So if you're playing a game that comes in a bit short of 60 fps, DLSS and FSR can both get you into the fully smooth 60+ fps range. But if you're playing a game at 120 fps and you have a 240 Hz display, our experience is that DLSS won't generally scale that high—it becomes the limiting factor. On the other hand, FSR has no qualms about scaling to higher fps, and if you don't mind the loss of image quality, running in Performance mode often more than doubles performance. (So does running at 1080p instead of 4K.)
That's an interesting point. While the current iteration of DLSS may have an edge at more moderate frame rates, it's possible that FSR could provide an edge for high frame rate gaming, due to the lower overhead. The best way to compare these two techniques would be to eventually test games that support both, and try to roughly match frame rates at a given native resolution using various levels of DLSS and FSR, then compare image quality at whatever levels those end up being for a given game. Performing some testing on more mid-range, GPU limited graphics hardware might make sense too, especially for lower resolutions. Something like an RTX 2060, for example.
And unfortunately, given the nature of the business, I'd be pretty surprised to see many games implement both FSR and DLSS.
I'm not so sure. All it takes is a major game engine to integrate FSR support to make it even easier for developers to include in their games. There's no real reason why a game couldn't support multiple solutions, particularly since each have their own strengths and weaknesses depending on the hardware and settings being used. Aside from Nvidia probably encouraging them not to, to help push their newer cards, of course. DLSS isn't going to do much good for the majority of gamers that currently don't have compatible hardware to run it though. Going by the Steam Hardware Survey, the number of people with RTX cards currently only amount to around 17% of their userbase, so that remaining 83% is a pretty large market that could benefit from a better hardware-agnostic upscaling solution, especially since a lot of that will be lower-end hardware struggling to run newer games well at native resolution. Games often include the option for some form of upscaling, but its often not particularly good, so having a decent standard to work off of could be helpful.
I believe the new Unreal Engine 5 actually includes its own upscaling feature that makes use of temporal data though, so they might just stick with that as their upscaling technique if it happens to work better than FSR. UE5 also has official support for DLSS as well though, so it might also be worth comparing those against one another.
Honestly I'm just wondering something: How come none of this is driver based yet?
That would be the killer feature.
FSR and DLSS should be performing their upscaling before things like interface elements are drawn to the screen, allowing things like text to be rendered clearer and without artifacts at native resolution, as the developers have control over what it gets applied to. A universal, driver-based method would be applying the upscaling and sharpening after the final image is rendered, meaning those elements would be getting upscaled and sharpened too, which is less than ideal. And that's really something that's already largely covered by using the existing Radeon Image Sharpening or Nvidia's similar sharpening option combined with other forms of upscaling. It's possible that AMD will update Radeon Image Sharpening to behave more in line with the developer-integrated options though, even if it wouldn't work quite as well. Though unlike the built-in FSR feature, that also wouldn't be of much use to those without AMD hardware.
The one thing that remains to be seen is if AMD takes ages to add supported games like Nvidia did or will they actually get a decent amount of games on board quickly, considering their current supported games are quite lackluster imo.
In Nvidia's case, the original implementation of DLSS required AI training to be performed on a per-game basis, which likely required more time and effort to implement. The original DLSS was arguably worse than some existing forms of upscaling as well, and relatively few people had hardware capable of utilizing it, so there wasn't much incentive for developers to get on-board, aside from in Nvidia-sponsored titles. DLSS 2.0 has largely addressed most of those issues, though the number of systems capable of utilizing it is still in the minority. FSR has an advantage in that all graphics cards should be able to use it. Of course, if a game engine includes its own capable upscaling technique, like with the aforementioned UE5, games using it won't necessarily need to utilize FSR specifically to provide something similar or potentially better.
I am really surprised that everyone seems to forget what these technologies do in the end. Compromise image quality to get performance...
If we cannot discern image quality while moving, in action etc, then why it is there at the first place? Were all those shiny graphics drawn in vain?
These technologies should be a search for better image quality with less performance impact, not sacrifice image quality for playable performance.
If we cheer to much on this I fear we may push amd and nvidia on wrong direction. Imagine with some dlss 3 or fsr pro you always get 200 fps.. By reducing quality, removing some stuff from the scene? Will we be happy?
These shouldn't be the leading factor of gpus.
The thing is, most people barely notice much difference between games rendered at 4K compared to 1440p at common screen sizes, at least when not carefully analyzing still-frames. But 4K requires significantly more hardware resources for what ultimately only amounts to a slightly sharper image. So if you can render a game at around 1440p, but upscale it to 4K, while applying algorithms to make it look near-indistinguishable from a scene natively rendered at 4K, that opens up more hardware resources for making the game look better in other ways. For example, with raytraced lighting effects, or more detailed environments, things that are likely to provide more noticeable improvements to visuals than just a slight increase in sharpness. So it's not so much "compromising image quality", but rather shifting image quality from areas that matter less to those that matter more, allowing games to look better on a given level of hardware, not worse.
Facebook
Twitter
Instagram