DLSS, FSR & XeSS, WTF Does All This Mean for Gamers?
Peter Donnell / 1 year ago
It wouldn’t surprise me if most of you had some knowledge of upscaling technologies now, they’re becoming increasingly commonplace in the gaming landscape. What they do and how they work varies, but they’re all typically some form of using AI to take a low-resolution image and make it a high-resolution image, with the main goal being to improve your gaming frame rates. Of course, my featured image above is totally an exaggeration of how this works, but oh, how we mocked movies for this… and now, here we are doing the “enhance” thing in real-time for games.
Upscaling Technologies and Supported Hardware
The core concept of upscaling technologies is easy enough for most of you, perhaps all of you to grasp; turn it on and your performance/FPS will increase. There may be some trade-off to image quality, but that’s true of virtually all graphics settings, turning things on and off will result in better visuals, or better performance. However, I think it’s fair to say that the upscaling world is getting complicated and with Intel ready to enter the market with more new GPUs, it’s getting more complicated still. There are now three GPU brands in play, and they all have their own upscaling technologies.
Nvidia has DLSS, AMD has FSR, and Intel has XeSS. You would think it would be as simple as I have X brand GPU so I’ll use their technologies? Right? RIGHT!? Wrong… ish. If you have an AMD graphics card, you can use AMD FSR, or Intel XeSS. If you have an Intel GPU you can use FRS or XeSS and even some of their iGPUs from 11-gen onwards support a version of it too, and if you have an Nvidia GPU you can use DLSS, FSR, or XeSS. Why? Well, it seems to be that AMD FSR and Intel XeSS are more open source, they use technologies that are present in common GPU hardware and drivers to achieve their goals, while Nvidia leverage a hardware-locked feature on their Tensor cores.
Interestingly, that means much older AMD GPUs can use these forms of upscaling, but for Nvidia, only their RTX series 2xxx, and 3xxx GPUs can use DLSS 1.0 and DLSS 2.0, but now it seems only their RTX 4xxx series cards can use DLSS 3.0. What a flipping mess, eh!?
What’s important to stress is that FRS or XeSS aren’t going to be a magic bullet that gives a performance boost on Nvidia cards, and likewise for any other brands technologies running on graphics cards that are not their own. They have their own in-house technologies for a good reason, and while some will benefit, optimisation issues and hardware differences will likely net different results, both in terms of FPS and image quality, especially on older cards where the DP4a technology that underpins many of these upscaling technologies may not run natively and ends up being sort-of emulated. That requires a much deeper dive and a lot of testing, but I think it’s safer to let Intel get their GPUs to market, and let drivers mature before we throw our full weight behind that project.