Shulou(Shulou.com)11/24 Report--

‏ every time NVIDIA releases a new generation of graphics cards, it will always cause a sensation among the player community. On the one hand, core performance tends to improve by leaps and bounds, and more importantly, every NVIDIA brings some amazing new technologies. This year's RTX 40 series graphics cards introduce DLSS 3 technology for the first time, which can increase the number of frames by up to 4 times using AI. This will undoubtedly be a revolution in the history of GPU development. ‏

‏ CTOnews.com received the flagship GeForce RTX 4090 Founder Edition in the RTX 40 Series GPU in advance. In this test, we filled up all the configurations to see what the limits of consumer PC are today. The specific configuration is as follows: ‏

‏ design ‏‏: when ‏‏ 's cold silver and black behemoth ‏‏ was first acquired, GeForce RTX 4090 gave people a full sense of power. In the past, high-end graphics cards are generally three fans and three slots, while the GeForce RTX 4090 has become a giant card occupying 3.5 slots due to the increase of power consumption and heat dissipation. This means that a larger chassis needs to be prepared when the machine is installed. At the same time, due to the increase in weight, it is best to use with the graphics card bracket. ‏

The TDP power consumption of ‏ GeForce RTX 4090 has been increased to 450W. In order to carry higher power consumption, the power supply interface has become 16 Pin. At present, there are not many power supplies on the market that support 16 Pin. Fortunately, four 8-Pin to 16-Pin adapters are included in the package. Therefore, it is also recommended that you choose at least 850W full module power supply. ‏

The ‏ GeForce RTX 4090 overall design inherits almost the same design as the RTX 30 series public edition GPU, with dense metal cooling armor and two huge fans in the silver-and-black armor. Vertical air ducts are formed one after the other. ‏

The GEFORCE RTX logo at the top of the ‏ also glows when lit. ‏

‏ due to the large size of the GeForce RTX 4090 video card, a screw hole is provided on the side of the video card, and the video card bracket can be installed by opening the cover plate. ‏

The ‏ GeForce RTX 4090 interface is still standard with 3 DP 1.4 x 1 HDMI 2.1, which can support displays with a resolution of up to 8K. ‏

‏ technology analysis ‏‏: ‏‏ core size skyrocketing, help AI leap forward ‏‏ GeForce RTX 40 series of the first three models are GeForce RTX 4090, GeForce RTX 4080 16GB, GeForce RTX 4080 12GB, our GeForce RTX 4090 is based on the most high-end AD102 core, there are 16384 CUDA cores, 512 Tensor cores, 128light-chasing cores, 512texture units, 176ROP units, more than half the size of the previous generation RTX 3090 core. There should be a true flagship RTX 4090Ti that unlocks all GPC units. ‏

The core frequency of ‏ GeForce RTX 4090 is 2230-2520MHz, which is also more than half that of the previous generation. The video memory is still 384-bit 24GB GDDR6X, but the bandwidth has been increased to 1TB / s. The TDP is 450W, which reaches the grade of RTX 3090 Ti and increases a full 100W compared to RTX 3090. The guess is that the power consumption has to be increased because the core size and frequency have skyrocketed. ‏

The architecture of ‏ GeForce RTX 4090 has also changed a lot, upgrading to the latest NVIDIA Ada Lovelace architecture. It is based on the TSMC 4N NVIDIA custom process, thus achieving a leap in performance and power consumption of up to 2x. Streaming multiprocessors have up to 83 TFLOPS of coloring capacity and twice the throughput of the previous generation. The effective ray tracing computing power of the third generation RT Cores is 191 TFLOPS, which is 2.8 times that of the previous generation. The fourth generation of Tensor Cores adds a new FP8 engine with up to 1.32 petaflops of Tensor processing performance, which is five times higher than that of the previous generation. SER brings up to three times the performance improvement for RayTracing, and up to 25% improvement for overall game performance. ‏

The data mentioned above in ‏ may be abstract. As for the concrete function, the new architecture brings a lot of new functions: due to the introduction of Ada optical flow accelerator, DLSS 3 can predict the motion changes between frames in the scene, realize AI frame insertion, and improve the frame rate while maintaining the image quality. At the same time, it supports the application of dual AV1 encoder, which can not only shorten the video export time, but also have better picture quality. In short, it is suitable for game production. ‏

‏ above is the analysis of the hardware specification and architecture of GeForce RTX 4090. As for the DLSS 3 technology that you are most concerned about, we will explain it in detail in the game testing session later. ‏

‏ theoretical ‏‏ performance: how much progress will be made in theoretical performance by almost doubling the running score and improving the amazing ‏‏ as a flagship? Then we conduct a theoretical test, the first is a single-baked doughnut, after 15 minutes, the core temperature is stable at 67.6 ℃, power consumption is stable at about 443W, in line with 450W TDP. The fan is also far from fully loaded, it seems that this heat dissipation is more than enough, and there is even a lot of overclocking space. But it also means that the GeForce RTX 4090 does have a special power supply. If equipped with a processor like i7 / i9, you should honestly use a power supply of more than 1000W. ‏

‏ then carries out the 3DMark stress test, which can detect whether the performance of the graphics card has declined under continuous running minutes. Generally, more than 97% of the graphics card is considered qualified. The score of the measured GeForce RTX 4090 is 98.9%, and there is almost no frequency reduction, which perfectly passes the test. ‏

‏ scored 34725 points for GeForce RTX 4090 graphics cards in the 3DMark TimeSpy DX12 test, compared with 18981 points for GeForce RTX 3090 and 21862 points for RTX 3090Ti. In other words, the increase of GeForce RTX 4090 has reached 82.95%, which is a terrifying increase. ‏

‏ scored 75013 points for GeForce RTX 4090 graphics cards in the 3DMark Fire Strike DX11 test, compared with 46045 points for GeForce RTX 3090 and 52630 points for GeForce RTX 3090Ti. The increase also reached 62.91%, which means that all games can be run smoothly at 4K resolution, even chase games and VR games can meet the needs of high refresh rates. ‏

In the 3DMark Portal Royal light pursuit test of ‏, GeForce RTX 4090 scored 25080 points, which is more improved than the performance of the grating. Almost double that of GeForce RTX 3090. In the past, GeForce RTX 3090 can experience light-chasing games smoothly at 2K resolution, while RTX 4090 can already play smoothly at 4K resolution. ‏

The performance of RTX 4090 in the theoretical running score is absolutely shocking. We have made a comparison table of GPU physical scores, and the RTX 4090 has almost doubled compared to the previous flagship card in all projects.

‏ DLSS 3 details ‏: AI power game, the number of frames skyrocketed ‏ although the theoretical performance has been improved by nearly 80%, but NVIDIA is not satisfied with this. One of the most important technology upgrades at the press conference was DLSS 3. The DLSS 3 is an AI-driven performance multiplier powered by the new fourth-generation Tensor Core and GeForce RTX 40 Series GPU's new optical flow accelerator, the latest version of NVIDIA's acclaimed deep learning supersampling technology and groundbreaking optical multi-frame generation that will usher in a new era of NVIDIA RTX neural network rendering games and applications. ‏

‏ 's NVIDIA Ada Lovelace-based Optical flow Accelerator can analyze two consecutive game images and calculate frame-to-frame motion vector data for objects and elements without the need for traditional game engine modeling. The multi-frame generation model simultaneously acquires the motion vector and optical flow data in the game engine, which greatly reduces the visual anomalies of AI when rendering elements such as particles, reflections, shadows and lights. ‏

‏ put it simply, in the past, DLSS 2 improved the image quality and the number of frames by rendering a low-resolution image and then AI zooming in to a high resolution. While the new DLSS 3 is compatible with DLSS 2, it also adds a frame generation function, which can insert a new frame calculated by AI between two real frames to double the number of frames. Combined with the super-resolution function of DLSS 2, AI can reconstruct up to 7/8 display pixels, and the game performance can be up to four times better than without DLSS! ‏

‏ said that there is no proof, let's go straight to the test. The latest version of 3DMark has been adapted to DLSS 3's BenchMark, and we can directly see the difference by comparing the test results of GeForce RTX 4090 with the control variables of DLSS 2. In DLSS 2 mode, the average number of frames is 134frames, which is more than two times higher than that without DLSS 2. ‏

‏ but if we turn on the latest DLSS 3 technology, the number of frames on the GeForce RTX 4090 graphics card goes straight to an average of 165frames, more than three times higher than without DLSS. This improvement is quite obvious. ‏

‏ currently, DLSS 3 is supported by many of the world's leading game developers, such as:

● Superhuman (SUPER PEOPLE): the preemptive trial version will be released at 13:00 on October 12, with support for DLSS 3

● reincarnation of Life and death (Loopmancer) will be updated on October 12, with support for DLSS 3

● "against the current" and "Justice 'Fuyun Court'": the new graphic display will be released on October 13, supporting DLSS 3

● "Microsoft simulated Flight" (Microsoft Flight Simulator): a beta version for Xbox Insider program members will be launched on October 17, supporting DLSS 3

● "Legend of the Plague: Requiem" (A Plague Tale: Requiem): will be released on October 18, supporting DLSS 3

In addition, more than 35 games and applications have announced that they will soon support the technology, and as long as they adapt to DLSS 3, they can be backward compatible with DLSS 2, so it is very easy to develop, so it is certain that more and more developers will join DLSS. We selected "Cyberpunk 2077", "Cold against the current" and "Unreal engine 5: Lyra", which have been adapted to DLSS 3, and Demo to test GeForce RTX 4090. In addition, we will test the Demo "Enemies" of a Unity engine. ‏

‏ is the most popular "Cyberpunk 2077" recently because of the trigger club animation of the same name. we ran its own BenchMark in 4K Super Light pursuit over drive mode and got the following data. From these data, we can draw three conclusions: ‏

The physical properties of ‏ 1 and GeForce RTX 4090 are strong enough to run the Super Light-chasing Cyberpunk 2077 at a native 4K resolution. ‏

The improvement of ‏ 2 and DLSS 3 is very obvious. On the premise that both of them are turned on to the super performance gear, the frame number of DLSS 3 is 50% higher than that of DLSS 2 and 2.5 times higher than that of DLSS. ‏

‏ 3. Due to the increase in the number of frames, you can experience improved fluency and better image quality by opening the DLSS 3 quality file in the actual game. ‏

‏, will DLSS 3 have a serious impact on picture quality? We use the NVIDIA ICAT tool for picture comparison, we can see that the difference between the left and right sides of the picture quality is not obvious, it can be said that DLSS 3 technology can really let us "pick up" nearly twice the number of frames. ‏

‏↑ left DLSS closed, right DLSS 3 balanced ‏

Ray tracing is also a revolutionary feature brought about by NVIDIA. It can simulate real panoramic light and reflection in the virtual world. For example, in the picture below, the wet ground after the rain can best show the charm of the light pursuit. If you take a closer look at the puddle, the difference is obvious. If the light pursuit is not turned on, there is only a vague projection, and when it is opened, it will produce a lifelike reflection.

‏↑ left ‏‏ RT ‏‏ closed, right ‏‏ RT ‏‏ enabled ‏

‏ in the domestic martial arts online game "against the cold" ‏‏ "Fuyunting" update ‏‏, also provides RTX ray tracing and DLSS 3 options. As can be seen from the data below, DLSS 3 is even better than Cyberpunk 2077 in "against the current". Leaving DLSS on is almost unplayable, and when DLSS 3 is turned on, the frame rate increases nearly fivefold. Compared with the DLSS 2 in the same gear, it has also increased by nearly 65%. ‏

‏ in terms of picture quality, we chose a detail in the scene. We can see that the difference between the balance between DLSS and DLSS is very small, and even the image quality of DLSS is clearer in some places. This is the charm of AI technology. ‏

‏↑ left DLSS closed, right DLSS 3 balanced ‏

‏ is followed by a game Demo "‏‏ Lyra ‏‏" made using the Unreal 5 engine. The Unreal 5 engine provides full support for RTX ray tracing, DLSS 3 and NVIDIA Reflex, so there will be a large number of 3A masterpieces based on the Unreal 5 engine (such as "Black Myth: Wukong"). ‏

‏ DLSS 3's performance in Lyra also met expectations. When DLSS 3 is turned on, the frame rate increases by nearly 1.7 times. Compared with the DLSS 2 in the same gear, it has also increased by nearly 20%. ‏

‏ in addition to the unreal engine, the Unity engine also provides full support for RTX ray tracing, DLSS 3, and NVIDIA Reflex. Finally, we tested an exploding Demo "Unity:Enemies". This Demo does not have so many gears to choose from, so you can only adjust the switch. When GeForce RTX 4090 opens DLSS 3, the number of frames soars by nearly 2.5 times, and the 4K rendering is silky and smooth. ‏

‏ in terms of clothing texture and hair rendering that is quite difficult for AI, there is almost no difference in image quality between turning on DLSS and turning off DLSS. It can be said that with the help of AI in DLSS 3, 4K light chasing games will eventually become popular and become standard in all kinds of games. ‏

‏↑ left DLSS closed, right DLSS 3 balanced ‏

‏, is DLSS 3 technology perfect? Of course not. The essence of DLSS 3 is that AI generates new frames, and one of the side effects of generating frames must be increased latency. After all, our mouse is the feedback on the frame we see, while AI adds additional frames, which has a great impact on competitive games. So when you turn on the frame generation feature of DLSS 3, NVIDIA Reflex will force it on to minimize the rendering delay caused by the generated frame. ‏

‏ game measured ‏‏: ‏‏ take all 4K masterpiece, challenge 8K play ‏‏ in addition to supporting DLSS 3 masterpiece, CTOnews.com also tested several mainstream games for everyone. GeForce RTX 4090 will be tested at the highest resolution of 4K, and the number of frames can even reach the upper limit of 600frame in the newly launched "Watch Pioneer" and "return". Even if it is a fierce battle, it can be kept at about 500 frames, and e-sports has an excellent experience. ‏

‏ also supports NVIDIA Reflex low latency, which allows CPU to start submitting rendering work to GPU before the previous frame is completed, greatly reducing or even eliminating rendering queues, resulting in faster response time and higher targeting accuracy. ‏

‏ in Tomb Raider: shadow, it is measured that BenchMark can run to 176frames at the highest resolution of 4K, which is good enough for 4K144Hz monitors to play. ‏

The ‏ game is also supported by DLSS, but the version is older, so the effect is certainly not as obvious as DLSS 3, with an average increase to 185frames under measured GeForce RTX 4090. ‏

At the end of ‏, we tested a light-chasing 3A masterpiece "Control". In 4K, the highest image quality and the highest light can only barely play smoothly without DLSS, while GeForce RTX 4090 doubles the number of frames directly after turning on DLSS, which is an excellent experience. ‏

‏ generally speaking, GeForce RTX 4090 can meet all the requirements of high frame number 3A at 4K resolution. With such a strong performance, coupled with the enhancement of DLSS 3, even 8K games can be challenged, and 24g's super large video memory makes it have no burst memory problems from beginning to end, so if you have a 4K144Hz monitor, then the RTX 4090 graphics card will be the best choice. ‏

‏ creative production ‏‏: ‏‏ consumer productivity peak ‏‏ to the GeForce RTX 4090 level, certainly not only playing games, creative production and deep learning ‏‏ is also the main factor. GeForce RTX 4090 is also a GPU accelerated by NVIDIA Studio. At present, there are more than 75 creative applications that can use AI computing power to greatly improve productivity. ‏

‏ in modeling and rendering software, GeForce RTX 4090 can use CUDA to provide hardware acceleration for work. At present, almost all modeling software will optimize the GPU of NVIDIA, so it can achieve the coexistence of efficiency, stability and compatibility. For example, in the commonly used rendering tool V-Ray, you can use RTX accelerated ray tracing to achieve high-performance final frame rendering. With the noise reduction feature of AI, GPU can further speed up interactive rendering. ‏

When ‏ V-Ray Benchmark opened the ‏‏ history ‏‏ rce RTX 4090 GPU CUDA acceleration, the score reached 4221 points, a record. ‏

‏ benefits from the improved performance of light pursuit, and the score of GeForce RTX 4090 in V-Ray GPU RTX has also increased to 5840, which is sufficient for heavy rendering of real-time light pursuit. ‏

‏ finally let's take a look at another rendering software, Octane, which can achieve fast interactive rendering through RTX accelerated ray tracing and OptiX noise reduction with AI. According to the final test results of GeForce RTX 4090, its rendering capability is equivalent to up to 26 GTX 980s working at the same time, and the improvement of GPU is really changing with each passing day. ‏

‏ in the later part of the video, the GeForce RTX 4090 is equipped with the eighth generation NVIDIA NVENC encoder, which supports AV1 coding and dual encoders. AV1 is the next generation video coding technology, which can provide faster video coding and higher quality streaming media transmission performance ‏‏ while occupying the same space. ‏

‏ NVIDIA works closely with a number of industry partners, and today's mainstream multimedia applications can support AV1. At present, the domestic professional version of editing and projection has supported the acceleration of NVIDIA's AV1 video encoder. OBS LVB software also supports AV1-encoded recording of NVENC. It is hoped that AV1-encoded LVB can be adapted as soon as possible, thus greatly saving LVB bandwidth. ‏

‏ commonly used video color editing software DaVinci also provides support for AV1 coding, as a free open source coding, there will be more and more mainstream software to join. ‏

‏ CTOnews.com tested PugetBench For Davinci for everyone. GPU acceleration effects based on NVIDIA CUDA technology can achieve faster real-time video editing and frame rendering, and the output speed using NVIDIA coding can be increased by up to 5 times. The final score of GeForce RTX 4090 can reach 2610, 8K video clips won't stutter, and real-time playback won't have to be an agent. ‏

‏ in addition, NVIDIA also provides NVIDIA Broadcast as an artifact for RTX GPU. Through artificial intelligence technology, it realizes some practical effects, such as audio noise reduction, green screen ‏‏ matting ‏‏, portrait focus and so on. ‏

‏ now I use a simple set of USB cameras and microphones for live streaming at home. Even if there is only one camera on such a computer, with the help of RTX GPU, I can choose NVIDIA Broadcast applications in all kinds of applications. ‏

‏ concluded: AI is the industrial revolution of our generation. Although the testing time of ‏‏ is very tight, the author is still shocked by GeForce RTX 4090. On the one hand, about 80% of its grating performance improvement is indeed huge, which seems to proudly announce a GPU performance breakthrough in the post-Moore's law era. On the other hand, it is lamented that NVIDIA integrates AI technology into all aspects of GPU, and even breaks through the physical laws of graphics cards. ‏ brings several times improvement to the game field through DLSS 3's AI technology and panoramic ray tracing. ‏ for the vast number of developers, how ‏ makes full use of the extraordinary AI computing power of the RTX 40 series GPU determines how we write the next generation. ‏

In terms of ‏ price, the suggested retail price of GeForce RTX 4090 Founder Edition is 12999 yuan. Compared with its huge performance improvement, this price is not expensive. If you are a game enthusiast who wants to experience 4K144Hz, or a professional enterprise user who lives on GPU, then RTX 4090 is definitely the only top choice right now. But if you are an average ‏ player, you don't need to use the mid-to-high-end models of the high-end GPU,NVIDIA Ada Lovelace architecture, or consider the RTX 30 series models whose prices are now stable.

JD.com NVIDIA GeForce RTX ®4090 FE12999 direct link ‏ is worth mentioning that this NVIDIA will launch a limited edition of GeForce RTX 4090 FE, which will be debuted in JD.com at 9: 00 p.m. on October 12, Beijing time, aiming to bring an unparalleled experience to enthusiastic players who love FE designs. At that time, players can choose from the NVIDIA FE version and the RTX 4090 GPU standard frequency version and super version from the top graphics card suppliers. In addition, the non-public version of the RTX 4090 graphics card has been launched by NVIDIA authorized partners Asustek, Seven Rainbow, Gengsheng, Yingchi, gigabyte, Yingzhong, MSI and Suotai, and also debuted on all major platforms at 9: 00 p.m. on October 12 Beijing time.

Click here to enter the RTX 4090 public domain purchase page

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.