mb*32Bit and 21Gbps specification GDDR6X particles (with ECC), and the particle FBGA number is: D8BZC. Although Micron has not officially announced the detailed specifications of this model to the public, interested netizens can refer to the GDDR6X particle specifications numbered D8BGX, which are also 21Gbps, but the particle density is only 256Mb*32Bit [MT61K256M32JE-21].
We all know why the improvement of video memory frequency is beneficial, but many netizens do not understand what the terms 512Mb*32Bit means. Let me briefly explain here that the RTX3090 currently uses 256Mb*32Bit and 19Gbps GDDR6X particles, numbered as D8BWW; in conversion, the capacity of a single video memory is 1GB. If you want to form the 24GB of the RTX 3090, you need to line up 24 video memory particles on both sides of the front and back PCB, which also causes most users who purchase RTX 3090 graphics cards to complain about the high memory temperature.
The GDDR6X particle D8BZC with 512Mb*32Bit specification is converted into a single 2GB video memory capacity, which only requires a total of 12 video memory on the front. This way, the PCB design is simpler, and the heat generation control of the video memory module is also better. After our actual measurement, the temperature of the video memory monitoring point is indeed much lower.
Improvement 4: Video memory with ECCh function
RTX 3090 Ti is the only graphics card in the consumer market with ECC (Error-correcting code). This function solves the video memory error problem that graphics card has in long-term high-load applications, and can provide better guarantees for the reliability of production applications.
By default, the ECC function is not enabled on the RTX 3090 Ti. Everyone needs to find the corresponding ECC activation option on the NVIDIA control panel; after the ECC function is enabled at the same time, the performance of the graphics card will be reduced accordingly. The ECC function generally only exists on industry graphics cards, such as the RTX A6000, which is also the core of GA102.
Improvement 5: 12VHPWR auxiliary power supply interface
12VHPWR auxiliary power supply interface, one of the hottest topics discussed on the Internet, this will be the power supply interface of PCIe 5.0 graphics cards in the future. According to the information we have learned, this 12VHPWR interface can provide up to 600W of power supply. The RTX 3090 Ti non-public graphics card is equipped with this 12VHPWR auxiliary power supply interface first, which means that players can replace the new power supply from now on?
The graphics card interface uses 12VHPWR, but we must connect to the 12VHPWR wiring to use it? The implementation of the small 4Pin definition in the 12VHPWR has not yet been set. We can directly use the Micro-Fit 12Pin interface interface to assist in power supply.
The 30 series FE graphics card configuration is all dual*8Pin auxiliary power supply conversion cables, and in order to ensure that the auxiliary power supply reaches 450W, the AIC manufacturers have equipped us with Micro-Fit 12Pin To 3*8Pin conversion cables, so that everyone does not have to worry about the conversion cables. They just need to see if the power supply has 3 8Pin power supply interfaces and whether the power consumption is sufficient.
The specification of this wire given away is really good. Generally, the power cord uses 18AWG specification. In addition to the NVIDIA faith LOGO, this wire also uses 16AWG specification wire, which is thicker and has a larger current conduction value.
Interested netizens can study the line definition of Micro-Fit 12Pin To 3*8Pin conversion line.
Improvement 6: Reserve NVIDIA NVLink
Same as the industry card RTX A6000 and consumer-grade RTX 3090, the RTX 3090 Ti retains the NVIDIA NVLink 3.0 high-speed GPU interconnection port, which is not as simple as the SLI interface we learned before. This NVLink is generally only used in industry graphics cards. This technology connects two NVIDIA graphics cards to achieve performance expansion of graphics memory and performance.
The author takes the NVIDIA NVLink technology animation provided by NVIDIA to refer to it. According to the characteristics of the GA102 core, the RTX 3090 Ti uses the third-generation NVIDIA NVLink. After using the NVLINK bridge, the two RTX 3090 Ti can achieve a maximum transmission speed of 112 GB/s, and the extended video memory is up to 48GB, which can accelerate graphics and computing workloads, and can handle larger data sets, which can quickly form a high-performance desktop workstation.
But since most non-public graphics cards of the RTX 3090 Ti are basically designed with three slots, if you want two RTX 3090 Ti to form NVLink, then you have to check whether your motherboard has enough space and cooling space. As shown in the two RTX 3090 Ti back-to-back, the cooling space is obviously not enough, and the four-slot space is your best choice.
Summary: From the perspective of specifications, the RTX 3090 Ti is more like a high-frequency, equipped with GDDR6X video memory, and higher power RTX A6000 graphics card, but the RTX 3090 Ti is only a consumer-grade product, and many industry-specific functions have been removed.
Even so, the RTX 3090 Ti is still the most powerful existence in the consumer market. The specifications are the complete GA102 core, the core frequency, video memory specifications, and performance are absolutely powerful, and of course the power consumption is also a terrifying 450W.
RTX 3090 Ti's future
In terms of the industry, the Ampere architecture is no longer a new thing, after all, it has been released for a long time. Especially NVIDIA released the Hopper architecture on its recent GTC 2022 that was quite tempting. But it is undeniable that the Ampere architecture is the best architecture on the consumer market at present. It has a brand new SM multi-unit stream processor, high-performance single-precision floating point (FP32) computing processing speed, and excellent performance power consumption ratio, which can significantly improve the performance of graphics and analog workflows.
At the same time, the addition of the second generation RT Core and the third generation Tensor Core makes the graphics card more powerful in running ray tracing, shading or noise reduction functions, greatly speeding up the speed of workloads; at the same time, the AI computing power is also stronger, and higher-speed AI computing and data science model training is achieved without modifying the source code.
In fact, it is really time to release the RTX 3090 Ti. In the just-in-the-GTC 2022, NVIDIA not only released a series of high-spec hardware, but also many technologies that industry users and developers lead the way, such as: NVIDIA OMNIVERSE, NVIDIA DRIVE Map, NVIDIA Maxine, NVIDIA AI acceleration, etc. Although most technologies are not very understood by consumers, these technologies and applications will eventually be fed back to our users. With its strongest consumer-grade graphics card and a 24GB large graphics card, the RTX 3090 Ti can naturally be used for future software and games.
For example, DLSS technology, more than 150 games have now supported NVIDIA DLSS, and is still increasing, including a series of recently released game masterpieces, Ghost Line: Tokyo, The Disappearing Light 2 People and Ren’s Battle, Shadow Samurai 3, etc. Of course, there are more RTX 30 series graphics card features: NVIDIA DLAA, NVIDIA Reflex and RTXGI. If you don’t have a powerful RTX 30 series graphics card, you can’t play with it.
Yingchi RTX 3090 Ti Xingyao OC Appearance & Disassembly
Yingchi RTX 3090 Ti Xingyao OC
Let's return to our theme, this time we did not get the RTX 3090 Ti FE public version, but we received Yingchi GeForce RTX 3090 Ti Xingyao OC [hereinafter referred to as "RTX 3090 Ti Xingyao OC"].
Family design, each Xingyao graphics card has the same outer packaging design, the same series of appearance designs, of course, different levels of graphics cards have different workmanship and materials, especially this Yingchi RTX 3090 Ti Xingyao OC material is the most powerful in the Xingyao series, no one.
How to identify the family design depends on the product model on the outer packaging. NVIDIA's unified product model logo: NVIDIA GEFORCE RTX 3090 Ti.
The back is a description of some functional features of this graphics card , including Xingzhuo II Plus radiator, Xuanyao lighting effect, PCB design, Magic disk software, and a three-year warranty (supports personal insurance).
two practical accessories: graphics card support rod, and Micro-Fit 12Pin To 3*8Pin conversion cable.
The appearance is up and down, RTX 3090 Ti Xingyao OC adopts the classic Xingyao transparent crystal shell element design in Yingchi RTX 30 series. The white air guide cover is paired with a transparent crystal shell. You can customize the pattern by removing the crystal shell. It cannot be said that this block RTX 3090 Ti Xingyao OC is really big, the two 10cm + one 9cm fan are full of force, which is very strong, heavy and scary.
Compare the RTX 3080 Ti FE graphics card, which is completely large in the circle. Other graphics cards are really brothers in front of the RTX 3090 Ti Xingyao OC graphics card.
Of course, it is not only big, but also really heavy. The weight of the RTX 3090 Ti Xingyao OC naked card reaches 2.1KG, and the RTX 3080 Ti FE is only 1.37KG. It is really a noble high-end brick that doesn’t like anyone.
The back plate of the graphics card is a white star back plate, made of high-strength aluminum alloy material. In addition to the PCB and heat dissipation, the entire graphics card can be equipped with a white host. The white back panel can also provide players with better DIY space, and the stickers and other stickers are also more personalized.
The rear of the graphics card is a porous heat dissipation design, which reduces the rebound effect of the wind pressure of the radiator and strengthens the heat dissipation efficiency.
NVLink interface
12VHPWR auxiliary power supply interface
can be used to power the gifted Micro-Fit 12Pin To 3*8Pin conversion cable
The output interface is three DP 1.4a and one HDMI2.1, all of which support a single interface to implement 8k video output , and four interfaces can also be used to form a four-screen use.
Yingchi RTX 3090 Ti Xingyao OC disassembly
Yingchi RTX 3090 Ti Xingyao OC graphics card PCB adopts a non-public design and is modified based on the public version of the PCB solution. This PCB solution is different from the general GA102 core solution. It has a more powerful power supply and is quite neat workmanship. It always has a feeling similar to the industry graphics card PCB solution.
The design of the overshoulder PCB makes the graphics card look larger, and the PCB that covers the components is really pleasing to the eye; at the same time, there is no longer a video memory position on the back of the PCB, and the tail has also been hollowed out. The only disadvantage is that the power supply position and the SP-CAP air welding position are more. It is estimated that this PCB solution will have a higher frequency RTX 3090 Ti, and these air welding positions should be completed at that time.
RTX 3090 Ti core number: GA102-350-A1, the production cycle is 21 years and 46 weeks, and there are 12 GDDR6X video memory particles in the four weeks.
The power supply module adopts the design of two power supply modules in the front and rear, with a total of 20+4 phases super digital power supply. The entire PCB uses SMD materials, the welding process is mature, the electrical performance is stronger, and the stability is better.
There are a large number of SP-CAPs for filtering processing on the front and back ends of the power supply, replacing the conventional solid-state capacitor . No solid-state capacitor can be found on the entire graphics card.
Each phase of power supply is equipped with DrMos with high efficiency and low temperature, model MP86957, produced by MPS American core source system , with a maximum continuous output current of 70A. With the 20+4 phase*70A design, do you think of how powerful the power supply module of this graphics card is?
The main control PWM control chip is also from the MPS American chip source system. The model is the latest MP2891, and the single-controlled 20-phase video memory is powered by power. The specifications are believed to be much stronger than the common MP2888A. The two us5650Qs in front and behind the PCB are 4-phase video memory powered controllers.
The 12VHPWR interface has significantly better workmanship and is more stable than the general 6Pin/8Pin interface; the rear of the graphics card is the voltage monitoring point of the graphics card, so players or engineers can quickly monitor the graphics card voltage, which is convenient for everyone to overclock or repair.
The radiator is the Xingzhuo II Plus heat dissipation module, which is quite large and heavy. Most of the weight of the entire graphics card comes from the radiator. Xingzhuo II Plus heat dissipation uses a large three-fan radiator with 6+2 heat pipes, two-stage heat dissipation fins, type.
GPU base is a nickel-plated copper base that can quickly absorb and transfer GPU heat. GPU, video memory, power supply module and other different parts adopt an integrated covering heat dissipation design, and are equipped with a heat dissipation pad with high thermal conductivity and viscosity to assist in power supply.
This angle allows you to clearly see the direction of the heat pipe of the Xingzhu II Plus heat dissipation body. Two 8mm+4 6mm thick heat pipes penetrate the two different heat dissipation fins, and the remaining 2 are U-shaped designs.
The workmanship of the radiator is still possible, the fin buckles Fin + two fixed rods to enhance the rigidity of the entire radiator.
At the same time, the heat pipes are fixed with the heat sink by reflow soldering, and the heat dissipation efficiency is well guaranteed.
This is the end of the disassembly. It is worth noting that the graphics card comes with a three-slot bezel standard. If the position of the bezel is limited by the chassis slot, or if the user who replaces the water-cooled dual-slot graphics card, he or she can replace the given double-slot bezel by himself.
Test platform introduction
Test platform introduction
We selected the top batch of hardware composition test machines in the consumer market to test the Yingchi GeForce RTX 3090 Ti Xingyao OC graphics card. The I9-12900K is paired with ROG MAXIMUS Z690 EXTREME motherboard + Zhiqi Trident Z5 RGB Fantasy Finger DDR5 + Dragon God II 360 integrated water-cooled radiator, which can not only meet the needs of gamers, but also the rich expansion performance of the motherboard can also meet the needs of content creators and micro computing workstations.
As for the display, since we do not have an 8K monitor and TV at hand, in this test, we used DSR technology to zoom K resolution for game testing, and at the same time, we will test the graphics card through decoding, rendering, and exporting of 8K videos.
As for power supply, we chose Xingu 's latest 80PLUS gold medal product GM1250W full module power supply. Although the 850W power supply product can meet the power supply requirements of the non-public graphics card of the RTX 3090 Ti, in order to avoid any accidents during the test, it is safer to choose a high-power 1250W power supply; the GM1250W is equipped with the native module cable of Micro-Fit 12Pin, but during the test, we prioritized the use of the official Micro-Fit 12Pin conversion cable to ensure the smooth progress of the test.
Theory & Game Performance
DMARK Theoretical Performance Comparison
From the theoretical performance of 3DMARK, NVIDIA is really a magical swordsmanship. A GA102 core is divided into five products, and each product has a performance gap of 5~7%. Compared with the RTX 3090 graphics card, the RTX 3090 Ti adds 2 sets of SM and higher frequency settings, giving it better performance. If you compare the RTX 3080 graphics card, the 24% performance improvement is still really good, provided that the price and power consumption are not considered.
It is worth noting that our comparison table just now does not include NVIDIA DLSS functional test, DLSS2 Performance 8K project, the RTX 3080 Ti with 12GB video memory capacity, directly explodes video memory in this test item, and its performance is naturally not representative. On the other hand, it has also proved that if you want to enable DLSS to play games at 8K resolution, it is not possible to do without a graphics card of RTX 3090 Ti or RTX 3090.
Game performance comparison:
Game test
Before testing, the author actually did not have much hope for the performance of the RTX 3090 Ti. After all, the performance of the RTX 3090 graphics card, which is also the core of GA102, is already quite good. Even if the full specification GA102 is coupled with high-frequency performance improvement, it is limited. But what I didn't expect was that the 450W TDP graphics card should not be underestimated. The performance of 2K alone is already about 10% ahead of the RTX 3080 Ti; and with the resolution reaching 4K, the leading range has increased to about 15%. This performance gap really satisfies me.
Of course, if you look carefully at the comprehensive proportion of game performance, you will find that the performance gap between the five graphics cards such as this is basically the same as the theoretical performance gap. I really suspect that Lao Huang used a ruler to cut the core, and he could achieve such accurate performance positioning.
Ray Tracing Game Test
We also conducted ray tracing performance tests for four games with 2K and 4K resolutions. DLSS balance mode is turned on by default during the test. In fact, the performance of the five graphics cards in ray tracing games is consistent with that in regular games. As the big brother, the RTX 3090 Ti is still the first.
When you look at the two games "Tomb Raider: Shadow" and " Watchdog Legion ", the game frame rate performance is still quite smooth, so we can directly turn on ray tracing at 4K resolution to play the game.
Let’s go to the 8K game test again. It can be said that RTX 3090 Ti is OK to play 8K games, but the high game picture quality will cause too low frame rate. Therefore, the author recommends that everyone pay attention to the picture quality settings when playing 8K games, which can appropriately sacrifice the picture quality to achieve higher game fluency.
PS: Blockade 2 in the entire country, extremely high picture quality, DSR 8K resolution when playing games
Productivity test
The creator performance test, the performance gap between RTX 3090 Ti and RTX 3090 graphics cards is about 1%. Some applications will indeed lead a little, but it is only a few seconds away. If the time cost allows, you can actually choose a graphics card with better performance and power consumption ratio.
It has to be reminded that the RTX 3090 Ti is currently the only consumer-grade graphics card with ECC technology. When rendering large-scale engineering files, the author recommends that everyone enable ECC technology to obtain more stable rendering performance.
PS: After the ECC function is enabled, the graphics card performance will decline
Because the RTX 3090 Ti cannot widen the gap in the comprehensive creator performance performance, the author directly pulled out each item in the PugetBench DaVinci Benchmark Extended test to compare. The Extended test adds 8K video tests compared to the default tests, so that everyone can see the performance gap more clearly.
If you are the main UP host of the main video creation, I believe that the RTX 3090 Ti will bring you faster rendering speed. If the budget is limited, I actually recommend purchasing an RTX 3080 graphics card, which can meet the needs of most video and creative applications.
Power consumption and temperature test
The power consumption test really makes the RTX 3090 Ti reveal its 450W TDP nature. The simple CPD power consumption has reached 326.4W, which is higher than the power consumption of the entire RTX 3080 FE card. If you look closely, you will find that the power supply of the Micro-Fit 12Pin interface is quite powerful, and it directly provides 409.7W power supply, and PCIe is only 36.7W; according to the maximum TDP value of this RTX 3090 Ti Xingyao OC graphics card of Yingchi , there is still room for frequency increase.
At the same time, I have to say that the cooling performance of the RTX 3090 Ti Xingyao OC graphics card is really good. The indoor room temperature is 25, and the temperature of the full-load core of Furmark is only 70.6 degrees. At this time, the frequency of graphics card can still be maintained at 1440MHz, the fan speed is only 52%, and the noise is not obvious.
Of course, the most powerful thing is the video memory temperature. The 512Mb*32Bit specification GDDR6X video memory particles are good. 12 pieces on a single side can meet the 24GB video memory capacity, so you don’t have to worry about overheating. The RTX 3090 Ti Xingyao OC graphics card has a 80-degree video memory temperature, while the RTX 3090 graphics card with double-sided video memory temperature reaches 90 degrees, which is 10 degrees lower. How to use up
4GB video memory?
How to use 4GB video memory?
When testing the RTX 3090 Ti graphics card, the author used HWINFO software to record the video memory usage during the test. It can be found that most applications and games have a capacity of more than 10GB of video memory, and only some games have a capacity of more than 10GB of video memory, such as "Local Blockade 2" and "PugetBench DaVinci Test". Of course, the video memory usage of 8K games is basically overturned, and the video memory usage in the game is over 10GB, among which the video memory is even more explosive in "The World Blockade 2".
Of course, in addition to games, the actual professional application scenario is where the RTX 3090 Ti, the 24GB video memory capacity is used. Next, we will use three small examples of video, computing, and modeling to "squeeze" the 24GB video memory of this RTX 3090 Ti.
DaVinci Resolve 8K real-time rendering
Under normal circumstances, DaVinci does not call too much video memory. Whether it is playback or transcoding, the RTX 3090 Ti only takes about 10GB during testing.
And we use the 8K RAW Video test project file provided by NVIDIA to make a series of settings for Motion Effects, so that we can call 24GB of video memory.
Same time, the video memory usage has reached 24GB when playing 8K RAW Video, which is really exaggerated. Of course, if you replace other low-capacity versions of graphics cards at this time, you will directly report an error and stop running. So after setting this way, is the RTX 3090 Ti an advantage? We also use this project file to directly export the video to compare the export time.
You can see the same Motion Effects setting. In general, RTX 3090 Ti, RTX 3090 and RTX 3080 have the same export speed, and the same video memory occupancy is 10GB. However, after strengthening the Motion Effects setting, the RTX 3080 Ti obviously reports an error and cannot load the motion effect. Only RTX 3090 Ti and RTX 3090 have export results, which are 1:34 and 1:39 respectively. Side description: 4GB large-capacity video plays a key role in some large and extremely complex video rendering projects.
Machine learning operation
In the practical application of machine learning, we use an open source CNN image data classification model for benchmarking. It uses the widely used machine learning library on the market - TensorFlow. After building the CUDA component environment, TensorFlow can use the GPU to accelerate computing. Large video memory can provide more computing resource support for graphics cards. Here we set the video memory footprint to "all", and then use 2 sets of batches and 32 samples as units. , run 100,000 sets of image data recognition, you can see that the 24GB video presence is directly consumed during the calculation process, and finally achieves the performance of identifying nearly 30,000 images in one second.
BLENDERScene Rendering
And on the popular 3D modeling software blender, the built-in Cycles renderer will use real-time RTX ray calculation and real-time loading of OpitX AI denoising technology during interactive rendering, which has high requirements for video memory. For heavy users, even with a 12GB video memory capacity, there will be insufficient video memory or rendering errors or failure to render under multiple workflows. RTX 3090 Ti's large video memory can realize real-time viewing and interaction of multiple complex workflows, and can also render the scene of the final frame. At this time, the video memory usage also reaches more than 21GB.
Summary: The belated new emperor
Summary: The belated new emperor
really strong performance
450W TDP-designed RTX 3090 Ti is destined to have extremely strong performance. From a series of tests, it can verify all this. It is basically leading the other four graphics cards that are also the core of GA102 with an absolute advantage. The complete GA102 core meets the needs of ashes players well, and full blood is true love!
Of course, 450W TDP has brought us the trouble of power supply. People who can afford the RTX 3090 Ti graphics card naturally have high power supply, but the 12VHPWR auxiliary power supply interface is destined to use a conversion cable, or customize a compatible Micro-Fit 12Pin module cable, which is a lot of investment.
But thankfully, the power supply and heat dissipation of non-public graphics cards are quite powerful. For example, the Yingchi RTX 3090 Ti Xingyao OC used in our test has a strong power supply design of 20+4 phases and an efficient multi-heat pipe three-fan radiator, which can ensure that the RTX 3090 Ti runs at high frequency for a long time.
Who will buy ?
Do you think NVIDIA GEFORCE RTX 3090 Ti is worth buying? It's hard to say this. Love it so much. The RTX 3090 Ti is quite worth it. It has a complete GA102, high core frequency, 24GB large capacity and high specification graphics card, and stronger power supply. Such a powerful graphics card can do whatever it does. If it doesn't work, then buy another one and connect it with NVLink. The price of RTX A6000 is less than half, so it can be bought. It would be strange if players with coercion do not suffer from death.
I hate you naturally. It is undoubtedly a RTX 3090 Ti has absolute performance strength, but it is not cost-effective. It is much more expensive than RTX 3090/3080 Ti, but the performance improvement is limited, especially the performance power consumption ratio is a bit worse. If the power supply is not strong, you have to replace the power supply, which is too much investment.
So, buying an RTX 3090 Ti is really within your ability, and it is the bottom line to see if you need to use such a top-notch graphics card.
How to buy ?
Users who want to buy can relax. It is understood that NVIDIA provides AICs with a large number of RTX 3090 Ti full-spec GA102 core and GDDR6X video memory particles. You only need to ensure that your wallet is thick enough to buy an RTX 3090 Ti graphics card!
