Nvidia RTX 4090 and 4080: we discover the specifications, strengths and weaknesses of the new graphics cards

Nvidia RTX 4090 and 4080: we discover the specifications, strengths and weaknesses of the new graphics cards

Nvidia RTX 4090 and 4080

"The power of our computers continues to increase thanks to AI", with these words the CEO of Nvidia Jensen Huang, rigorously with his leather jacket, inaugurated the GTC 2022, the event that presents all the news of Team Verde . Artificial intelligence was at the heart of the entire presentation both on the hardware side (with the presentation of the 4000 series graphics cards) and on the software side with the introduction of new cloud services to "increase the reach of Nvidia platforms".

An important appointment for all PC gamers, but not only, as the new cards will enhance the computers of professionals and companies all over the world. For this reason we are going to recap the specifications, strengths and weaknesses of the Nvidia RTX 4090 and 4080, the new flagship graphics cards of the American giant.

By Riccardo Lichene.

The new Ada Lovelace architecture

The new Nvidia RTX 40 GPUs Since its introduction in 2018, Nvidia's RTX line has revolutionized the way video games are designed and experienced, in particular in terms of light management and its bounces on surfaces (so-called ray tracing). Ada Lovelace is the third generation of this architecture and promises to be the future of gaming. To show what it is capable of, a video was shown that Huang said was rendered by a single GPU with Ada Lovelace architecture. It's called RacerX and shows how the new line of graphics cards is able to simulate the most diverse scenarios: from smoke to water, from fabrics to light conditions.

Ada Lovelace, daughter of Lord Byron, is considered one of the first computer programmers in the world because already in the mid-19th century she wrote what today we can consider an algorithm for the Analytical Engine, the first mechanical computer in history. Returning to the present, Huang wanted to reassure fans and investors by specifying that for this new architecture a new production process has been developed with TSMC, the Taiwanese semiconductor giant, which will guarantee the availability of the GPUs. An element that should, therefore, avoid the problems encountered with the previous generation of cards.

Geforce RTX 4090

The RTX 4090 The RTX 4090, the new top of the range GPU from Nvidia, it is twice as fast as its predecessor, the 3090Ti, both in demanding games like Microsoft Flight Simulator and in creative apps like video editing or 3D modeling. Not much detail has been given regarding the graphics card architecture itself, but very reliable analysts already say that even with just the announced specs, the new 4090 will be a marked improvement over the previous generation. Before the conference, there were rumors that the new 4090 would require 600 Watts of power to run, but Nvidia managed to double the performance of the 3090 without increasing the required power. To make it work, however, it will still need 450 Watts, which is no small feat for PSUs under 750 Watts.

The price of 1600 dollars (1949 euros in Europe) may seem like a huge amount but if we consider inflation and production costs, the 4090 should cost more than 1800 dollars so Nvidia should be recognized for the effort to stay efficient in design and production. That said, a new 3090Ti is readily available between $ 1,100 and $ 1,200, so Team Verde may find itself competing with itself.

RTX 4090 datasheet

CUDA Core: 16384 Memory: 24GB GDDR6X Tensor-TFLOPs: 1.321 RT-TFLOPs: 191 Shader-TFLOPs: 83 Over clock: "over" 3Ghz Boost clock: 2.52 GHz Base clock: 2.23 GHz Memory bandwidth: 384- bit TDP: 450 Watt Price: 1949 euro Available from 12 October

Geforce RTX 4080

Geforce RTX 4080 Things get less clear when we analyze the RTX 4080 because it will bea> released in two versions: one with 16 GB of memory and one with 12. Looking at the specifications, however, a strange detail immediately catches the eye: they are completely different cards. First of all, what is perplexing is the fact that the 12 GB version has over 1000 fewer CUDA cores than the base 3080, and the 16 GB version does not have the same number of CUDA cores as the Ti version of the 3080. clocks are much higher than the previous models, but both Gamers Nexus and Linus Tech Tips (two YouTube channels of American analysts among the most authoritative in terms of hardware) have pointed out that in the values ​​presented by Nvidia, especially testing Microsoft Flight Simulator, the performance of the two graphics cards was so similar that it did not justify the price difference of almost 400 euros. The 12 GB RTX 4080 will cost € 1,099 in Europe ($ 899 in the US), while the 16 GB RTX 4080 will cost € 1,469 in Italy and $ 1119 in the States.

Add to this the fact that we do not know how these graphics cards will behave on games without ray tracing because Nvidia hasn't given any information about it.

16GB Geforce RTX 4080 datasheet

CUDA Core: 9728 Momoria: 16GB GDDR6X Tensor-TFLOPs: 780 RT-TFLOPs: 113 Shader-TFLOPs: 49 Boost clock: 2.51 GHz Base clock: 2.21 GHz Memory bandwidth: 256-bit Price: 1469 euros Available in November

12GB Geforce RTX 4080 datasheet

CUDA Core: 7680 Momory: 12GB GDDR6X Tensor-TFLOPs: 639 RT-TFLOPs: 92 Shader-TFLOPs: 40 Boost clock: 2.62 GHz Base clock: 2.31 GHz Memory bandwidth: 192-bit Price: 1099 euro Available in November

The DLSS3 revolution

The performance of the RTX 40 D LSS3 (Deep Learning Super Sampling) is the third generation of the model that increases the framerate of a game without putting the hardware under stress. The backbone of this process is artificial intelligence which, with the computing power of the Lovelace architecture, allows a "quantum leap" in improving the gaming experience. The DLSS2 was going to predict and insert single pixels to increase the number of frames per second using some of the computing power of the CPU. The DLSS3, on the other hand, generates entire frames without using the processor and with a final result 4 times better than just the rendering done by the hardware without the help of AI.

3 games have been the flagship of innovations presented during GTC 2022. The first was Cyberpunk 2077, used to showcase the computing power of the Lovelace architecture thanks to its many neon environments and dynamic city scenarios. Then, to showcase the power of DLSS3 in games that require a very powerful CPU, a Microsoft Flight Simulator trailer was shown running at over 120fps with ray tracing on. Finally, Portal RTX has been announced, a visually stunning remaster of the first installment of the series released in 2007 that will be free for owners of the original.

Modders' paradise

Improving performance of the RTX 40 "9 of the 10 most played competitive video games in the world are due to the modders," said Huang. "We have created an application called RTX Remix dedicated to modifying games". The process is simple: future modders will just have to start the game and do a sort of 3D scan of the environments they want to modify. Then the AI ​​will take care of increasing the resolution of game assets, surfaces and materials as well as giving more accurate physics to objects. Then creating your own mod will be as simple as exporting a file. Many titles don't support ray tracing and it will take a few more years for it to become a true industry standard. RTX Remix, however, will allow the creation of thousands of mods to bring this feature to titles like Morrowind or Fallout 3.

What's next for Nvidia

Nvidia's new GPUs The RTX 4090 showed its muscles regarding the top of the range from Nvidia, but there are still some uncertainties about the RTX 4080: the two versions do not convince in terms of performance / price ratio. On the shoulders of Team Verde, then there is still the problem of the thousands of 3000 series left unsold after the collapse of the GPU market. This happened because the Chinese government has banned many forms of mining, i.e. the calculation of mathematical operations that forms the backbone of the cryptocurrency ecosystem. Computers that do this need a lot of video cards, and this, coupled with the pandemic, has led Nvidia to produce far more GPUs than it could sell. On November 3 AMD will unveil its video cards that will almost certainly not reach the levels of the 4090, but could be much more attractive in the middle price ranges thanks to greater energy efficiency and a better power / price ratio.

Will Nvidia be able to keep its huge market shares unchanged?

Have you noticed any mistakes?

Powered by Blogger.