Although it has existed for a long time, it was not until recently – specifically from the beginning of 2020 that DLSS was really valued and used more widely by users. Especially for those who love Ray Tracing, this algorithm will be indispensable to provide a balanced experience between graphics and frame rate. Just earlier this year, a series of major games were announced to support DLSS, demonstrating how important this technology is also being taken seriously by publishers and NVIDIA at this time.
So what is DLSS after all? How does it work? How great is the use that it is so valued? Let’s find the answers to all these questions through the article below.
In fact, besides benefiting the game experience, DLSS is also promoted by NVIDIA for many other purposes related to technical graphics. In the article, the relationship between DLSS and the game will be emphasized more because it will also be popular and of interest to many brothers.
What is DLSS?
DLSS (or Deep Learning Super Sampling) is an advanced video rendering algorithm NVIDIA put on RTX graphics cards, helping our game experience optimize display quality and speed frame rate. When DLSS is enabled, the image in the game will be rendered with a lower resolution than the original setting in but still ensure sharpness because AI automatically fills the empty pixels. As a result, we will have good graphics – often useful with complex effects like Ray Tracing and a vastly improved frame rate when DLSS is not enabled.
DLSS is considered an alternative to many basic anti-aliasing techniques in the past, including Temporal Anti-Aliasing (TAA) – one of the most popular solutions in PC games, Mobile, etc.
According to NVIDIA, with the most popular version of DLSS 2.0, you will improve the frame rate (or FPS) in the game from 200 to 300%. DLSS 1.0 is not that effective (even criticized for its lack of stability) but has also been confirmed with 70% FPS plus by the Blue team.
Current types of DLSS
The original DLSS, or DLSS 1.0, had its debut to the gaming community around 2018 with many promises related to the game experience. But unfortunately, this is not a very well-received version, mainly due to the lack of stability and many limitations in some blockbuster products. For example, with Battlefield V, DLSS 1.0 can only be enabled when the game is in 4K resolution, instead of right from FHD or even 2K to optimize for gamers.
Released in March 2020, DLSS 2.0 is by far the most popular version of this technology on current games – mainly thanks to its impressive performance and significantly more stability than its predecessor. DLSS 2.0 can be twice as powerful as DLSS 1.0, with higher picture quality and superior frame rates – up to 2-3 times more with many games running at 4K Performance settings. In addition, DLSS 2.0 also allows choosing between many quality modes, Quality, Balanced, and Performance – making the game experience more diverse.
In September 2020, NVIDIA launched DLSS 2.1 with a new Ultra Performance mode – targeting the needs of high-resolution gaming and virtual reality experiences. DLSS 2.1 is sometimes referred to with the same name, “DLSS 2.0”.
How does DLSS work?
If you stop at the first part, you have grasped the working mechanism of DLSS. But if you want to understand more deeply, know the differences between DLSS 1.0 and 2.0, then… please continue reading below.
With DLSS 1.0, NVIDIA will set up an AI grid using its supercomputer, NGX. The network will then be shown a series of before and after images using x64 anti-aliasing, and its task will be to learn how to render with “asymptotic x64” quality. “This process will be repeated many times, and the AI will automatically adjust the algorithm to produce the final result. Problems like ghosting, blurring, etc., like when using TAA, will also be avoided by DLSS” – NVIDIA explained above in 2018.
In addition, some techniques such as “Temporary Feedback” are also applied along with DLSS to help the sharpness of the game’s images remain stable through each frame. Temporal Feedback is the process of applying motion vectors – which are used to describe the direction of movement of objects in the frame – with a resolution greater than or equal to the original setting, helping the system ensure ready for the arrival of the next frame.
As for version 2.0, the processing speed of DLSS has increased thanks to improvements to the AI network significantly. That helps it utilize Tensor Core more efficiently, increasing FPS and removing limitations related to GPU, Settings, and resolution. NVIDIA says that DLSS 2.0 only needs to render 25-50% of the original pixels (even 11% with DLSS 2.1 in Ultra Performance mode) and will come with a new Temporal Feedback style to increase sharpness and stability drastically.
The NGX supercomputer will still train DLSS 2.0’s AI system but will change how it “teaches” it to render images compared to before. Instead of providing anti-aliased and anti-aliased frames in parallel like DLSS 1.0, NGX will feed the AI network “a series of low-resolution, anti-aliased frames,” along with its motion vectors to serve Temporal Feedback.
Based on low-resolution frames, DLSS 2.0’s AI compares them with a previous high-resolution frame (up to 16K – 15360 x 8640) “dissecting pixel by pixel between the two to learn how to render optimally,” explains NVIDIA. This process will also be repeated by NGX thousands of times until the final result.
DLSS 1.0 and 2.0 will be trained according to the above process every time a game that supports this technology is released. After learning, the supercomputer will send the AI versions to the graphics card products through the NVIDIA Game Ready Driver. That’s why when you read the patch notes (Update Notes) of the drivers; sometimes you will see a few lines like “Support Cyberpunk 2077”, “Assassin’s Creed Valhalla Support,” etc.
How to experience DLSS in the game?
To experience DLSS at present, you will need a PC equipped with a graphics card from RTX 2000 or higher. But in the future, this technology may also appear on many devices, and the new Nintendo Switch is a bright name – with a high probability of continuing to stick with NVIDIA Tegra hardware. In addition, on the other side of the front line, the AMD “red team” is also said to be developing AMD FidelityFX Super Resolution (FSR) – a video rendering technique with a similar mechanism.
You need to know some information about DLSS: Mechanisms of action, types, benefits for our game experience, and more.