News AMD FidelityFX Super Resolution Image Quality Investigated

[digitalgriffin]

I have to admit, it turned out better than I thought it would. I stand corrected on my previous assessment of it being a blurry mess.

 

[RodroX]

I must be another one getting old, cause at least from those images posted I don't see a real horrible difference between them. I did notice some minor difference while playing and trying DLSS CBP 2077 on my system.

Anyways, with the GPU prices and availability not improving I guess anything that can help with FPS is really welcome, as long as the quality remains decent/good.

 

[aalkjsdflkj]

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.

I don't think anyone is forgetting that image quality is lower with these technologies turned on when compared with them off. But there's a lot more to image quality than just any degradation that comes with use of these technologies. The most obvious is that this may allow a player to improve framerates enough to turn a game from essentially unplayable to playable, and image quality is a whole lot less important if the difference is between not playing and playing.
But beyond that, while I'm not convinced that ray tracing is ready for prime time yet there are some games where it makes a difference, and if these technologies allow you to turn on ray tracing while offering minimal degradation in other areas, then this can let you get a higher overall quality image with ray tracing on and these technologies on compared with ray tracing off and running native resolution. You can replace ray tracing with higher quality textures, higher draw distances, better water effects, etc. So these technologies may allow you to get a better overall image quality if you can turn up the dials in other areas that matter more to you than the fuzziness that these technologies seem to add. In other words, if I get things like ray tracing, better water, higher quality textures, and/or greater draw distances that I DO notice, but there's an added fuzziness that I DON'T notice during gameplay, then these technologies allow for better perceived image quality. I'm not saying that's always the case, but given how little difference I see between the native, ultra, and high quality images vs. the differences I see with high quality textures I think these techs do often allow for higher overall image quality at a given framerate.
The other factor is that I don't think there's any danger of these technologies pushing anyone in the wrong direction. The people who make games and game engines are not going to just stop improving all of the various things that go into image quality because these techs are available. If anything these technologies will spur additional development in those other areas because they know that if people don't have hardware that can handle those increases they can always turn on these techs to get reasonable framerates.

 

[Teeroy32]

Me and my RX480 8Gb are getting excited about this. Currently to play Warzone at 120 fps 1080 I'm running on medium with some low settings, it looks like complete arse and I get texture pop in. Maybe with this I can render the game at 720 high or ultra to get the frames and upscale to 1080. Plus with the new Battlefield coming out I think I'll need FSR bad.

 

[ottonis]

Does it work with low resolution input, too? E.g. 720p - - > 1080p or 360p - - >720p?

 

[Lorien Silmaril]

Does it work with low resolution input, too? E.g. 720p - - > 1080p or 360p - - >720p?

I imagine you could, but keep in mind the lower the resolution the less information the upscaler has to work with.

from 1080p to 1440p or 4K the results may be decent: upscaling from 720p or 360p is unlikely to look good even at 1080p (720p may be ok). just temper your expectations is you are upscaling from very low res.

 

[JarredWaltonGPU]

FYI, we're updating the article and changing the URL. We'll have our own test results now, which aren't wildly different from what AMD reported but it's good to have confirmation. I also tested FSR on Intel UHD 630, running at 1600x900. It does indeed work, and performance mode doubled performance in Terminator: Resistance. As you'd expect, upscaling from 800x450 to 1600x900 results in significant loss of image quality, but then anyone running on Intel IGP is probably looking for whatever they can get to boost performance. The updated article should be live soon, at which point future comments will go there and these comments become 'archived' (no direct link from an article that's live on the site).

 

[alceryes]

I find it odd that some other sites put a much more positive (or at least neutral) spin on FidelityFX.

Just using the word 'Investigated' in the title adds a negative connotation before the article even starts!

Here's the Hothardware review, for comparison - https://hothardware.com/reviews/amd-fidelityfx-super-resolution-tested

Edit: Note that I'm not against either DLSS or FidelityFX. Any technology that helps get better performance at no visible loss to quality (while gaming) is welcome.

 

[JarredWaltonGPU]

I find it odd that some other sites put a much more positive (or at least neutral) spin on FidelityFX.

Just using the word 'Investigated' in the title adds a negative connotation before the article even starts!

Here's the Hothardware review, for comparision - https://hothardware.com/reviews/amd-fidelityfx-super-resolution-tested

I love how people try to read connotation in words like "investigated." Put bluntly, you're reaching. I used "investigated" because that's why I did: I looked at image quality first. Due to other things going on, I wasn't able to get all the benchmarking done in advance. That will be in place shortly (it's just going through the editing pass). The updated article will say "AMD FidelityFX Super Resolution Performance and Image Quality Tested" (unless some other editor comes up with a better headline), because now I've actually had a chance to run more benchmarks -- it's what I was doing most of Tuesday and Wednesday.

Anyway, headlines ultimately come down to SEO: Making an article that will hopefully get people to click through and read it. We try to avoid clickbait as much as possible, but a good headline is still critical. What's a better word to use, that still fits within the 40~60 character limit we try to stick to? I could have written "AMD's FSR Gives Your GPU an Upscaling and Enhancement Pill" for a more entertaining slant, or "Move Over DLSS; AMD's FSR Is the Solution for Non-RTX GPUs," but "tested" and "investigated" are basically synonymous in this case. And since I hadn't completed "testing" of performance yet, I went with "investigated."

 

[hotaru.hino]

I find it odd that some other sites put a much more positive (or at least neutral) spin on FidelityFX.

Just using the word 'Investigated' in the title adds a negative connotation before the article even starts!

Here's the Hothardware review, for comparison - https://hothardware.com/reviews/amd-fidelityfx-super-resolution-tested

Edit: Note that I'm not against either DLSS or FidelityFX. Any technology that helps get better performance at no visible loss to quality (while gaming) is welcome.

To me testing means simply running a repeatable sequence of steps and comparing it against objective criteria. e.g., Running 3DMark, reporting the score, and doing nothing else.

Investigating means diving deeper into what you're testing. e.g., if a card running 3DMark is getting higher scores than expected, looking into why that is. Whatever information that comes up from that can be useful for people to see where said information lines up with what they want. If it turns out this particular card has a better factory overclock or cooling system so it can retain boost speeds higher, that information can be more useful than just "this card got 3000 bogo marks while the average is 2500"

In this case, "testing" simply implies reporting just the performance difference with FSR on and off. "investigating" dives deeper into telling me what other things change with it on and off.

Or, they're both basically the same word in the land of hardware review sites.

 

[JarredWaltonGPU]

The updated article with benchmarks is now live: https://www.tomshardware.com/news/amd-fidelityfx-super-resolution-fsr-performance-tested

Carry on discussions wherever, but the site will only show comments from the new thread (https://forums.tomshardware.com/thr...lution-performance-and-image-quality.3709785/)

 

[ottonis]

I find it odd that some other sites put a much more positive (or at least neutral) spin on FidelityFX.

Just using the word 'Investigated' in the title adds a negative connotation before the article even starts!

To "investigate" is a technical term. It includes not only objective measurements (e.g. fps) but also the subjective evaluation of image quality, sharpness, the overall quality of visual impression. I would love if more hardware review sites woud "investigate" thoroughly.