• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Fidelity Super Resolution in 2021

Caporegime
Joined
12 Jul 2007
Posts
40,861
Location
United Kingdom
Again, that is my intent, to show the ghosting issues in the worst possible way to show why that guys screenshot should be taken with a massive pinch of salt....

So essentially you think my screenshots are on the same level as this one?
fKGcBHJ.png
If so, oh lordy.... specsavers time for you :p :D


Ain't nobody got time for spoiler tags ;)

+Added spoiler tags.

His are worse in that they have more tails, but even 1 tail is terrible so my point was it doesn't matter whether it's 1 or 4.

Neither is acceptable, but yours are less bad (but still bad) if that's what you are looking for.

Anyway don't answer lets get back on topic. :)
 
Associate
Joined
1 Oct 2009
Posts
1,033
Location
Norwich, UK
For one thing he doesn't show performance numbers vs TAAU or other methods so people can't possibly judge whether they feel it's worth the image quality cost (the title of their video mentioned a performance boost at the cost of image quality and then they only talk about image quality). The particularly bad example of FSR's image quality was in a game called Kingshunt that I've never heard of and may just be an outlier (it was most obvious in the character's dress). There are some not so good DLSS 2.0 implementations too, the quality varies depending on the game but focussing on just one or two examples may give a false picture.

KitGuru tested FSR vs TAAU in Godfall and found it produced better image quality and performance than TAAU.

Alex goes out of his way to present things in a negative way by saying that the only reason that Terminator looks good with FSR is because the dark scenes hide artefacts from running the game at a lower resolution while he provides no performance numbers or context.

This is just wrong. He shows performance metrics regarding TAAU at the end around the 12:25 mark, showing the GPU utilization to hit 60fps when comparing traditional upscaling, to FSR and TAAU. And he clearly states that TAAU requires only a bit more utilization of the GPU. Is this point somehow wrong or in dispute?

Wrong again, he didn't focus on one or two examples, he's shown FSR in what, 4 of the 7 games it's supported in? Godfall, Riftbreaker, Terminator and Kings Hunt. That seems like a good sample size relative to what was released, it's over half. In fact you cite KitGuru as a source here who uses 1 less game in their analysis, so you have a bias application of your standards here.

Even if he were to focus on only 1-2 examples, which he did not, that only may give a false picture, it doesn't mean it necessarily has. And it's funny that you have this standard for sample size but then you're totally OK with a claim that there's 1 game in which you claim goes against that. In fact just watching the Kitguru video now and the TAAU comparison and what he says, quote: "I would say that the UE4 upsample looks fractionally sharper, however it does come with a caveat of some noticeable shimmering" and then switching to a more aggressive upsample scale says "unreals implementation is just about sharper overall". So I'm not sure that really agrees with what you're claiming here.

What's more the scene used for comparison does not have any noticeable use of transparent details like was used in Alex's comparison shots, and that's kind of important because one thing that seems to be true about FSR is that edge detection seems pretty good but it does very little or nothing for quality of textures. Most notably KitGuru is only using a 67% TAAU scale comparison to Quality FSR which is looking at internal resolutions of 2560x1440, and I think the point that Alex was making is that when you get down to 1080p internal resolution the quality of FSR is kinda trash vs TAAU. The performance numbers broadly agree with what Alex was saying which is that TAAU has a very minor increase on GPU load over FSR at an equivalent internal resolution. So is any of that actually not true, what SPECIFICALLY is in dispute here?

*edit* From what I'm reading godfall doesn't have integrated TAAU support and requires trickery to unlock using 3rd party tools. May explain why Alex didn't use that game.

Nothing about what he said about Terminator is wrong. He explained the technical reasons why differences are harder to see in this particular game, noting things like lack of contrast due to darkness, lower quality textures on internal surfaces, and heavy use of post processing. Again you're taking a bunch of technically true things and then just making an accusation that it's done in a negative light without disputing anything. And the blatantly strawmanning that he said "the only reason Terminator looks good..." and that's NOT how he phrased it, that's just how your heard it.

It's a completely legitimate point that games are going to differ in how obvious they mask artefacts, again are you actually disputing this? His literal summary for Terminator is that "FSR fares well here" and you literally hear "the only reason terminator looks good". Absolutely astonishing.

They also didn't use the Ultra Quality pre-set for comparisopn as it would have ruined the narrative.

What narrative precisely? He spoke at length that Ultra Quality preset for FSR looks good and his summary of the entire video. "My conclusion based upon what i have seen with FSR is that it is most useful at 4k in it's Ultra quality mode. But its utility drops off rather quickly at resolution and modes lower than that."

Again do you actually dispute anything he's saying here? Can you explain what the narrative is and where he's wrong?
 
Last edited:
Associate
Joined
1 Oct 2009
Posts
1,033
Location
Norwich, UK
DF are the outliers, is it they are just better than everyone else? is that what we are supposed to believe? Or are THEY just a bit crap? or is this a deliberate editorial?

How are they outliers? I mean...be specific.

If people are going to claim things like this it would be really handy if they actually just gave an example of what they're talking about so that we can verify 1) that it's true 2) doesn't have a legitimate explanation.
 
Soldato
Joined
12 May 2014
Posts
5,250
Well when you and people make posts like this:

"Great so just tell everyone not to drive around any corners then."

Then yes, I will call it out because simply put, it is completely wrong.... again, if you had played the game, you would see that you can drive around corners without the need for having to do extreme parkour camera angles and even when turning corners (and you can tell from the car speed, tyre smoke + tyre burnout marks, that I wasn't doing the slowest possible turn...), there still is no ghosting except one ever so slight bit in one screenshot.....

Images Removed

Did you take those images in the highest graphical settings?
 

DDH

DDH

Associate
Joined
29 Jun 2016
Posts
173
How are they outliers? I mean...be specific.

If people are going to claim things like this it would be really handy if they actually just gave an example of what they're talking about so that we can verify 1) that it's true 2) doesn't have a legitimate explanation.


Well, what was the purpose of showing FSR performance vs TAAU 1080p? Why not FSR ultra vs 1662p? Why just kingshunt?

There are a few issues with the DF vid imo, first like mentioned above they used a mode that most people probably wont ever use. They used kingshunt, a game in beta with poor graphics compared to godfall.
Kingshunt has no option in game to enable TAAU, so DF had to use some trickery to enable it. If they went to these lengths, why not also do that in godfall? Both use UE4, godfall has higher graphics fidelity.
I am hesitant to believe that if DF knew how to enable TAAU in kingshunt, they didnt know how to enable it in godsfall

Then there is also the complete lack of mention of this tech on older gen amd hardware, older and current gen nvidia hardware, intel igpus and AMD APU's. I found this to be a glaring omission.

And finally, TAAU may produce better image quality than FSR but of the two games it has been compared to FSR in, neither have TAAU as an option for the end user to enable. The only way to enable it
is through config editing. So sure its better in some regard except its not available to anyone when FSR is? So really its a mute point

And finally, why talk about the alternatives to FSR like TAAU and how they compare without also speaking on their respective implementation potential which touches on my point above. FSR
can be broadly adopted in many engines and offers comparative image quality to UE TAAU.


I still remember the RTX3000 PR vid they did prior to its launch. This vid having no mention of FSR on nvidia hardware, and that pr video closely align in editorial direction imo
 
Soldato
Joined
30 Dec 2011
Posts
5,545
Location
Belfast
I said it before and I will say it again. Don't try to reason someone out of a position they didn't reason themselves into. You are trying to convince someone who believes one single negative review while ignoring ALL of the mostly positive ones, yet he is asking you to prove for him why the single outlier review is the outlier.

The real question is why to you believe the outlier and not the majority?

FSR at 4K has been proven to work and work well and the only way to tell if it is worse than native 4K at Ultra Quality is by pixel peeping. I literally have a PSD file with Ultra FSR on vs off in The Riftbreaker and another with High Quality DLSS 2.0 on vs off in Death Stranding. Both show that there are some small blurring but both are close enough that I have to really stare at the very iner details to notice. The fact I have to do that proves both FSR and DLSS 2.0 do a very good job at 4K.

At this point you are never going to convince these people that FSR isn't crap. Let it go and enjoy.
 
Last edited:
Man of Honour
Joined
13 Oct 2006
Posts
91,650
And finally, TAAU may produce better image quality than FSR but of the two games it has been compared to FSR in, neither have TAAU as an option for the end user to enable. The only way to enable it
is through config editing. So sure its better in some regard except its not available to anyone when FSR is? So really its a mute point

And finally, why talk about the alternatives to FSR like TAAU and how they compare without also speaking on their respective implementation potential which touches on my point above. FSR
can be broadly adopted in many engines and offers comparative image quality to UE TAAU.

TAAU is fairly easy to implement (the shaders, etc. used by Quake 2 RTX can be seen here https://github.com/NVIDIA/Q2RTX/tree/master/src/refresh/vkpt/shader ) - but a bit harder to tune to avoid undesirable effects than something like FSR - but IMO worth it given the overall better image quality/performance balance.
 
Associate
Joined
1 Oct 2009
Posts
1,033
Location
Norwich, UK
Well, what was the purpose of showing FSR performance vs TAAU 1080p? Why not FSR ultra vs 1662p? Why just kingshunt?

There are a few issues with the DF vid imo, first like mentioned above they used a mode that most people probably wont ever use. They used kingshunt, a game in beta with poor graphics compared to godfall.
Kingshunt has no option in game to enable TAAU, so DF had to use some trickery to enable it. If they went to these lengths, why not also do that in godfall? Both use UE4, godfall has higher graphics fidelity.
I am hesitant to believe that if DF knew how to enable TAAU in kingshunt, they didnt know how to enable it in godsfall

Then there is also the complete lack of mention of this tech on older gen amd hardware, older and current gen nvidia hardware, intel igpus and AMD APU's. I found this to be a glaring omission.

And finally, TAAU may produce better image quality than FSR but of the two games it has been compared to FSR in, neither have TAAU as an option for the end user to enable. The only way to enable it
is through config editing. So sure its better in some regard except its not available to anyone when FSR is? So really its a mute point

And finally, why talk about the alternatives to FSR like TAAU and how they compare without also speaking on their respective implementation potential which touches on my point above. FSR
can be broadly adopted in many engines and offers comparative image quality to UE TAAU.


I still remember the RTX3000 PR vid they did prior to its launch. This vid having no mention of FSR on nvidia hardware, and that pr video closely align in editorial direction imo

The purpose is to demonstrate that FSR isn't the only upscaler on the block and that many engines already have some kind of upscaler support which is superior in image quality. And that this difference is most noticeable at lower internal resolutions and which is where FSR really struggles. They weren't the only reviewer to do this KitGuru made the same comparison but they failed to bother testing lower resolutions. It had already been established in the DF video that FSR at high resolutions like 1440p internal upscaling to 4k looks really good, so the comparison isn't really needed. The same criticism could be used against the other reviewers regarding TAAU and only using 1 game to test it in.

I don't own either Godfall or Kingshunt so I don't know for certain, but reading about Godfall discussions people were using 3rd party hacks to enable TAAU as well. I don't know that Kingshunt being in beta of having worse graphics in any way invalidates upscaling comparisons. It's extremely clear in all FSR reviews that the upscaling from resolutions like 1080p to 4k is extremely bad and Alex's point is that TAAU simply does a much better job for more or less the same performance impact. Again this is not an outlier position, other reviewers did the same comparison they just conveniently avoided looking at low resolutions. As for use case and how common that will be, who knows...that's just pure speculation at this point, I've already said that I'd love to see some actual data on this because my suspicion is that the opposite is true.

One telling thing is that general demand for good/better upscaling has never been very high, we've technically been able to do things like FSR for many generations now, probably at least since things like FXAA were invented. We've even had widespread engine support for TAAU for years but devs can't be bothered to take a few hours to enable it in games. FSR didn't appear until after nvidia did DLSS. And let's face it the only reason DLSS was invented was because no enthusiasts are going to enable RT if it means dropping from 1440p or 4k down to 1080p. So I actually suspect that 1080p upscaling will end up being a more common use case in the long run. Let's put it this way, none of these games do both FSR and DLSS and that's almost certainly not a coincidence. The moment AMD starts to push RT hard and we start to see RT+FSR vs RT+DLSS at likely internal resolutions of 1080p upscaled to 4k, AMD are going to get murdered on image quality comparisons. And that's a large part of why 1080p quality matters.

Discussion about older tech is generally missing from all the reviews that I've watched, again I don't see this as them being outliers. The reason is likely because this kind of thing is very much like FXAA/SMAA it's performance profile is very stable across hardware/games. If it's working correctly it should spit out the same kind of image quality changes for the same kind of cost, these are just glorified filters, there's no real interaction with the rest of the rendering it's just taking a final output image and applying a filter to it.

Well FSR is not available to everyone, it's enabled in 7 games none of which are especially high profile. TAAU is available in probably at least that many games, Alex did give examples at the end of his video of the option in games on the market. But the point is that it's just as trivial to implement as TAAU is because in most cases it'll just mean exposing the engine setting to the end user, and maybe a few quality tweaks. It's not specific to the unreal engine as Alex pointed out it's supported on many mainstream engines already and gave examples of that. He also made the point extremely clear that the argument should be why not have both and let people choose. That seems extremely reasonable to me.
 
Soldato
Joined
30 Dec 2011
Posts
5,545
Location
Belfast
Every review I watched bar the DF review mentioned FSR ran on Nvidia and on AMD and most also mentioned (some in detail) that it supported older GPUs.

HardwareCanucks mentioned it
Gamersnexus mentioned it
Kitguru mentioned it
Hardware Unboxed Mentioned it
PC Gamer mentioned it
Techpowerup mentioned it
 
Last edited:
Associate
Joined
26 Dec 2020
Posts
156
Then there is also the complete lack of mention of this tech on older gen amd hardware, older and current gen nvidia hardware, intel igpus and AMD APU's. I found this to be a glaring omission.

And finally, TAAU may produce better image quality than FSR but of the two games it has been compared to FSR in, neither have TAAU as an option for the end user to enable. The only way to enable it
is through config editing. So sure its better in some regard except its not available to anyone when FSR is? So really its a mute point
I mentioned these points in an earlier post as well. What's the point of alternatives if they can't be used by people? Also, Ignoring the majority of gamers who are on older GPUs is suspect as well. Maybe he is too German/Tech audience-centric and thinks that everyone in the world has 3080/3090s in their systems.
 
Soldato
Joined
30 Dec 2011
Posts
5,545
Location
Belfast
Just watched the excellent, thoughtful and considered reviews by HUB and GamersNexus. Told me everything I need to know.

Yep, they both said it did a very good job but ultimatley needs more support. I liked how HUB also debunked the "this is just a filter" nonsense. I read some reviews, tried the free Riftbreaker demo at 4K and 1080p and concluded it is closer to DLSS 2.0 than 1.0 and a good first attempt.
  • It's open source
  • It's easily implemented
  • It works more than well enough at higher resolutions
  • It works on older GPUs
  • It works on AMD, Nvidia and Intel
I think the naysayers are more just annoyed that AMD did something that the wider tech community are actually happy with.
 
Caporegime
Joined
17 Mar 2012
Posts
48,255
Location
ARC-L1, Stanton System
Yep, they both said it did a very good job but ultimatley needs more support. I liked how HUB also debunked the "this is just a filter" nonsense. I read some reviews, tried the free Riftbreaker demo at 4K and 1080p and concluded it is closer to DLSS 2.0 than 1.0 and a good first attempt.
  • It's open source
  • It's easily implemented
  • It works more than well enough at higher resolutions
  • It works on older GPUs
  • It works on AMD, Nvidia and Intel
I think the naysayers are more just annoyed that AMD did something that the wider tech community are actually happy with.

/
 
Soldato
Joined
30 Aug 2014
Posts
5,984
I've just been using Techpowerup's excellent image comparison tool on my 1440P monitor and to my eyes FSR looks like this compared to native (in static scenes obviously, it may be different in motion):

1. Equivalent in Anno 1800 at Ultra Quality
2. Worse in Godfall, textures are less sharp, but it's not that noticeable at Ultra Quality
3. Better in Kingshunt (ironically given the DF review), perhaps it's the sharpening filter improving the image but you have to go all the way down to Performance for it to look equivalent to native.
4. Worse in Riftbreaker but not much worse, ground textures look blurrier.
5. Better in Terminator, the textures look significantly sharper, perhaps it's due to the sharpening filter

Overall it's pretty good for a first attempt. I hope Techpowerup will do a similar comparison when it comes to DLSS.

Here is the tool: https://www.techpowerup.com/review/...solution-quality-performance-benchmark/4.html
 
Associate
Joined
1 Oct 2009
Posts
1,033
Location
Norwich, UK
I said it before and I will say it again. Don't try to reason someone out of a position they didn't reason themselves into. You are trying to convince someone who believes one single negative review while ignoring ALL of the mostly positive ones, yet he is asking you to prove for him why the single outlier review is the outlier.

The real question is why to you believe the outlier and not the majority?

FSR at 4K has been proven to work and work well and the only way to tell if it is worse than native 4K at Ultra Quality is by pixel peeping. I literally have a PSD file with Ultra FSR on vs off in The Riftbreaker and another with High Quality DLSS 2.0 on vs off in Death Stranding. Both show that there are some small blurring but both are close enough that I have to really stare at the very iner details to notice. The fact I have to do that proves both FSR and DLSS 2.0 do a very good job at 4K.

At this point you are never going to convince these people that FSR isn't crap. Let it go and enjoy.

The assertion that the review is an outlier or that DF are shilling was not my assertion, it was made or implied by other people. I'm simply asking for evidence of this and so far replies have been filled with outright falsehoods, things that you can trivially verify are factually incorrect by watching the video again. What you're engaging in right now is just assuming the validity of your own conclusion, it's a type of circular logic. It wasn't me who reasoned into a position, the position is held by other people, I'm just asking for examples of why they believe that.

Stating that FSR at 4k has been proven to work well in no way contradicts what Alex said nor my views that broadly reflect his. He quite clearly stated, and I have quite clearly stated, that FSR at quality modes in 4k do a good job. You also mischaracterize these points of view as FSR being "crap"... It's extremely weird to me considering people can just check these claims for themselves. What Alex believes and what I believe is documented clearly for anyone to just check what you're saying. I dunno if at this point it's just downright intellectual dishonestly or people are just trolling, but characterizing my opinion on FSR as it being crap is really reprehensible.
 
Soldato
Joined
9 Nov 2009
Posts
24,910
Location
Planet Earth
Correct, officially supported products can be found here (including supported Nvidia products).

However, there are reports from R9 280 users that it works just fine on 21.5.2 so YMMV.

Thanks for linking that list!

You're right but you can see your self what's happening, no one wants to hear "its quite good there is nothing fundamentally wrong with it" that's boring.

All it takes is one corrupt crank to create a fake "Gotcha" and the internet blows up with it and that then becomes the overriding thing associated with it, this has Nvidia's subversive marketing all over it, its A-Typical to how they operate, they use internet commentators wherever they can find them taking their money.

AMD especially on the Graphics side have an ingrained mindshare that they are a bit crap, Nvidia's marketing is at least partially responsible for that, as well as some of AMD's own #### ups, Nvidia's job is to keep that subverted mind share going, its what keeps them at 80% marketshare despite over priced products that also have built in obsolescence to a greater extent than AMD.

No mater how good AMD get you will always get this and with it no actual competition unless these useful idiots are known to be a untrustworthy.

This has Nvidia's modus operandi written all over it.

I think its important for AMD to not only keep on with their good gains in CPU,but also keep releasing competent GPUs. AMD doing well in both will start to bring back more mindshare with their GPUs.
I think because of the limited wafer supply they have and need to supply consoles,AMD otherwise might have sold much more GPUs. Even if they can't beat Nvidia in sales share,getting closer to their historic 30% to 40% share which they used to have during their ATI an early AMD days would be a decent achievement IMHO.
 
Soldato
OP
Joined
6 Feb 2019
Posts
17,831
I've just been using Techpowerup's excellent image comparison tool on my 1440P monitor and to my eyes FSR looks like this compared to native (in static scenes obviously, it may be different in motion):

1. Equivalent in Anno 1800 at Ultra Quality
2. Worse in Godfall, textures are less sharp, but it's not that noticeable at Ultra Quality
3. Better in Kingshunt (ironically given the DF review), perhaps it's the sharpening filter improving the image but you have to go all the way down to Performance for it to look equivalent to native.
4. Worse in Riftbreaker but not much worse, ground textures look blurrier.
5. Better in Terminator, the textures look significantly sharper, perhaps it's due to the sharpening filter

Overall it's pretty good for a first attempt. I hope Techpowerup will do a similar comparison when it comes to DLSS.

Here is the tool: https://www.techpowerup.com/review/...solution-quality-performance-benchmark/4.html


that tool is good in theory but the images are too low resolution - I tried Rigtbreaker on my PC and the differences between FSR and native were much more noticeable than the screenshots on the techpowerup tool. Techpowerup should be using much higher resolution images, allow for zoom in and 3 second animated playback
 
Back
Top Bottom