NVIDIA RTX 4070 SUPER graphics card first test AIGC generation speed increased by 38%

At CES on January 8 this year, NVIDIA released RTX 4070 SUPER, RTX 4070 Ti SUPER and RTX 4080 SUPER Graphics card 。 Today, let's introduce NVIDIA GeForce RTX 4070 SUPER Video card evaluation.

NVIDIA GeForce RTX 4070 SUPER's domestic recommended price is from 4899 yuan, which is 100 yuan more expensive than RTX 4070. However, the official suggested selling price of these two graphics cards is $599, and the difference of 100 yuan comes from the exchange rate.

When it comes to the SUPER series graphics cards, the latest is the RTX 20 series. The performance of both SUPER series has been greatly improved. Therefore, this release of SUPER series can also make the overall RTX 40 series more cost-effective.

This time, the ban on the MSRP version of RTX 4070 SUPER was lifted at 22:00 on January 16, while the ban on the OC version of graphics cards was lifted at 22:00 on January 17. Please also pay attention to the follow-up evaluation.

Nvidia GeForce RTX 40 Super series is now on the market. Please purchase GeForce RTX 4070 Super/80 Super public version and partner graphics cards at JD DIY Super New Year Festival

one Overview of NVIDIA GeForce RTX 4070 SUPER FE

First of all, let's look at the appearance. The NVIDIA GeForce RTX 4070 SUPER graphics card is still packaged in a gift box with a black lid, but the font and logo are all black. The font can only be seen in different light.

After opening, the decorative pattern around the graphics card has its own "shock" effect. The card body is semi embedded in the packaging box, which is more convenient for collection and display. What's more, unlike the earliest RTX 40 Series FE version, the metal frame and font of this RTX 4070 SUPER are also black, which makes it more mysterious.

This RTX 4070 SUPER package comes with a 16 pin to dual 8 pin adapter cable.

The overall size of NVIDIA GeForce RTX 4070 SUPER graphics card is about 267 × 112 × 42mm (excluding the baffle), occupying 2 slots.

The overall appearance of NVIDIA GeForce RTX 4070 SUPER graphics card has not changed much, but it can be seen from the comparison of the above two pictures that the FE version is mainly different in the metal frame and font.

The size of the NVIDIA GeForce RTX 4070 SUPER fan is larger than that of the RTX 30 series, and has basically reached the diameter of the overall frame of the graphics card. On the basis of the increase of the cooling fan, the maximum airflow dynamic increases by 20%, and the airflow dynamic increases by 15% at the same noise level.

NVIDIA GeForce RTX 4070 SUPER also adopts the design of double fans with front and back bright faces. The back is full of exposed fins, and the words "RTX 4070 SUPER" can also be seen on the metal frame.

Although the overall design has not changed, the black metal frame used in the SUPER series is still as exquisite and beautiful as art.

On the video output interface, the four interface design of HDMI 2.1+DP 1.4a * 3 is still adopted.

In addition, because the public version uses a dual axial flow cooling system, a large number of cooling fins can be seen at the video output interface, which is the same as that of the previous generation.

The whole card power consumption of NVIDIA GeForce RTX 4070 SUPER FE is 220W, and a single 16pin auxiliary power supply is adopted. The recommended power supply is 650W and above.

two Analysis of NVIDIA GeForce RTX 4070 SUPER Architecture

The GeForce RTX 40 series graphics card released this time is built by the new NVIDIA Ada Lovelace architecture, and adopts the TSMC 4N NVIDIA customized process. The flagship core AD102 has achieved 76 billion individual transistor And 28 billion RTX 30 series graphics cards.

Compared with the previous generation NVIDIA Ampere, NVIDIA Ada Lovelace has More than twice Performance improvement up to 90-TFLOPS The shader data throughput of.

The RTX 4070 SUPER released this time has 7168 CUDA cores, providing 35.5-TFLOPS Computing power; 56 third-generation Ada RT cores have 82.1 RT-TFLOPS ； 224 fourth generation Tensor Cores available 568 Tensor-TFLOPS（AI TOPS） 。

The L2 Cache capacity of RTX 4070 SUPER is 48MB, and the bandwidth is 504 GB/s. NVIDIA officials also showed the importance of L2 Cache. We use two figures to briefly illustrate the role of L2 Cache.

In a real GPU, the kernel is where all computing takes place, and This is the role of L1 data cache. Each SM has an ultra-low latency L1 data cache, next to its processing kernel, making L1 the first choice for GPU to find information. However, since L1 cache needs to be very close to the kernel, it cannot be very large.

If the data required for kernel computing cannot be found in the L1 cache, the GPU will look in the L2 data cache. This video memory system is located on the GPU chip and connected to all GPCs (graphics processing clusters) through a very high-speed crossbeam system. Each GPC includes multiple SMs. If the information is found in the L2 cache, the GPU will pick up the data and put it into the kernel.

If no information is found in L2 cache (called cache missing), GPU will search VRAM through video memory interface. This creates a lot of extra work in the entire GPU storage subsystem, and reduces performance and power efficiency.

For more detailed parameters, we have summarized a table for you. It can be seen that in terms of specifications, the RTX 4070 SUPER is closer to the RTX 4070 Ti, but its power consumption is only 20W higher than the RTX 4070.

In addition, the main price is only 100 yuan more expensive than the recommended price of the RTX 4070. Let's take a look at the test results.

three Introduction to test platform

First, let's introduce the test platform. In order to ensure the performance of NVIDIA GeForce RTX 4070 SUPER, our platform has also been comprehensively updated.

Although the recommended power supply for RTX 4070 SUPER is only 650W, it is recommended to reserve 20% - 30% redundancy for the video card and CPU to be fully loaded and to facilitate subsequent hardware upgrades. Xingu Kunlun MU-850G power supply is made of ATX3.0 specification, with 850W rated power and 80PLUS gold certification. It is equipped with a native 12VHPWR graphics card power supply interface, which is safe and stable without switching. At the same time, this power supply uses high-quality all Japanese capacitors and LLC Pro+DC-DC pure blood architecture, which brings strong dynamic overload performance and prevents high-power PCs from being overloaded, causing blue screen, restart and other situations.

With high-performance CPU and graphics card, the heat dissipation of the chassis should also keep pace, while taking into account the beauty and durability. Xingu Boundless 1 case, with 4mm high permeable tempered glass on the front and side, creates a real "sea view room" effect at 270 ° wide angle, which can not only monitor the operation of hardware in real time, but also provide first-class "light pollution" effect with RGB light effect accessories; In terms of heat dissipation, it is equipped with 9 fan slots and supports 240/360mm water-cooled radiators. At the same time, it is also equipped with a hand-operated special booth to make your chassis enjoyable every day. It is worth mentioning that this chassis uses a コ type mechanical support structure with high-strength hardware accessories, which brings strong load-bearing performance, and it is stable without columns.

Current GPU-Z The 2.56 version can recognize some basic specifications of the graphics card, but the chip itself has not yet recognized them.

NVIDIA GeForce RTX 4070 SUPER has 7168 CUDAs, the Boost frequency is 2475MHz, which is the same as the RTX 4070, but the basic frequency is slightly higher.

12GB GDDR6X video memory is adopted, with a bit width of 192bit, a video memory bandwidth of 504 GB/s, and 80 and 224 grating units and texture units.

four Theoretical performance test

In this test, NVIDIA GeForce RTX 4070 SUPER mainly compares its upper and lower models, RTX 4070 Ti, RTX 4070 and the previous generation RTX 3070.

The first step is to measure the graphics card DX11 The 3DMRKFS package with theoretical performance: FS, FSE and FSU correspond to the theoretical performance of the graphics card at 1080P, 2K and 4K respectively. The actual test results of the graphics card scores are as follows:

In the FS test, NVIDIA GeForce RTX 4070 SUPER is compared with RTX 4070, where FS is improved 18% ； FSE lifting 19% ； FUS promotion 25% , comprehensive improvement 21% 。 Compared with RTX 4070 Ti, the comprehensive scores are different 3% about.

In the Time Spy and Time Spy Extreme tests for the DX12 environment, the improvement of NVIDIA GeForce RTX 4070 SUPER over RTX 4070 is: TS improvement 18% ； TSE lifting 18% , about 18%。 Compared with RTX 4070 Ti, the comprehensive scores are different 5% about.

PortRoyal is 3DMARK For the test items of the secondary specialized school for the performance of light pursuit, NVIDIA GeForce RTX 4070 SUPER has an improvement of about 17% 。 Compared with RTX 4070 Ti, the difference is 4% about.

Now let's take a look at the tests of some new specific application scenarios in 3DMARK.

Speed Way test combines real-time ray tracing and traditional rendering technology to measure the performance of video card. The scene contains raytraced reflections, real-time global lighting, mesh shaders, volume illumination, particles, and post-processing effects. So SW testing can basically be regarded as the benchmark of next-generation 3A games.

NVIDIA GeForce RTX 4070 SUPER, compared with RTX 4070, increases from 1080p resolution to 4K resolution as follows: 18%/17%/17% 。

Compared with the RTX 4070 Ti, the performance is almost the same, the difference is not more than 1%, and even exceeds the 1080p resolution. Although the gap is small and can be regarded as error, on the whole, the performance of this RTX 4070 SUPER is basically comparable to that of RTX 4070 Ti.

In addition, we used the DLSS 3 just updated by 3DMARK for relevant performance tests.

NVIDIA GeForce RTX 4070 SUPER Under 2K resolution, the DLSS switch is improved to 15%/20% ； 4K resolution, increased by about 16%/22% 。

From the overall theoretical test results, NVIDIA GeForce RTX 4070 SUPER has a comprehensive improvement of about 19% about; Compared with RTX 4070 Ti 3% The latest SW comprehensive test also shows the same results.

Compared with the RTX 3070, the test of traditional FS/TS/PR and next generation scenario SW is comprehensively improved in 50-55% about.

Although our test platform has been updated from 13900K to 14900K, everyone who has seen the CPU test must know that the 14 generation Core is pulling its legs. In addition, our theoretical test only selects the score of the graphics card itself, and the physical score is not included.

five The game test of "Swallow Cloud Sixteen Voices"

This time we also tested NetEase's latest "Yanyun Sixteen Voices", but the game has not yet been publicly tested, and all test data are for reference only.

The Sixteen Voices of the Swallow Cloud is an open world with unique Chinese martial arts feelings RPG Game , it will be divided into single person mode and multi person mode, and published on the whole platform.

The game provides a rich face pinching system, but there are no female characters in the internal testing stage.

In addition, a large number of Chinese style elements are integrated into the game, so players can unlock the corresponding historical documents according to the elements found in the game, so as to truly teach and enjoy.

Moreover, in addition to the archaic elements in the game environment, the BOSS settings are very sophisticated. The shadow puppeteers outside the novice village can be called the "big tree guards" of our domestic games, and they failed after a few simple attempts.

What is amazing is the brilliant picture of "Sixteen Voices of the Swallow Cloud". The photos taken at different times in the same place are not only simple changes in ambient light, but also the rendering of various atmospheres will vary with time.

At present, "Yanyun Sixteen Voices" has supported DLSS 3 technology in the internal test version. Let's take a look at the actual frame number test using NVIDIA GeForce RTX 4070 SUPER.

It is easy to see from the table that DLSS 3 has significantly improved the low resolution when encountering CPU bottlenecks, with 73% improvement at 1080p resolution, and 43%/21% at 2K and 4K respectively.

In addition, in the CPU bottleneck of the game, the addition effect of DLSS 2 is relatively limited, but compared with DLSS off, it can still have a good number of frames.

In the detailed test of DLSS 3, we chose the 2K resolution of the main game scenes of RTX 4070 SUPER. Among them, DLSS 2 is 25% higher than DLSS 2; Compared with DLSS 3, DLSS 3 has increased by 93%, which is amazing.

six General game performance test

Since the new DLSS 3 technology is added to the RTX 40 series this time, a separate test will be carried out later. Here we still select several mainstream 3A masterpieces for game performance comparison.

In Extreme Racing: Horizon 5, NVIDIA GeForce RTX 4070 SUPER has significantly improved in 1080p resolution compared with RTX 4070, reaching 13%, and 2K resolution has increased by 14%; 4K resolution increased by 18%, comprehensive improvement 15% 。

On《 Assassin's Creed In Fantasy, NVIDIA GeForce RTX 4070 SUPER increases by 8% compared with RTX 4070; 2K increased by 9%; 4K increased by 12%, comprehensive improvement 10% 。

In Land Without a Master 3, NVIDIA GeForce RTX 4070 SUPER has increased by 19% compared with RTX 4070: 1080p; 2K increased by 17%; 4K resolution increased by 16%, comprehensive improvement 17% 。

Light pursuit test of bright memory: infinity Software It is a test tool independent of the game, which is more than the ray tracing technology used in the game. The test condition is "RTX highest/DLSS quality".

In terms of performance, compared with RTX 4070, NVIDIA GeForce RTX 4070 SUPER increases by 19% at 1080p; 2K increased by 22%; 4K increased by 23%, comprehensive increase 21% 。

In the score running software of another domestic game Border, the situation is basically the same as that of Bright Memory: Unlimited, and the test conditions are all conducted under the "highest RTX/DLSS quality".

In Frontier, NVIDIA GeForce RTX 4070 SUPER has increased by 17% compared with RTX 4070: 1080p; 2K increased by 18%; 4K increased by 19%, comprehensive increase 18% 。

In the final test, we will experience different light and shadow modes in Cyberpunk 2077.

From top to bottom, it is super image quality/ray tracing super/ray tracing: overspeed mode. You can see the ray tracing: the overspeed mode is closer to the real effect than the ray tracing super mode. It simulates the real ray path. In fact, compared with the previous ray tracing, it simulates the reflection of more rays on different surfaces, completely calculates the real lighting of the scene, and avoids the "dead black" situation in the previous generation of ray tracing.

In addition, Cyberpunk 2077 has updated DLSS 3.5 to support light reconstruction technology. Like the stone bridge in the above figure, when light tracing is turned off, the mottled marks on the stone bridge are almost invisible. When light tracing is turned on, although the surface state is reflected, the performance details are lost. After DLSS 3.5 is turned on, the surface texture is more detailed, which improves the overall image quality to a higher level.

seven DLSS 3 performance test

Up to now, more than 530 games and applications have supported RTX technology.

The DLSS 3 technology will be supported by Dragon Creed 2, Grey Zone War, Half Life 2 RTX, Horizon: The West's Dead End, Layers of Fear, Dragon 7 in Man: The Unknown Dragon, Dragon 8 in Man, Nake Garden: The Last Paradise, Pax Dei, Star Miner and Kingship and Freedom, Some of these games will be released this year, while others will update their support for DLSS 3 this year, bringing unprecedented performance and higher image quality to gamers equipped with GeForce RTX 40 series graphics cards.

The test chart of DLSS 3 is complicated, and 1% low FPS and delay tests are added, 1% Low FPS arranges the number of frames in a period of time from large to small, selects the smallest 1%, and then averages the 1%.

Starry Sky has experienced the downturn and bad comments after its first launch, and has received a good response recently. After the 1.8.86 version update, DLSS 3 was added, which finally enabled users holding N cards not to open FSR, and the overall effect was very good.

In 1080p resolution, DLSS 3 is 64% higher than DLSS 2; 2K resolution increased by 51%; 4K resolution increased by 39%. It is easy to see that the lower the resolution, that is, the CPU bottleneck, the greater the frame generation bonus of DLSS 3.

Although the frequent image reading of Star Sky, which has been criticized, has not been modified, at least for users who use the N card, the smoothness of the image can be greatly improved.

Horizon 5 is also an old friend of video card testing. Its excellent optimization enables many entry-level products to run at a higher frame rate. It is also evident in NVIDIA GeForce RTX 4070 SUPER. In the three resolutions, the improvement of DLSS 3 over DLSS 2 is 39%/27%/9% 。

Finally, we also tested Wizard 3: Crazy Hunt. After ray tracing was added to this game, the requirements for configuration rose sharply, almost reaching the point where the game could not be played smoothly without DLSS. Among them, NVIDIA GeForce RTX 4070 SUPER enables DLSS 3, which is 125% higher than closing DLSS; Compared with DLSS 2, it has also increased by 51%.

eight Productivity tool testing

V-Ray

In terms of productivity tools, we tested the latest version of V-Ray renderer, and selected CUDA test and RTX test using GPU rendering. The results are shown in the figure below.

Blender

Blender is a professional 3D rendering software, which has launched a fixed benchmark Score running software, eliminating installation software download The trouble of material. The software will automatically render three scenes of the test monster/junkshop/classroom.

The above figure shows the score of NVIDIA GeForce RTX 4070 SUPER graphics card, 3281/1548/1645 points respectively, with an average of 2158 points.

D5 renderer

D5 is a domestic real-time ray tracing renderer, which supports direct reading of various modeling software, including SketchUp 、 3Ds MAX , Rhino, Revit, Archicad, Blender, C4D, etc. At present, DLSS 3 is also supported. Let's take a look at the frame number and power consumption respectively.

Select the default scenario of the software [Simple Living Room]

In terms of real-time frame number, as D5 uses real-time GPU rendering, the pressure is all on the graphics card. It can be seen that at 4K resolution, even NVIDIA GeForce RTX 4070 SUPER is in PPT status. After DLSS 3 is enabled, the frame number can be increased to 470% Around, very violent.

In terms of power consumption, the TDP of NVIDIA GeForce RTX 4070 SUPER is 220W. When DLSS is turned off, the actual average power consumption is 205W, while when DLSS 3 is turned on, the power consumption is only 173W. With lower power consumption and higher frames, it can be said that DLSS 3 is not only very helpful for game players, but also an effect of 1+1>2 for content creators.

nine AIGC test

I don't know when the AI era has become closer to people's lives, ranging from text rewriting to various generative AI. We have tested the AI drawing ability of Stable Diffusion before, but now NVIDIA has launched a new TensorRT plug-in, which can fully call Tensor computing power. Let's test it NVIDIA GeForce RTX 4070 SUPER's performance in this respect.

[Note: xformers optimization scheme is enabled by default for all tests]

The first is the relatively basic SD1.5 model test. In SD1.5, the output resolution of the model is 512 × 512 and 768 × 768. We will export the [TRT] engine and record the default model rendering and the batch time of [TRT] model rendering.

Output according to the unified parameters provided by NVIDIA, so as to record the results. The sampling method is Euler a; Sampling steps: 50; Batch Count：10

[Default SD1.5 model] RTX 4070 SUPER operation time 41.3s, about 4.13s

[TRT SD1.5 model] RTX 4070 SUPER operation time is 31.5s, which is about 3.15s

The second test is the SDXL model test of. The SDXL model will be trained at a resolution of 1024 × 1024. Compared with SD1.5, the image is twice larger. It can provide more accurate keyword matching, thus obtaining more realistic results. However, due to the higher original resolution and model complexity, the computational strength of the SDXL model is significantly higher.

We set according to the same parameters, but the size is specified as 1024 × 1024. Let's see How does NVIDIA GeForce RTX 4070 SUPER perform.

[Default SDXL model] RTX 4070 SUPER operation time is 2m49.7s, which is about 16.97 seconds

[TRT SDXL model] RTX 4070 SUPER operation time 1m47.7s, about 10.77s

In more complex models, the advantages of TensorRT engine are increasingly obvious, and the output speed of a single image can be increased by about 6 seconds. And the utilization of video memory has reached 100%, reaching 12GB.

Generative AI is getting closer and closer to our life. Content creators are the most helpful. AI's unconstrained thinking is often more expansive than people's, and AI can inspire more sources of inspiration. For ordinary users, he can also let people without painting skills independently create a work no less than that of an artist, so that everyone can become an artist in their own life.

At present, generative AI is only in the initial stage of development. Later, for personalized education, health care, language processing, even financial analysis and other scenarios, I believe that AI can be used.

ten RTX Video Super Resolution Test

At present, the RTX Video Super Resolution has been partially browser First, players need to update to the latest NVIDIA drive In the NVIDIA control panel, you can see the latest RTX video enhancement superresolution in [Adjust Video Image Settings].

RTX video super-resolution is a breakthrough in AI image processing. It goes beyond the traditional edge detection and feature sharpening technology and greatly improves the quality of live video content.

Enabling RTX video super-resolution requires not only the latest driver, but also the use of RTX 40 or 30 series GPU, and is almost suitable for Google Chrome And Microsoft Edge browser (the browser also needs to be updated to the latest version).

After opening, currently known users can enjoy the bonus of RTX VSR effect by opening YouTube or Station B.

If you are not sure, you can open the task manager when playing the video in full screen. When you see the GPU load increases, it means that the startup is successful.

We open YouTube to watch videos at will. After opening RTX VSR, we can clearly see that the quality of underwater coral has improved significantly, the edges are clearer, and the distortion is greatly reduced.

(Click to enlarge the original picture)

In the comparison of the following picture, it is obvious that after the RTX video super-resolution is turned on, the beard is obviously smoother, and the mosaic is no longer visible.

However, it should be noted that the effect of RTX video super-resolution is more obvious at low resolution. If the video you watch is 1080p or above 2K, it is difficult to see the difference.

This is very helpful for some users whose network speed is not fast enough, who can only watch low resolution videos, or the video source itself is not clear enough.

eleven Temperature and power consumption test

In the power consumption test, we selected FurMark software for copying test, and used GPU-Z to detect the temperature. The power consumption only calculated the graphics card itself.

You can see NVIDIA GeForce RTX 4070 SUPER This video card has passed the copy machine test for about 40 minutes. The temperature is about 71 ℃, and the hot spot temperature is about 83 ℃.

Game dynamic power consumption test

In this copy test, the maximum on-board power consumption is about 220W. But it is much lower in the actual game test.

Due to different game loads, the actual power consumption of GPUs is dynamic. Similar to the change of FPS over time, it is difficult for the RTX 40 series to reach the power consumption wall.

NVIDIA GeForce RTX 4070 SUPER Average power consumption of 3A games

Average power consumption of RTX 3070 Ti 3A games

In the actual game power consumption test, we chose Bright Memory: Unlimited, with its own benchmark, the highest light pursuit, DLSS quality, and 2K resolution, to pull the performance limit of the video card and test the power consumption of our actual application scenarios.

You can see NVIDIA GeForce RTX 4070 SUPER The average power consumption is only 197W RTX 3070 Ti even reaches 288W ， 90W lower This is indeed an amazing achievement.

The average power consumption of the game is less than 200W. For users who use the computer for a long time or don't like to shut down the computer like me, it can effectively reduce the electricity charge, especially in the south, where the electricity charge is extremely high.

Or in some relatively extreme use environments, such as Internet cafes, computers are basically fully loaded for 24 hours. The ultra-low power consumption of the RTX 40 series is really good news for Internet cafe owners.

twelve Performance approaching RTX 4070 Ti

After testing the whole article, I believe you have roughly understood this picture Although the performance of NVIDIA GeForce RTX 4070 SUPER is only 100 yuan more expensive than that of RTX 4070, its performance is greatly improved.

To be specific, the comprehensive performance of this RTX 4070 SUPER is improved by 15-20% compared with the RTX 4070, and the gap with the RTX 4070 Ti is about 5%.

As a 70 level product, although it is positioned as a 2K resolution high frame light chase game, whether it is a daily online game, independent game, or a traditional 3A game, it can reach the level of fluency or even electronic competition at 4K resolution, and the price is still within 5000 yuan. This SUPER series is really gratifying.

If we put aside the concept of 70 and 80, it can even compete with the price performance king of the RTX 30 series RTX 3080 for the title of "price performance king".

Previously, we said that the overall performance improvement of the RTX 4090 series was quite explosive in terms of the performance of the RTX 4090 series. However, the performance of the RTX 40 series was severely out of gear from the entire 70 level, which also led to the difficulty of opening the gap between the 60 level product and the RTX 30 series, which was criticized.

This upgrade of the SUPER series can be said to pull the overall performance level of the RTX 40 series back to the normal range, forming a more regular isochromatic series.

However, from the actual performance of the game, this card can achieve the E-sports frame rate at 2K resolution, and also has excellent performance at 4K resolution, of which Horizon 5 is 110 frames; The Assassin's Creed: Illusions is 79 frames; The Land Without a Master is 67 frames. When DLSS 3 is turned on, Starry Sky can reach 85 frames; The Sixteen Voices of Yanyun can reach 105 frames.

So although this is still a level 70 graphics card, you can feel the 4K fluency standard, and the game picture is improving every year. If you use the old standard to measure this one NVIDIA GeForce RTX 4070 SUPER， It's really a bit expensive. If you take it as the RTX 3080 of the previous generation, it will be much cheaper.

This RTX 4070 SUPER is equipped with 12GB video memory, so in addition to being a game card, it also has good performance in content creation and AI performance. Especially with the TensroRT plug-in, the single picture of the SDXL model is about 38% faster than Mr. Yuan. The speed increase of batch processing multiple sheets is calculated in minutes.

Of course, pictures are only the basic play method of generative AI. I believe that in the future, no matter voice, video or complex scenes, games can not be separated from generative AI, so a video card with high investment and high return is very cost-effective.

In addition, when it comes to price increases, in recent years, it should be more outrageous to play games. Some standard cans 3A have doubled or even tripled in price, which makes people can't afford to play.

The above is the entire content of this NVIDIA GeForce RTX 4070 SUPER. What do you think of the upgrade of SUPER?

This article is an original article. If it is reproduced, please indicate the source: NVIDIA RTX 4070 SUPER graphics card first test AIGC generation speed increased by 38% https://diy.zol.com.cn/851/8518342.html