DLSS4 New Flagship: Sauda RTX5080 AMP Extreme Infinity evaluation

Jan 31 2025

We have introduced a number of products based on RTX5080 GPUs, and this evaluation is the new flagship graphics card product launched by the SotX50 series GPU-Saita GeForcertx 5080 16GB AMP Extreme Infinity. When RTX5080, this flagship GPU with a new flagship product line, what kind of spark can it collide? Let's take a look with our order to download the footsteps of the hardware group.

Infinite mirror!

We tested a variety of different graphics programs in the Sotai family in the Sotai special program of "Black Mythical Goku" before. AMPEXTREMEINFINTY is a new flagship product line launched with the RTX50 series. It is located in the middle of the Apocalypse and PGF. It belongs to the authentic flagship positioning. It also adds a new set of design language in Sotai's product system-infinite mirror. This unique and charming design creates an unlimited extension and deep and mysterious sense of space with the assistance of ARGB lights, bringing a surrealist style to the whole machine, which can be regarded as a new style in the current graphics card market.

However, from the front, this graphics card is quite sharp in terms of observation, but the lines on the metal middle frame effectively reduce the aggressiveness of this design, although it is not prominent, it is still quite durable. There are almost three new generation of ring -blade fans on the front. We will talk about the heat dissipation part.

In fact, the two points of the design language of unlimited mirrors are not on the front, but on the framework of the graphics card. Due to the use of gray and gold color matching design, the appearance of the entire graphics card is quite luxurious. In addition to making the airflow more effective, the intermediate dense metal fins also separate the entire golden frame visually, allowing the entire Zhang to let the entire Zhang, so that the entire Zhang to let the whole piece The color of the graphics card is very balanced. At the same time, whether it is the word Zotac Gaming in the lower right corner or the ARGB light bar below, it can be detailed in Firestorm 5.0 software, and the experience is excellent.

In addition to the heat dissipation fins and ARGB lights mentioned earlier, there are 12V-2X6 power interfaces, BIOS fast switching keys and ARGB synchronization lines. Introduced three new technologies: 12V-2X6 power supply safety indicator, reverse power interface for convenient lines, and gold-plated design that allows the interface to be more stable, which can effectively improve the installation and line experience of high power consumption GPU, so that players can maintain themselves to maintain themselves Combat platform.

On the right side of the graphics card is a complete mirror panel, which is the biggest highlight of the design of unlimited mirror. After litting, this panel shows the Zotac Gaming Logo of ARGB lighting, becoming the biggest contributor to the sense of space.

The back of the graphics card has a high -strength alloy backboard, which significantly increases the structural strength of the graphics card. Because of the characteristics of its hollow lines, it can also help the overall heat dissipation performance, so that the wind can easily circulate.

On the interface, this Sotai RTX5080 AMPEXTREMEINFINITY also continues the standard of the RTX 50 series. The three sets of DP 2.1 80Gbps interface+1 group of HDMI 2.1 48Gbps full blood interface can easily meet the current and future high -end display needs.

Icestorm 3.0

The cooling system with the Sotai RTX5080 AMPEXTREMEMEINITY is over ICESTORM 3.0 upgraded on behalf of. In addition to the high -intensity alloy materials used by the three ring blades, middle frames and back panels we saw earlier, the ICESTORM 3.0 cooling system also introduced a giant VC average hot plate+nickel -plated composite thermal pipe+dense fin array. Ax can significantly improve the cooling efficiency of the graphics card. Even large games such as Sai Bo Pochu 2077 and Black Mythical Goku can remain calm, fully release performance, and provide players with a cheerful game experience.

In our 30 -minute FurMark pressure test, the graphics card is stable at 360W at 25 ° C room temperature room temperature room temperature room temperature room temperature room temperature room temperature room temperature room temperature The GPU temperature is only 66 Around ° C, at this time the fan speed is only about 50%. It can be said that there is sufficient space for pressure permeability in this setting system, which is quite good. img src = "https://www.haolingxiazai.com/uploadfile/thumb/7b/d8/75Def59735310F5605ADF3410b86b/1024_width.jpg"

In the EED WAY pressure test, this card's The frame rate stability has reached 99.3%, which is also an equally highly praised result. Firestorm overclocking software has always been known for its excellent experience. This time, the 5.0 version launched with the RTX 50 series GPU has ushered in a considerable upgrade. Make performance tuning easier

Open Firestorm 5.0. The first to introduce the GPU performance page. Users can monitor the core parameters of the GPU here, and perform frequency, voltage and power consumption/temperature wall optimization. And adjust and save three sets of preset for switching. /46/A649FC13F76577F0DA574A1645848369/1024x1024_width.jpg "/>

Users can adjust the three different light effects on the graphics card. They can be synchronized or adapt to their favorite colors respectively. , With the ARGB synchronization line experience in the box.

The leftmost status bar can allow users to monitor the performance parameters of the GPU more precisely, and use the intuitive line diagram to allow users to understand the operation of the current GPU more intuitively, and facilitate the detection of faults or monitoring software BUG.

In the fan bar, users can monitor and adjust the speed of the fan, and switch between overclocking and mute VBIOS. At the same time, we can also see that the Firestorm 5.0 version is in the UI language with the Thai RTX5080 AMPEXTREMEINFINITY that we tested this time, which makes people feel that the details of Sotai's details are quite in place.

Super Revolution of the graphic architecture

ELL architecture, based on NVIDIA in history The great gods named the GPU micro-architecture. This time, Blackwell is also a tribute to the famous mathematician and statisticians, the theorem of Laol Blackwell's theorem.

Due to the semiconductor manufacturing process and the business -feasible chip size gradually approaching the physical limit of the physics, the end of Moore's Law is already a problem that cannot be ignored on the table. On the RTX 50 series Blackwell GPU, NVIDIA introduced a new concept -Neural Shaders neural color device.

It can also be seen from the name that the color unit of this Blackwell GPU is highly dependent on AI computing power. Each set of Blackwell SM units is a specialized product for neural network color device architecture, and the previous generation ADA ADA Only half of the GPU SM unit has different INT32 cores. On Blackwell, the core of each set of SM units can execute the FP32/int32 instruction, with the fifth -generation Tensor Core and the improved sequential execution architecture, which can significantly significantly can significantly. Improve the efficiency of various graphics/AI loads including the running of the graphics card.

In practical applications, the Neural Shaders based on the BLACKWELL GPU, developers can deploy a neural network -based model, texture, material and volume cloud in the game. Developers can train a basic model to achieve high -quality, high efficiency and small -volume picture effects on real -time reasoning on the user end. Compared with the local complete baking light and packaging distribution, the use of AI coloring based on Neural Shaders can also significantly improve the efficiency of the game developer side. It can be said that it is a solution that everyone is happy.

On the Tensor Core of various AI applications including DLSS, the fifth -generation Tensor Core.Holding a lower accuracy FP4 specification data, this can further compress the memory pressure during loads, and can significantly improve the effects of technologies such as DLSS on the consumer -grade game GPU restricted by the memory pool. At the same time, for the Transformer model that has gradually become the focus of the industry in recent years, the effect of the fifth -generation Tensor Core is also quite brilliant. This will also be mentioned when talking about DLSS 4 later.

When it comes to video memory, the RTX 50 series GPU has introduced revolutionary GDDR7 video memory. Due to the use of the PAM-3 signal modulation technology, the GDDR7 memory can significantly reduce power consumption while achieving a higher frequency, and the improvement is considerable: according to NVIDIA, the GDDR7 particles equipped with the RTX 50 series can bring two two GDDR7 particles. At the same time, the performance of GDDR6 memory is only half of the energy, which is quite significant.

The energy consumption of large -scale GPUs is becoming more and more significant today, and the Blackwell architecture can be said to greatly optimize this through more fine voltage control and frequency switching.

For consumer GPUs, graphic output specifications and video codec ability are also a very important parameter. In this regard, Blackwell is naturally not too much. The RTX 50 Series GPU fully supports advanced video output specifications including DP2.1 80Gbps, which stands in the Consumer GPU. At the same time, NvenC and NVDEC have been upgraded to the ninth and sixth generations, respectively, supporting the codecs of AV1 UHQ, MV-Hevc and 422 specifications, and the H.264 decoding efficiency of the basics is also amazing. Both multi -screen users and multimedia creators can benefit from the significant improvement of the Blackwell architecture.

More light, go online faster

Since the RTX 20 series Turing architecture shows the real time tracking technology of the world for the first time Since the revolution brought by expression, the construction of games based on light tracking technology has almost become a default option for 3D games. With the listing of the RTX 50 series, the fourth -generation RT Core is also launched, and the collaboration with the GPU architecture has brought a more colorful pursuit experience.

The fourth -generation RT Core divided the triangular engine into a sequential and decompression engine on the basis of the third -generation ADA light pursuit unit, and introduced a linear scanning ball engine, which can greatly improve the calculation efficiency of the light chasing unit. In conjunction with the improvement of the GPU micro -architecture and video memory technology mentioned earlier, the increase cannot be underestimated.

Under the blessing of the advanced light chase unit, the RTX 50 series Blackwell GPU introduced the new RTX Mega Geometry and the hardware acceleration curve base. The neural network color device mentioned in the previous architecture analysis can provide developers with provision More diverse possibilities bring revolutionary screen performance ability and gaming experience.

dlss 4: Life two, regeneration three, regeneration everything

In -depth cooperation, RTX The DLSS 3 function equipped by the 40 Series ADA GPU can be said to be widely praised. The DLSS 3.5 light reconstruction technology launched last year has allowed players to enjoy in games such as "Cyberpunk: 2077" and "Soul Killer 2" For hundreds of hours of visual feast, the author himself made a special trip to experience the remedy Entertainment ten -year -old god -made sword of the RTX 4080 Super GPU for ten years. With the new technology -DLSS 4 released by the RTX 50 series game GPU, it has further broaden the upper limit on the existing frame generation pipeline.

The highlight of this DLSS 4 is that the transformer model is introduced to replace the previous convolutional neural network model architecture. The traditional convolutional neural network model architecture has already touched the performance limit in many years of update, and the performance of the new Transformer architecture has not only been confirmed again and again in large language models such as AI pictures such as Stable Diffusion. And it has the ability to reason according to the context of the picture, which can bring more stable and logical screens, reduce the phenomenon of ghost shadows and flickering with the launch of the DLSS 4, including the full set of super resolution and frame generation functions. The more advanced Transformer model driver, which brings a significant improvement.… Picture performance.

In this comparison chart of the backyard of the Liang Waterfall Police Station of "Soul Killer 2", we can see that after introducing the transformer model, the screen performance of the DLSS light reconstruction function has been significantly improved. Since the Transformer model has the ability to realize the overall understanding of the picture through the "context", we can clearly see that the texture of the wire is more real on the fence, and the DLSS light reconstruction based on the convolutional neural network model is realized because its model depends on the convolution The characteristics of the nucleus are not outstanding in such scenarios.

Even in games like "Horizon: West" do not support DLSS light reconstruction, we can only see similar improvements in games that support DLSS super resolution. The leather lines on the Eroy backpack still maintain a very obvious three -dimensional sense after the DLSS super resolution blessed by the Transformer model. Players can significantly see the sewing lines of this backpack, which is significantly improved for the overall performance of the game. Essence

In addition to the innovation of the underlying architecture, another highlight of DLSS4 is that it introduces multiple frames to generate functions, allowing games to generate more frames on the basis of DLSS3 frame generation technology. At the same time, because the game screen is needed, the DLSS 4 also introduces a new AI optical current algorithm to replace the hardware light stream accelerator to accelerate the optical field calculation. In summary, this new DLSS 4 pipeline can bring 40%of the speed improvement, and at the same time, it can also reduce the memory occupation by 30%, which can be said to be quite powerful.

In "Battle Hammer 40K: Dark Tide", DLSS4 multiple frames can achieve the maximum game experience of 4K137 frames at the highest quality, which is up to 10%. At the same time Provide more space to introduce high -quality screen technology.

When we consider the DLSS 4 frames to generate more than all previous DLSS technologies, we will find that currently DLSS technology has been able to make 15 of each 16 pixel dots from AI generate. Today, today, which is constructive in the innovation of traditional graphic rendering architecture and the evolution of process evolution, DLSS relying on AI accelerates instead of violent pile of hardware (although NVIDIA is also a master of martial arts in this regard) can bring a more significant gaming experience Improvement.

For players who have previously purchased NVIDIA GeForcertXGPU, the improvement of this DLSS4 on the underlying architecture can also benefit all of them, as we have seen in DLSS3.5 light reconstruction.

Reflex 2: lower delayed gaming experience

Since the official launch of 0 years, this item is optimized The technology of game graphics rendering pipelines to reduce system delay is widely praised throughout the industry. Both game developers, peripheral manufacturers, players and players are happy to see and widely accepted this prestigious technology. A few days ago, more than a hundred popular games have been supported, including "Counter -Strikes 2" and "Fearless Contract" When it comes to popular e -sports masterpieces, it is also an important cornerstone of high availability of DLSS 3 frames to achieve high availability.

As the complexity of graphics technology becomes higher and higher, NVIDIA Reflex 2 also came into being. It introduces a new Frame Warp technology on the basis of the first version of Reflex, which can further reduce the system delay based on the mouse input to predict the next frame.

针对Reflex 2实现首发支持的两款游戏分别将会是The Finals与无畏契约，而这项技术尽管首发支持RTX 50系GPU，将会在未来扩展到全部RTX GPU阵容，让所有Modern NVIDIA GeForce GPU game user experience to the next -generation e -sports experience.

RTX 5080: Complete GB203

The RTX 5080 GPU we evaluated this time is based on the complete GB203 core, with 7 groups of GPC units, 42 groups of TPC units, 42 groups of TPC units, 42 groups. 84 groups of SM units, a total of 10752 groups of CUDA cores.. Although it is still created based on the TSMC 4nm 4n customization process, the RTX 5080 GPU has significantly improved the Tensor Core, RT Core, NvenC/DEC codecs mentioned earlier. In the premise of maintaining a width of 256bit, the memory bandwidth is increased from 716.8GB/s of the previous generation RTX 4080 to 960GB/s. It can bring a considerable improvement in modern games, graphic professional applications and AI loads required by high bandwidth. The comprehensive improvement brought a slight increase in the power consumption of the entire card, from 320W TGP to 360W, but considering that the increase was only 40W, the current RTX 4080/4080 Super users should be able to upgrade directly without changing any accessories without changing any accessories. Essence

We assembled the test platform with the configuration as shown in the figure and conducted some basic tests.

In the 3DMark's grating rendering pipeline test, we see that this RTX5080 AMPEXTREMEINFINITY graphics card is completely surpassed by the previous generation of RTX 4080 Super in five projects.

In two light chasing tests based on NVIDIA RTX technology, we can also see a considerable performance increase. Considering that GB203 does not have a significant improvement in digital specifications compared to AD103, this performance improvement almost depends on the improvement of the Blackwell micro -architecture as mentioned earlier. Even without considering the progress related to DLSS, the improvement of this generation of graphics card alone from the traditional grating and Guangzhuang is quite unavoidable.

and the performance improvement of DLSS 4 can make people down. NVIDIA provides the media with a 3DMark copy that supports DLSS 4 performance testing. We also used this set of software for a round of testing. It can be seen that no matter which mode, the performance after turning on DLSS has been significantly improved, and in the test based on the test of DLSS 3 and traditional convolutional neural network pipelines, the performance of the RTX 5080 is still significant. It is stronger than RTX 4080. After opening the DLSS 4 frames based on the Transformer model, even if only one frame is still generated, it can further improve the maximum performance. The performance is amazing.

Come on, sweet game time

RTX 5080 specifications and performance is so powerful, with DLSS 4 frames generated, panoramic light rays, panoramic light, Tracking cash graphics technology, it can be said to have one battlefield and only one: 4K high brush high -quality 3A game experience. Especially with the high -quality 4K 160Hz high brush display is becoming more and more popular, the selection of RTX 5080, especially this Sota's GeForce RTX 5080 16GB AMP Extreme Infinity graphics card can be said to be a very reasonable choice.

Of course, the performance of the multi -frame generation of DLSS after testing is so good. It is not appropriate to click the game to see it. Considering the scale and positioning of the GPU, this evaluation will only focus on 4K performance. At the same time, the performance gap between the performance gap between the convolutional neural network model and the Transformer model is not the focus of this evaluation, and the evaluation time is indeed limited. I will not target The performance of convolutional neural network+DLSS multiple frames or DLSS 3+Transformer is fully tested and analyzed. Please forgive me.

"Soul Killer 2" is the first batch of games that bring DLSS 4 supporting games that are quite prominent. This is mainly because it is the first game to introduce NVIDIA RTX Mega Geometry technology. This technology can intelligently intelligently can intelligently intelligently Concentrated processing and updating complex geometry. In real time, light tracking calculation can not only reduce the CPU occupation rate, increase the frame rate of the game, but also reduce the memory consumption of the complex light tracking scenario. At the same time, at least in the preview version of the developer used by our tests, Remedy opens the permissions of choosing convolutional neural networks or Transformer models, which enables us to compare the performance in several modes. However, due to its native characteristics of the development of DLSS 3.5, this is also the only game we test under high -screen preset+high -light tracking preset.

It can be seen that even if the multi -frame generation is not opened, it rely entirely on the architecture upgrade of the RTX 5080 GPU itself and the DLSS pipeline based on the Transformer model rather than the convolutional neural network. Pursuing setting up the easily realized 1% LOW frame has 4K 60 frames, and in the highest performance DLSS 4X multi -frame generation mode, the most mainstream 4K 160Hz display can become the bottleneck of the game. This is for the spiritual killer 2 This game is really shocking.

" Cyberpunk 2077 "is an essential work for the innovation of graphic technology innovation in NVIDIA in recent years. The DLSS 4 technology is officially introduced in the latest 2.21 version update, and the same is the same. Allowing players to choose to use the traditional convolutional neural network model or the Transformer model for games, which also provides us with a good way to explore the two in actual games. ; "

In gram 2077, we use it Light tracking: Speed preset testing, and we can also see that the performance improvement of the DLSS pipeline based on the Transformer model rather than the convolutional neural network is improved under the collaboration with high -performance RTX 50 series GPU: DLSS performance is almost doubled. The original average frames will increase the 4K 30 frames to 1% LOW frame can be close to 4K 60 frames. This performance improvement can be perceived quite significantly for a first -person ARPG game like 2077. It is really important. A26/ 1024x1024_width.jpg "/>

In DLSS 4 game works, we can also see similar performances. In addition to the game, the work scenarios of various multimedia content creation, 3D modeling and other work scenarios are also a good place for high -performance GPU exhibitions, especially the appearance design of the Saita GeForce RTX 5080 16GB AMP Extreme Infinity. The scene is not prominent, and it is very suitable for the purchase of effective professionals.

Readers who know more about industrial AI must have heard of the name of MLPERF. As an industry standard AI test kit, we introduced an AI test kit for the MLPERF Client this time. Based on the Meta open source LLAMA 2 model, it can effectively test the theoretical performance of the system under common AI load. We compare the token generation rate value as the benchmark. It can be seen that among all four loads, RTX5080 AMPEXTREMEINFINITY significantly surpasses the RTX 4080 Super GPU as the benchmark, which means that in various AI applications, this new RTX5080 AMPEXTREMEINFINITY can bring more efficient energy capacity Performance.

< / In addition to the theoretical Mlperf Client, we can also test the actual performance of the RTX 5080 GPU under the PROCYON kit RTX 5080 GPU under the actual performance of the LLM application. Let's take a look at Wensheng. PROcyon's Wenshengtu test is based on the FP4 and FP8 specifications based on the FP4 and FP8 specifications. This table records the average time required to generate a picture. It can be seen that RTX5080 AMPEXTREMEINFINITY has considerable performance improvement under FP4 or FP8, especially under the first native FP4 format of the RTX 50 series, performance improvement is quite significant. For the daily use of AI loads, the use of a low -precision format such as FP4 can significantly improve efficiency under the premise of not loss of loss, which is very easy to use.

In the test session of the big language model, we have also seen performance progress that cannot be underestimated, especially in LLAMA models with higher memory throughput. Considering that there are already many plans to deploy large language dialogue models such as Qwen or Deepseek locally, the performance improvement brought by RTX5080 AMPEXTREMEINFINITY is obviously very gratifying.

Of course, not all productivity applications have introduced AI acceleration. As a computing power base, how does this RTX5080 AMPEXTREMEINFINITY manifest? We used the industry standard SpecViewPerf 2020 V3.1 for a full set of performance testing. It can be seen from the results that the RTX 5080 has achieved considerable and significant lead, especially in the tests of the two individual users of Maya and Solidworks.

and Vray6 rendering test, we also see quite good performance.

Summary

The whole set of test runs down, we appreciate this RTX5080 AMPEXTREMEINFINITY graphics card from Sotai. On the one hand, this new set of unlimited mirror design language can be said to be a positive example with excellent creativity and good execution. The overall design of the graphics card is quite harmonious. It is really amazing with ARGB lighting and infinite mirror effect, but at the same time it will not be in Seeming scenes appear to be in serious occasions. On the other hand, the full set of advanced features of the RTX 50 Series GPU is displayed on this RTX5080 AMPEXTREMEINFINITY, and it can also make various users feel the obvious experience improvement. At the same time, with the Firestorm 5.0 software released by this generation of new GPUs, it also feels refreshing through the innovative UI design. In general, this RTX5080 AMPEXTREMEINFINITY is a high -standard flagship GPU worth starting. If you are a gamer, or you plan to buy a high -performance GPU for work occasions, then this RTX5080 AMPEXTREMEINFINITY must not be missed.