NVIDIA officially released a new RTX 40 SUPER series GPU on CES 2024. The RTX 4070 Ti SUPER is the only model in the NVIDIA GPU that has both Ti and SUPER naming suffixes, and it is also the only member of the RTX 4070 series that has 16GB super large video memory. It is officially positioned to play 2K ray tracing games at high frame rate under extreme image quality. Can it become the new 2K game overlord?
↓↓↓ Click to learn about RTX 4070 Ti SUPER sculpture ↓↓↓
Same as "RTX 4080", the specification has been upgraded
Although RTX 4070 Ti SUPER also belongs to the RTX 4070 series, it actually uses the same AD103 core as the RTX 4080, and its physical scale is larger than the AD104 used by the RTX 4070 Ti, which also means that it has the specifications of the "80" series, which is a cross level upgrade compared with the RTX 4070 Ti.
Of course, in order to accurately divide the product positioning, RTX 4070 Ti SUPER still has a gap with RTX 4080, with the overall scale reduced by about 13%. Compared with the RTX 4070 Ti, the specification upgrade of the RTX 4070 Ti SUPER is about 10%, but the video memory bit width is increased from 192bit to 256bit, and the capacity is increased to 16GB. Therefore, it will perform better in games and applications that eat video memory bandwidth and capacity.
In terms of power, the average game power of the RTX 4070 Ti SUPER is about 226W, and the power of the whole board is 285W, which is consistent with the RTX 4070 Ti and lower than the RTX 3070 Ti. It can be seen that its energy efficiency ratio performance is also quite excellent.
In terms of price, the initial price of RTX 4070 Ti SUPER is the same as that of RTX 4070 Ti, which means that the price is increased without increasing the quantity, and the cost performance is improved sharply.
For old users who are still using RTX 3070 Ti, the upgrade range of RTX 4070 Ti SUPER is a leap. Not only are RT Cores and Tensor Cores all upgraded, but also the frame generation function of DLSS 3 is supported. While the performance increases dramatically, the power even decreases. Old players do not need to change the power supply to upgrade.
When the RTX 3070 Ti is a little powerless in the face of the latest light chase game masterpiece and needs to reduce the image quality to maintain fluency, the RTX 4070 Ti SUPER is undoubtedly an upgrade option that can provide a better experience.
It is worth mentioning that RTX 4070 Ti SUPER will not launch Founders Edition, so we used AORUS RTX 4070 Ti SUPER 16G from Gigabyte in this test.
Gigabyte AORUS RTX 4070 Ti SUPER 16G
The overall shape of RTX 4070 Ti SUPER Super Sculpture keeps the family design style of RTX 4070 Series Super Sculpture. It is equipped with a wind power cooling system, and has three bionic shark double ball bearing fans, which support forward and reverse, 3D start and stop technology, effectively increasing heat dissipation efficiency and reducing noise.
The radiator part adopts the design of soaking plate direct contact GPU, and has 9 composite heat pipes. The fin of the radiator adopts the bent angle design, and is equipped with air inlet grille, which effectively improves the heat dissipation efficiency of the whole machine.
A full metal backplane is installed on the back of the graphics card, and a custom graphics card bracket is included with the card, which effectively enhances the ability of the graphics card to resist physical damage, and is safer and more reliable to install in a vertical chassis.
In terms of personalization, the fan of RTX 4070 Ti SUPER Super Sculpture is equipped with dazzling light wheel RGB light, and the top of the graphics card is also equipped with a super large AORUS LOGO light, which supports RGB FUSION light effect synchronization technology. In addition, the graphics card also provides a dual BIOS switch, which can quickly switch between mute mode and OC mode.
In the performance specification part, the RTX 4070 Ti SUPER Super Sculpture is equipped with 12+3 phase digital power supply. The GPU acceleration frequency is up to 2670 MHz, higher than the official 2610 MHz. The power limit can also be manually unlocked to 128%. In terms of video memory, 256bit/16GB GDDR6X can fully meet the requirements of 2K high frame light tracking games and efficient productivity.
Let's take a look at the actual performance of RTX 4070 Ti SUPER Super Sculpture.
Performance measurement: 2K takes all 4K smoothly, and AI plot performance is more powerful
Test platform
Graphics card: AORUS RTX 4070 Ti SUPER 16G
AORUS RTX 4070 Ti Super Sculpture 12G
RTX 4080 Founders Edition
Processor: Core i9 14900K
Memory: AORUS DDR5 6000 16GB × 2
Main board: Z790 AORUS MASTER
Hard disk: WD_BLACK SN850X 2TB
Power supply: Platinum carving AP1200PM
Operating system: Windows 11 Professional 23H2
Benchmark performance test
In the benchmark performance part, in the FireStrike series test of DX11, RTX 4070 Ti SUPER is about 8%~9% ahead of RTX 4070 Ti, and in the TimeSpy series test of DX12, RTX 4070 Ti SUPER is about 7%~9% ahead of RTX 4070 Ti. In the SpeedWay, DXR and Port Royal tests, the RTX 4070 Ti SUPER can lead the RTX 4070 Ti by about 16% at most, especially the DX12U project SpeedWay with heavier load. In addition, in DLSS project testing, RTX 4070 Ti SUPER has obvious advantages after DLSS is enabled, up to 13%. In a word, the benchmark performance of RTX 4070 Ti SUPER compared with RTX 4070 Ti has increased by nearly 10% on average, reaching 87% of RTX 4080 and 1.91 times of RTX 3070 Ti.
Rasterized Game Test
The performance of RTX 4070 Ti SUPER is also eye-catching in the practical part of raster games. From the test data, we can see that the RTX 4070 Ti SUPER can make all the tested game masterpieces run at an average of more than 80 fps under the 2K limit image quality setting, and many of them exceed 100 fps. The smoothness can be said to be quite OK.
Even with 4K extreme image quality, RTX 4070 Ti SUPER can keep most of the game masterpieces above 60fps. You should know that DLSS is still not turned on, and the frame rate will further increase after it is turned on.
Compared with RTX 4070 Ti, the maximum increase of RTX 4070 Ti SUPER is close to 10% under 2K limit image quality, and even higher under 4K limit image quality. The increase of some games has reached more than 10%.
Light Tracing and DLSS Test
The performance of the RTX 4070 Ti SUPER has changed again after the optical tracking and DLSS are enabled. At 2K resolution, RTX 4070 Ti SUPER can lead RTX 4070 Ti by about 8% at most and 10% at 4K resolution at most. Compared with the RTX 4080, the RTX 4070 Ti SUPER can achieve about 90% performance at 2K and 87% performance at 4K, which is also quite good.
DLSS 3/3.5 performance test
DLSS 3/3.5 is the exclusive black technology of the RTX 40 series, which can bring a huge increase in the game frame rate for the RTX 40 GPU. It can be seen from the actual measurement that after the DLSS 3/3.5 is turned on, the game frame rate of the RTX 4070 Ti SUPER has increased by 148% at the highest level, and most games have nearly doubled or even doubled, which can instantly raise the frame rate of the light chase game masterpiece below the average 60fps to more than 100fps. The effect can be said to be quite powerful, so it is recommended that players must turn on DLSS 3/3.5 support in the game.
▲ In Mind Killer 2, after light reconstruction is enabled (left), the details of the reflection on the water surface become richer and the picture becomes clearer
In addition, DLSS 3.5 also adds light reconstruction technology, which can further enhance the details of the game screen. From the screenshot of Mind Killer 2, we can see that after the light reconstruction technology is turned on, the reflection details of the ground puddles are richer and the picture is clearer, effectively improving the visual experience of players. At present, the number of games supporting DLSS 3.5 is also increasing. It is recommended that players using RTX 40 GPU must open DLSS 3.5 (including ray reconstruction) in the game.
▲ After DLSS 3.5 support is turned on in the D5 renderer, the noise of the picture almost disappears, and the whole picture is more exquisite and more realistic
DLSS 3.5 can also significantly improve efficiency and image quality in productivity software such as 3D design. For example, the D5 renderer provides support for DLSS 3.5, including ray reconstruction, super-resolution scaling, and frame generation. From the actual measurement, after DLSS 3.5 is turned on, the rendering efficiency of RTX 4070 Ti SUPER is increased by 356%, and the noise of the rendered image is reduced to almost invisible. Therefore, designers and users are strongly recommended to turn on DLSS 3.5 support to enjoy the most perfect rendering efficiency and effect.
Game Delay Test
DLSS 3 (including Reflex) can not only significantly increase the game frame rate, but also reduce the frame delay and operation delay, and improve the response speed of player control. From the PCL delay monitored by FrameView, after the DLSS 3 is enabled by the RTX 4070 Ti SUPER, the delay of most games decreases significantly, with a maximum decrease of 55%, and the fluency increases significantly. The effect is very powerful.
According to the overall game test, the overall performance of RTX 4070 Ti SUPER is about 10% higher than that of RTX 4070 Ti, and it can reach about 87%~90% of the performance of RTX 4080. The positioning is really quite accurate. At the same time, the RTX 4070 Ti SUPER can maintain a very high frame rate of the mainstream light chase game under the 2K extreme image quality setting. After DLSS is turned on, it can achieve an extremely smooth level. It is indeed the flagship of the 2K light chase game with a high frame rate. Of course, even if the resolution is increased to 4K, RTX 4070 Ti SUPER can keep most games above 60 fps on average, and DLSS 3/3.5 (including light reconstruction) black technology support can further improve the frame rate and image quality. The experience is also very good.
For RTX 3070 Ti old users, the overall performance of RTX 4070 Ti SUPER has been improved by up to 90%, and it also supports DLSS 3/3.5. With the support of frame generation and light reconstruction, the experience of RTX 4070 Ti SUPER is more than double that of RTX 3070 Ti, so it is worth upgrading.
Productivity performance test
The RTX 4070 Ti SUPER not only has the GPU specification significantly upgraded compared with the RTX 4070 Ti, but also has the video memory bit width increased to 256 bits and the capacity increased to 16 GB, so it will also have a stronger performance in terms of productivity and performance. From the actual measurement, in Blender, RTX 4070 Ti SUPER can lead RTX 4070 Ti by about 12% at most, V-Ray by about 13% at most, Octanebench by about 16% at most. On the whole, in 3D rendering, RTX 4070 Ti SUPER can lead RTX 4070 Ti even more than the leading range of games. In the SPECView test, RTX 4070 Ti SUPER leads RTX 4070 Ti by up to 10%, which also compares the composite theoretical data.
Compared with RTX 3070 Ti, the productivity performance of RTX 4070 Ti SUPER is a qualitative leap. In terms of 3D rendering output performance, RTX 4070 Ti SUPER is about 88% ahead of RTX 3070 Ti, with an average leading range of 77%. If the renderer also supports DLSS 3/3.5, the leading range of RTX 4070 Ti SUPER can be increased several times (refer to the previous D5 renderer test).
In the video transcoding part, RTX 4070 Ti SUPER not only upgrades the efficiency but also increases the number of codecs compared with RTX 3070 Ti, and adds support for AV1 hardware coding. Therefore, from the Da Vinci coding test, RTX 4070 Ti SUPER improves the H.265 coding speed of RTX 3070 Ti by 136%~157% compared with RTX 3070 Ti, and AV1 hardware coding is a function that RTX 3070 Ti cannot achieve. Therefore, for users who often need to do video clips, it is time to upgrade their RTX 3070 Ti to RTX 4070 Ti SUPER.
AIGC performance test
NVIDIA GPU has always supported Stable Diffusion, which not only has perfect compatibility, but also can provide far ahead computing performance. NVIDIA also launched NVIDIA TensorRT, which enables NVIDIA GPU to run trained neural networks more efficiently. From the official reference data provided by NVIDIA, through TensorRT optimization, the Stable Diffusion plotting efficiency of RTX 40 series GPUs can even be improved by up to twice.
The RTX 4070 Ti SUPER has upgraded the video storage capacity to 16GB, which is really suitable for the local plotting task of Stable Diffusion. Here we select Stable Diffusion Automatic1111 WebUI to test the mapping efficiency of RTX 4070 Ti SUPER, and at the same time examine how much the mapping efficiency will be improved after using NVIDIA TensorRT optimization.
In the benchmark part of Stable Diffusion, we used the standard settings officially recommended by NVIDIA to compare with the previous test data. The sampling method is Euler a, the number of sampling steps is 50, the CFG Scale is 7.5, the number of plots per batch is 2, and the number of plots per batch is 10. Finally, the number of plots per minute is counted (the model is V2-1_768-ema-pruned.ckpt).
It can be seen from the test results that the number of maps produced by RTX 4070 Ti SUPER per minute has reached 12.37, significantly surpassing the 10.71 maps produced by RTX 4070 Ti, and about 78% faster than that produced by RTX 3070 Ti. Of course, compared with the Core i9 13900K, the mapping efficiency of RTX 4070 SUPER is 159 times higher.
Next, we use NVIDIA TensorRT to optimize the models of Stable Diffusion XL and Stable Diffusion V1.5 respectively, so as to compare the changes in the mapping efficiency of RTX 4070 Ti SUPER before and after optimization. When using the model of Stable Diffusion V1.5, we choose 512 × 512 and 768 × 768 plot resolutions respectively, while when using the model of Stable Diffusion XL, we choose 1024 × 1024 plot resolution (the same as the image resolution of the training model). For other parameters, the sampling method is Euler a, the number of sampling steps is 50, the CFG Scale is 7, the number of plots per batch is 1, and the number of batches is 10.
After optimization with NVIDIA TensorRT, RTX 4070 Ti SUPER increased by more than 101% in 512 × 512 plot test of Stable Diffusion V1.5, and by nearly 150% in 768 × 768 plot test. In the plot test of Stable Diffusion XL, the increase rate was also up to 106%, which shows that NVIDIA TensorRT effect is quite powerful.
Power and temperature test
In terms of power and temperature, we used Furmark to test the RTX 4070 Ti SUPER Super Sculpture, The maximum GPU temperature is 55.1 ℃, the maximum hot spot temperature is 65.7 ℃, and the maximum video memory temperature is 36 ℃. It can be seen that the power of the three fan cooling system is quite powerful, which ensures that the graphics card can keep cool enough even when the output is nearly 300W. At this time, the fan speed is only 36%, and the mute effect is also excellent. In fact, when playing the game, the game power of the RTX 4070 Ti SUPER Super Sculpture can't reach the limit most of the time. Most of the time, it stays at about 230W, so the video card will be cooler and quieter. This is also the advantage of the RTX 40 GPU's energy efficiency ratio.
After a simple attempt, we further improved the GPU and video memory of the RTX 4070 Ti SUPER Super Sculpture by 200MHz. In the test, the maximum frequency of the GPU reached 3105MHz, the TimeSpy score further increased to 26468, and the maximum power reached 330W. It can be seen that the performance output of the RTX 4070 Ti SUPER Super Sculpture is considerable.
Summary:
The performance is close to RTX 4080, and the game AI is suitable. This "70 is too super"
Finally, let's summarize. As the second member of RTX 40 SUPER series, The positioning of RTX 4070 Ti SUPER in SUPER series is very ingenious. It has the AD103 core of the same origin as RTX 4080, the same 256bit/16GB GDDR6X video memory configuration, and its actual performance is also closer to RTX 4080. It is not too much to say that it is "RTX 4080 Youth Edition", while the price remains at the same level as RTX 4070 Ti, so the cost performance ratio is very outstanding.
At the same time, DLSS 3/3.5 black technology has also brought a huge increase in game frame rate and picture quality for RTX 4070 Ti SUPER, such as the frame rate of Mind Killer 2 and the frame rate of Plague Legend: Requiem, which have even increased by 148% and 133% respectively. The new light reconstruction technology in DLSS 3.5 has greatly improved the rendering accuracy and brought more refined game pictures to players.
In addition, DLSS 3/3.5 can also play a powerful role in many 3D design tool software. For example, after the D5 renderer starts DLSS 3.5, not only the rendering speed is greatly increased, but also the noise reduction effect of the rendered image is quite powerful. Therefore, for designers and users, the RTX 40 series GPU is also a must for upgrading.
In terms of professional applications, RTX 4070 Ti SUPER has upgraded video storage to 256bit/16GB, which can provide more powerful adaptability and higher work efficiency than other RTX 4070 12GB video storage series GPUs in various productivity tools.
More importantly, various mainstream productivity tools have been optimized for NVIDIA GPU. CUDA's ecological environment is far ahead, which makes RTX 4070 Ti SUPER the only choice for designers and users.
In addition, in terms of AIGC, NVIDIA TensorRT has greatly improved the mapping speed of RTX 40 GPU in Stable Diffusion, which has greatly improved the efficiency of AI users who choose RTX 4070 Ti SUPER to complete AI mapping tasks, and further consolidated the position of RTX 40 GPU as the preferred AI acceleration tool.
In general, as the flagship of the RTX 4070 family, the RTX 4070 Ti SUPER not only reproduces the classic GPUs of the SUPER family, but also brings an all-around cost-effective solution for 2K high frame rate game experience and efficient AI design. The AORUS RTX 4070 Ti SUPER 16G under Gigabyte, with its flagship materials and cooling configuration, cool appearance and personalized design, is undoubtedly the best choice for enthusiasts and AI users to upgrade.
