AMD's Radeon version of DLSS is called FSR and due out this year
- Thread starter pax
- Start date
pax
Well-known member
yea about that the various video presentations talked 1st quarter 2022 so sounds like a bit delayed... by end of march at the latest I hope
isn't there alway a big driver release in january or first part of february ?
https://videocardz.com/newz/amd-to-discuss-next-generation-image-upscaling-at-gdc-2022
Videocardz - AMD presenting its latest research results for image upscaling for games said:
pax
Well-known member
Some tidbits are coming out about it now. Sounds interesting:
https://twitter.com/CapFrameX/status/1502375332829188097
Story is its about as good as DLSS 2.0.
https://twitter.com/CapFrameX/status/1502375332829188097
CapFrameX
@CapFrameX
FSR 2.0 to come/announced soon. Saw some footage today. Very impressive.
Temporal upscaling + optimized AA
AMD even claims that it can be better than native. ^^
#FSRGen2
4:06 PM · Mar 11, 2022·Twitter Web App
Story is its about as good as DLSS 2.0.
Of course FSR is going to improve alot with the next iteration. Keep in mind though, that DLSS 3.0 is right around the corner. It's highly unlikely that DLSS will be leap-frogged by a competitor. I guarantee that nvidia is investing ALOT more $$ and has ALOT more top-talent working on this than AMD. I've mentioned it before, but actually the lead architect of DLSS is a friend of mine.
Although I also think that we'll get to a point where all of the upscaling technologies are great and most people can't tell the difference. I already can't tell much difference between FSR and DLSS, unless somebody zooms way in and points out exactly which feature to pay attention to. I doubt I'd notice during actual gameplay.
So far it appears that DLSS will maintain it's lead as the best-of-the-best option, while FSR will become the ubiquitous democratized good-enough option. We'll see where intel's stuff falls in after that.
Well there's lots of other stuff that a DLSS 3.0 could address, beyond just graphics quality:
- Reduce the time/difficulty of implementation in games. That would help alot, and this is an area where AMD currently has a large advantage.
- Move toward a single AI model based on a single training set for everything, rather than individual models for each game. A universal model would be gold here.
- Implement at the driver level, make it available for every game regardless of specific implementations... of course this relies on the previous bullet.
- Improve overall performance, especially for mobile devices. FSR is a good deal more performant than DLSS, making it ideal for portable devices like SteamPad, Switch, or Quest2.
- Include DLSS fall-backs for older graphics cards and non-Nvidia hardware. That's what intel is doing, and likely what AMD is going to end up doing eventually. Sure you'll probably get the best quality & performance on newer nvidia cards, but it would be smart to get a DLSS-branded solution that can run on everything, especially xbox & playstation.
Well they're definitely going to show something this month, their presentation is titled "We're showing off our next generation upscaling", so it'd be embarrassing if they come with blank slides. That being said, they may not be announcing a 2.0 release date, or even anything branded "FSR" at all. It's totally conceivable that they'll just say "here's some stuff our researchers have been poking around with in our labs".