AMD: Fidelity FX Super Resolution, June 22nd

I'm interested to see how many developers implement it... hopefully its a easy add. The open source aspect and fact it will run on Nvidia Intel ect is golden. Go AMD.

Also exciting this will work on AMD APUs... the Vega APUs may be long in the tooth and all. But damn if they can hit 60fps in a performance mode that would be something.
 
Can't wait to try this out and compare image quality to native 4k. I have no doubt there will be growing pains.
 
This looks really good. Hard to tell the quality from a compressed video, but it looks decent enough and huge performance gains.

Would really help on my Ryzen APU laptop. Also GTX 1060 support, lol, stick it to the man.
 
Sharpness of textures really suffers, but hey, it's always a balance. DLSS has its share of issues as well. Nice that it's supported on multiple platforms.
 
As you drift into the event horizon and begin to partially fuse with infinity, your last thoughts are "holy fuck the performance is good, but that texture looks a little blurr..."
Or, like the movie Event Horizon, you realize you maybe should have just stayed with a Pascal card instead of opening a literal gateway to hell.

I tell myself this periodically.
 
Oh? This reminds me of Freesync vs GSYNC.

Digital Foundry will definitely be tearing this one apart when it launches.
 
Needs independent testing. What they showed on the 1060 was bad. The other comparison was decent but it was in what looks like an ugly unreal engine game (can't believe the guy was talking about life like photorealistic or w.e. he was saying about those messed up graphics). If its close enough to the target resolution then we're in business.
 
Needs independent testing. What they showed on the 1060 was bad. The other comparison was decent but it was in what looks like an ugly unreal engine game (can't believe the guy was talking about life like photorealistic or w.e. he was saying about those messed up graphics). If its close enough to the target resolution then we're in business.
Godfall is a decent looking game... and the developer has been heavily in the AMD camp. Which makes sense as it was a 6 month PS5 exclusive title. Which is worth noting... AMDs super resolution will now be coming to many console titles.

IT looks like it will also work with pervious gen consoles.... which brings the next question. Could this actually save games like Cyberpunk on last gen hardware ? (this is probably something developers wish AMD had got out the door prior to the console launches)
 
Last edited:
RX500 series folks are gonna be happy about this one me included.😁
So are 5700 owners. :)
If a Quality mode nets me 20-30% I'm golden till perhaps even the gen after the next. lol Especially seeing as the only games that push the 5700s (at 1080/1440) really are the exact type of games that will have this implemented.
 
Looking forward to seeing how much of a performance gain in Resident Evil Village we get on an Nvidia card with this. Was disappointed that the game didn't support DLSS.

Thanks AMD!
 
Umm.. Unreal Engine looks great, maybe one of the best out there. I guess everyone is entitled to their taste, but the graphics are quite good.

Not to derail the topic, but FidelityFX (the old one) actually is okay. Nowhere close to DLSS, but for lower spec machines (that don't support DLSS) or on AMD, it helps a lot.

I did some tests with Rage 2, and I found that setting FideltyFX to around 80% was a good trade-off and netted some nice performance boost while still looking okay.

Obviously it is a more rudimentary algorithm, but it is definitely useful. If you have an older machine you really don't have many options, so it's a nice tool in the box.

Look at the screenshots here if you are interested: https://hardforum.com/threads/more-dlss.1990492/page-9#post-1044651743
 
This is the kind of thing that continues to encourage my support of AMD. They could have made it proprietary, but instead they made it analogous to FreeSync - totally FOSS and not limited to any particular hardware or software and for that they get massive, massive respect. Provided its results are anywhere near DLSS in terms of both aesthetics and performance (which looks to be the case so far) hopefully it will be widely supported and implemented in pretty much all games to come. Being FOSS, open to not DX11, DX12 and even better Vulkan support, and likely platform independent (AMD's Linux compatibility is a feather in their cap, making the best FOSS GPU driver on Linux plus a design so that users who want absolutely maximum 3D performance can use the proprietary extension as a plug-in ontop of the FOSS base! , should make it a first choice for the vast majority of software.

Between Super Resolution and raytracing devs finally optimizing for AMD GPU style hardware there will be some pushback against NV's attempt to handwave victory entirely upon RT/DLSS presence and features this generation. I would however like to see AMD put forth an equal or superior alternative to NVEnc encoding - it allows GeForce Now and similar streaming, so it would be nice if AMD could offer an open source parity alternative ; doubly so if Steam could make use of it for SteamPlay Together / Steam Link and similar streaming
 
Only once you hit 88 fps.
b497a5591b52ea317e8f4313d5075d54.gif
 
Anything that helps graphics quality and keep fps, I'm all for.

Doesn't matter if it's amd or nvidia. All I care about is for more graphic quality, speed, availability, and stability.
 
They could have made it proprietary, but

That may have been difficult in their current market position.

Provided its results are anywhere near DLSS in terms of both aesthetics and performance (which looks to be the case so far) hopefully it will be widely supported and implemented in pretty much all games to come.

From what I've seen it's not a DLSS replacement, they don't look or perform similarly. But something may be better than nothing, especially with improvements. Software-only and agnostic tend to be limiting.
 
Last edited:
Not in their current market position


It's not a DLSS replacement, and from what I've seen they don't look or perform anywhere near the same or even function similarly. But something may be better than nothing.

So would you have preferred they make it entirely proprietary and exclusive (the newest?) AMD hardware, in the hopes of having a "killer feature" directed at both game devs and card owners? Not a fan of that approach, which Nvidia seems to favor. AMD was even in a weaker market position when they made FreeSync's core tech an open standard. It became widely adopted, was more cheaply applied to monitors and often gave an equal or better experience as GSync monitors but without the licensing fee or hardware requirement. Over time, there were more FreeSync monitors out there than GSync to the point that even Nvidia broke with their preference for lock-in to allow their GPUs to access VRR on FreeSync monitors. I imagine

How do you figure? It does a very similar task and sounds as though it is easier to integrate into any particular game or piece of software than DLSS 2.0 (which in turn was more compatible and required less specifics and mess than DLSS 1.0). It certainly seems like a DLSS alternative technology, given that it provides a "upscale" (term used loosely) to give greater performance than native resolution, but with graphically indistinguishable (or close to it depending on the preset level, much like DLSS - different settings to choose the degree of performance boost vs possible visual limitation ) quality. The exact methodology used and technology may differ, but from at least the description given the use case and results seem similar.
 
Many of us will be able to try this out soon enough and make our own decisions on how to best implement it for our own benefit. Awful nice of amd to be inclusive of team greens hardware in their up scaling solution knowing there will never be any reciprocation.
 
How do you figure? It does a very similar task and sounds as though it is easier to integrate into any particular game or piece of software than DLSS 2.0 (which in turn was more compatible and required less specifics and mess than DLSS 1.0). It certainly seems like a DLSS alternative technology, given that it provides a "upscale" (term used loosely) to give greater performance than native resolution, but with graphically indistinguishable (or close to it depending on the preset level, much like DLSS - different settings to choose the degree of performance boost vs possible visual limitation ) quality. The exact methodology used and technology may differ, but from at least the description given the use case and results seem similar.
This isn’t using machine learning. It's just upscaling/reconstruction. My guess is the real DLSS-like tech may come with RDNA3 which should have ML hardware on board.
 
Looks like Intel is wanting to push into open source DLSS with AMD, well according to the twitter post between Kyle and Intel's GPU god
 
Back
Top