If you’ve purchased an NVIDIA 2000-series or 3000-series GPU, you’re probably already aware of the big, new feature of these product lines: ray-tracing. What you may not know is that under the umbrella of NVIDIA’s ray-tracing features is another even more game-changing feature: Deep Learning Super Sampling (DLSS). While it sounds like a mouthful, what it does is remarkably simple and totally awesome for gamers. Read on to learn everything you need to know about DLSS, how the technology has evolved, and how to start using DLSS today!
What is DLSS and how does it work?
In short, DLSS lets you run games at higher resolutions and higher framerates for free.
In version 1.0 of DLSS, it took stills of frames of games running at low resolutions and with a lot of aliasing, then used these images to generate high-quality, higher-resolution versions of these images. Finally, it fed the stills and their high-quality counterparts into a supercomputing cluster that trained itself to recognize aliased and low-resolution frames and generate high-quality, high-resolution frames in response.
Though each individual game required “training” first on NVIDIA’s end, once NVIDIA’s AI could successfully recognize low-quality frames and use those to generate high-quality copies, by turning DLSS on, gamers had a lot more control over their experience. If you want to run a game at higher graphical settings or resolutions but are unsatisfied with the performance, DLSS lets you render your game at a lower resolution, getting the performance benefit of doing so. It can output a much higher quality image.
Version 2.0 of DLSS refined the technique’s application. The higher-quality images generated are even more detailed than before, and on the backend, NVIDIA has moved to a game-agnostic system, so every individual game doesn’t need to be tested and trained before DLSS can be implemented. With DLSS 2.0 came some customization of the feature itself: instead of choosing between On and Off, you can pick between Quality, Balanced, and Performance modes, which offer upscaling from a variety of resolutions.
If this sounds a little like black magic to you, that’s normal, but it’s important to remember that DLSS is not a piece of software. It’s not something that can be downloaded or patched into your game. The Tensor cores on NVDIA’s last two GPU lines that enable ray-tracing features are the same cores that power DLSS, which is why DLSS is considered an RTX feature, even if it itself is not actual ray-tracing.
How to use DLSS
Luckily, DLSS can be found with all your other graphical settings in games. This means there isn’t an NVIDIA setting you have to toggle or a Windows setting you have to mess with. Assuming you have an RTX NVIDIA GPU installed and your drivers (alongside the GeForce Experience application) and copy of Windows are up to date, the only thing you have to do is find the DLSS switch in-game.
It’s important to remember to set your game to whatever resolution you’re looking to use for output. If you’re planning on using DLSS to upscale to 4K, your resolution should be set to 4K. By turning DLSS on, under the hood the game will run at a lower internal resolution and will then be upscaled to your desired resolution. DLSS is best used at higher resolutions, like 1440p and 4K, which are particularly expensive, especially when you want to run at framerates higher than 60 frames per second.
Lastly, if you’re already running a game at a high framerate at a high resolution or if your computer is bottlenecked by a different component like your CPU, the benefits of DLSS will be far less pronounced. It’s also important to keep in mind that performance gains will always depend on your card’s load as well as the rest of your hardware.
Top supported games
Not all games will support DLSS! At the time of writing (mid 2021), only 41 games in total support it. However, its presence in many of these games greatly improve performance and visual quality, especially with DLSS 2.0 or newer integration. It even helps offset some of the performance lost by enabling ray-tracing in supported games. Right now, the best implementations of DLSS we’ve seen can be found in the following 10 games:
- Death Stranding
- Cyberpunk 2077
- Amid Evil
- No Man’s Sky
- Metro Exodus Enhanced Edition (Note: the original Metro Exodus supports ray-tracing and DLSS 1.0, while the Enhanced Edition requires ray-tracing and supports DLSS 2.1)
- Minecraft: Bedrock Edition
- Call of Duty: Warzone
- Crysis Remastered
What is FidelityFX Super Resolution?
FidelityFX Super Resolution is a competing open-source standard pioneered by AMD, Nvidia’s main competition. Like DLSS, its goal is to use upscaling technology to enable gaming at higher resolutions without bringing your graphics card to its knees.
Unlike DLSS, however, FidelityFX Super Resolution is NOT hardware-accelerated by Tensor cores or other specialized hardware. It’s an open-source in-software solution for tackling the problem that DLSS is made to solve, and since it’s open source, Intel is likely to make use of it in future GPU releases, too. (And yes, Nvidia cards will support it too.)
Where supported, FidelityFX and DLSS will do a lot to reduce the performance requirements demanded by next-generation games on high-resolution displays. We especially recommend getting a card that supports one or both of these features if you plan on using a 1440p or 4K monitor, since that is where they’re the most useful.
Are you interested in new, modern features like DLSS, HDR, or ray-tracing, or do you think they’re less game-changing than everyone wants to believe? Let us know in the comments down below!