2023-07-15 14:45:00
O DLSS (Deep Learning Super Sampling) is a technology developed by Nvidia to improve the visuals and performance of games. By rendering the game at a low resolution and using AI algorithms to increase it, DLSS offers improved image quality and higher frame rates. This article will explore how DLSS works, what the differences are between versions, what games are compatible, and also the positives and negatives.
What is DLSS?
We here at Oficina da Net are familiar with DLSS: we use this technology in all game tests with Nvidia cards, as in the analysis of the RTX 4090, 4070 Ti and the Notebook Avell A70 HUB, to name a few examples. But, after all, what is DLSS?
O “Deep Learning Super Sampling” is a technology developed by NVIDIA. It’s an “upscaling” method that transforms a low-resolution image into a high-resolution image, all in real-time using AI. This technique uses learning algorithms to improve graphics quality in games while maintaining or even increasing performance.
How does the DLSS work?
With DLSS turned on in the game’s options, instead of the video card rendering the image in the resolution of your monitor, the game frames are rendered in a lower resolution, usually 720p, and then, algorithms of upscaling in real time to reconstruct the image in a higher resolution, with greater sharpness and details.
In case of DLSS 3.0frame interpolation is also used, where the video card creates artificial frames between the real frames, often doubling the game’s FPS rate.
Cyberpunk 2077 Without DLSS on the left and with DLSS on the right. Notice how GPU consumption has decreased and FPS has increased.
The main goal of DLSS is to allow players to enjoy high quality graphics without compromising game performance. An analogy that we can use is that of an Instagram filter: just like the filters, the DLSS also acts in the post-production of the frames, making the final image sharper and more detailed.
The result of this is a final image with a quality comparable to that of a native rendering, but with much less demand on the graphics processing resources of the video card.
Related news
What is the difference between DLSS 1.0, 2.0 and 3.0?
Each DLSS release has been a complete evolution of the previous generation, but there are compatibilities. Check out the differences between versions below:
02/2019 – DLSS 1.0 – It has two steps, the first one improves the serration of the image using neural networks. The second applies upscaling, using a single low-resolution image, artificially improving it until it reaches the desired resolution. The main negative point of the first version of DLSS is that it needs to be trained specifically in the game that will be applied – that is, few compatible games. But the results were promising, and Nvidia continued to develop the technology.
08/2019 – DLSS 1.9 – It is a DLSS 1.0 adapted to run on cores CUDA instead of cores TENSOR, was used experimentally in the game Control. With unsatisfactory results, NVIDIA decided to go back and keep activities in cores TENSOR.
04/2020 – DLSS 2.0 – Accelerated by artificial intelligence, uses TENSOR cores and is generically trained, being compatible with a much larger number of games, such as Cyberpunk 2077, Crysys Remastered, COD: Warzone, among dozens of others. This is, until today, the most used DLSS, as it is compatible with the RTX 2000 and RTX 3000 cards.
09/2022 – DLSS 3.0 – It is a modified version of DLSS 2.0 with added frame interpolation, only compatible with RTX 4000 series cards. In addition to improving image resolution and sharpness, version 3.0 creates fake frames between real frames. The result is a surprisingly fluid experience with few artifacts.
RTX 4090, card with DLSS 3.0 technology support. Source: Oficina da Net
The downside of DLSS 3.0 is the no input lag increase and incompatibility with graphics cards that are not part of the RTX 4000 series.
What’s the secret that makes DLSS so great?
The secret lies in the way Nvidia uses its supercomputers to train artificial intelligence: DLSS 1.0 was trained on brute force: Nvidia showed the AI thousands of screenshots of the game at 64x supersample-anti-aliasing activated. Then they displayed screenshots of the same game, but this time, without anti-aliasing. The artificial intelligence compared the “before and after” and learned to transform the low quality image into an image close to the one originally presented.
DLSS 2.0 is trained in a broader way: it uses what Nvidia calls “temporal feedback techniques”, which is the process of applying motion vectors to frames, interpreting the direction and speed that objects or characters are moving. moving in the game. Thus, the artificial intelligence knows where the object was and where it is going, resulting in a more fluid movement and with fewer errors.
Since DLSS 2.0 is trained overall rather than on a specific game, it means that the algorithm is able to improve over time, carrying the upscaling knowledge from one game to another. This results in much greater compatibility.
Control with and without DLSS enabled. Source: Nvidia
After AI training, Nvidia sends the data to consumers’ graphics cards via GeForce Game Ready Drivers. With them updated, we can use the Tensor cores of the cards to take advantage of the upscaling technology, even when not connected to the internet.
Games compatible with DLSS 3.0
Compatibility with Nvidia’s upscaling technology is increasing day by day. Check out the complete list of games where DLSS 3.0 can be used:
A Plague Tale: Requiem
Atomic Heart
Black Myth: Wukong
Bright Memory Infinite
Chernobylite
Conqueror’s Blade
Cyberpunk 2077
Dakar Desert Rally
Deliver Us Mars
Destroy All Humans! 2: Reprobed
Diablo IV
Dying Light 2 Stay Human
F.I.S.T. Forged in Shadow Torch
F1 22
F1 23
Forza Horion 5
Hitman III
Hogwarts Legacy
Icarus
ILL Space
Jurassic World Evolution 2
Loopmancer
Marauders
Marvel’s Midnight Suns
Marvel’s Spider-Man Remastered
Marvel’s Spider-Man: Miles Morales
Microsoft Flight Simulator
Midnight Ghost Hunt
Mount & Blade II: Bannerlord
Naraka: Bladepoint
Need for Speed Unbound
Perish
Portal with RTX
Redfall
Ripout
Stalker 2: Heart of Chernobyl
Super People
Sackboy: A Big Adventure
Scathe
Sword and Fairy 7
Synced
The Cycle: Frontier
The Day Before
The Finals
The Lord of the Rings: Gollum
The Witcher 3: Wild Hunt
Throne and Liberty
Tower of Fantasy
Warhammer 40,000: Darktide
Warhaven
Witchfire
WRC Generations
1689439025
#DLSS #benefits #NVIDIA #technology