Fidelity Super Resolution in 2021

[ltron]

I've just played the beginning of the Riftbreaker demo at 1440P and can confirm my thoughts from the screenshots of the game, Ultra Quality looks worse than native but not much worse, ground textures and vegetation are blurrier and shimmering/crawling is slightly accentuated but it's not hugely noticeable for the 20-30% extra performance.

I don't think it's as good as DLSS at 1440P going by this game, but I'm not sure it needs to be given its other advantages.

CAS (Fidelity FX sharpening) at a 75% resolution scale (which gives similar performance as Ultra Quality) looks significantly worse than Super Resolution so it's definitely doing something.

 

[ltron]

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

Yes it's more obvious or not depending on time of day (the lighting from early morning makes the vegetation appear more blurry with it on). I can definitely notice the difference although I am looking for it.

 

[fs123]

I've just played the beginning of the Riftbreaker demo at 1440P and can confirm my thoughts from the screenshots of the game, Ultra Quality looks worse than native but not much worse, ground textures and vegetation are blurrier and shimmering/crawling is slightly accentuated but it's not hugely noticeable for the 20-30% extra performance.

I don't think it's as good as DLSS at 1440P going by this game, but I'm not sure it needs to be given its other advantages.

CAS (Fidelity FX sharpening) at a 75% resolution scale (which gives similar performance as Ultra Quality) looks significantly worse than Super Resolution so it's definitely doing something.

Have you tried the CAS reshade to further sharpen the FSR quality. Here's a comparison I did with FSR UQ with and without CAS vs native :

Spoiler: FSR Ultra Quality CAS 1440P comparison

 

[ltron]

Have you tried the CAS reshade to further sharpen the FSR quality. Here's a comparison I did with FSR UQ with and without CAS vs native :

Spoiler: FSR Ultra Quality CAS 1440P comparison

No I haven't, that's a good point. The sharpening is probably set too low in this game. I imagine it's easy for the developer to fix. Your screenshot definitely looks sharper, how is it in motion with that level of sharpening?

 

[Harlequin]

Wow Digital Shilleries sold their ass to nvidia over this one - they insulted the rest of the tech tubers and publications, fallout already happening - will be another PCPer with the fall.

 

[fs123]

No I haven't, that's a good point. The sharpening is probably set too low in this game. I imagine it's easy for the developer to fix. Your screenshot definitely looks sharper, how is it in motion with that level of sharpening?

Looks almost as good as native even when moving since the extra sharpness makes it look nice and crisp. There is a slight blur but not as bad as TAA type upscaling. It's actually very similar to DLSS blur which I notice on my 1440P monitor.

 

[ICDP]

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.

I really am astounded at the levels of cognitive dissonance on display here. If 9 out of 10 people say the sky is blue and one says the sky is green. The person saying the sky is green is by definition an outlier. Being an outlier does not always equate to being wrong but upon investigation any rational person should realise that there has to be a logical reason for the outlier.

The very definition of outlier is:

a person or thing differing from all other members of a particular group or set

The DF review is the only mainstream review I have read or watched that is negative about FSR. That is why it's an outlier, people are asserting it as a fact because it is a fact by the very definition of the word. So the question becomes "why is it an outlier". It is a fact that DF have been proven to have taken payment from Nvidia in the past to promote their hardware under very favourable conditions. The article they did where they stated the RTX 3080 is up to 100% faster than an RTX 2080 for example. So when I (and many many others) read or hear DF say anything I instanty dismiss their "findings" as suspect and not worth of taking seriously.

That's how reputation works and quite frankly DF have a very deserved poor reputation as Nvidia shills. If it looks like a duck and quacks like a duck, then it is a duck.

This is the DF disclaimer taken straight from their RTX 3080 Early look article. Now try to tell me this doesn't make DF look like an Nvidia PR site.

Full disclosure: I can bring you the results of key tests today, but there are caveats in place. Nvidia has selected the games covered, for starters, and specified 4K resolution to remove the CPU completely from the test results and in all cases, settings were maxed as much as they could be. The games in question are Doom Eternal, Control, Shadow of the Tomb Raider, Battlefield 5, Borderlands 3 and Quake 2 RTX. Secondly, frame-time and frame-rate metrics are reserved for the reviews cycle, meaning our tests were limited to comparisons with RTX 2080 (its last-gen equivalent in both naming and price) and differences had to be expressed in percentage terms

 

[BobShark]

Wow Digital Shilleries sold their ass to nvidia over this one - they insulted the rest of the tech tubers and publications, fallout already happening - will be another PCPer with the fall.

Yup. Riding Jensen's cash hard.

 

[ICDP]

Double post

 

[Godhand007]

Wow Digital Shilleries sold their ass to nvidia over this one - they insulted the rest of the tech tubers and publications, fallout already happening - will be another PCPer with the fall.

I can't recall exactly what happened with PCper, did it have something to do with shilling for Intel. Also, about fallout already happening, can you provide one or two links if you have some time. I want to follow this as this might become a reference for what happens to such media outlets/content creators in the end.

 

[ICDP]

I can't recall exactly what happened with PCper, did it have something to do with shilling for Intel. Also, about fallout already happening, can you provide one or two links if you have some time. I want to follow this as this might become a reference for what happens to such media outlets/content creators in the end.

I have read some reddit posts where DF has stated it is not them who are wrong, but every other tech site is wrong because they don't know how FSR really works.

 

[ICDP]

Out of interest I set RiftBreaker to run at 75% resolution with sharpening enabled on a 4K screen. The IQ was quite a bit lower than FSR Quality let alone Utra, I didn't even have to do a A + B comparison, it just looked instantly worse.

FSR is demonstrably better than normal upscaling and even HUB review showed this to be the case.

 

[Hotwired]

I have read some reddit posts where DF has stated it is not them who are wrong, but every other tech site is wrong because they don't know how FSR really works.

I think it was HUB who I last recall being drastically different to other reviewers on something. Believe they took a different approach to claiming everyone else did it wrong.

Then HUB (i think it was) walked it back and said they dun goofed. Which wasn't unexpected.

 

[Rroff]

I really am astounded at the levels of cognitive dissonance on display here. If 9 out of 10 people say the sky is blue and one says the sky is green. The person saying the sky is green is by definition an outlier. Being an outlier does not always equate to being wrong but upon investigation any rational person should realise that there has to be a logical reason for the outlier.

Not my intention to defend DF here as I've not really been following much of the goings on and just made my own conclusions from having a play with the various tech but GN and DF are outliers in that they usually deep dive and/or do their research and/or have people with expert level of knowledge of what is going on and I've found many of the others tend to run with a story based on not having a full understanding of the situation i.e. what happened initially with the Ampere stability situation and loads of these channels running very authoritatively with a factually wrong version until GN and 1-2 others put out their findings having done a proper analysis.

Out of interest I set RiftBreaker to run at 75% resolution with sharpening enabled on a 4K screen. The IQ was quite a bit lower than FSR Quality let alone Utra, I didn't even have to do a A + B comparison, it just looked instantly worse.

FSR is demonstrably better than normal upscaling and even HUB review showed this to be the case.

In-game resolution scaling won't come close even with sharpening to any kind of upscaling system - in many cases they even still just use bilinear scaling.

 

[oguzsoso]

well because console games use special recontrusction methods but they deliberately leave those out of pc versions...

almost everyone who plays ac:valhalla on series s says that it looks near native, despite it rendering in the range of 720-900p. it proves that series s/x version uses a much more different reconstruction method compared to PC, because even a slight %10 resolution downscale at 1080p makes the image look horrible, blurry and crappy.

either console people are delusional, or just like in re:village, PC version doesn't get the "good" reconstruction method, so that people are forced to upgrade more. see how efficient and brilliant checkerboarding works on re:village on consoles, and how horrible and ****** "interlaced" mode works on PC. this is one of the best examples there is to give.

one of the positive things about FSR will be this. a respectable "upscaler" that works somewhat good.

 

[Rroff]

well because console games use special recontrusction methods but they deliberately leave those out of pc versions...

almost everyone who plays ac:valhalla on series s says that it looks near native, despite it rendering in the range of 720-900p. it proves that series s/x version uses a much more different reconstruction method compared to PC, because even a slight %10 resolution downscale at 1080p makes the image look horrible, blurry and crappy.

either console people are delusional, or just like in re:village, PC version doesn't get the "good" reconstruction method, so that people are forced to upgrade more. see how efficient and brilliant checkerboarding works on re:village on consoles, and how horrible and ****** "interlaced" mode works on PC. this is one of the best examples there is to give.

one of the positive things about FSR will be this. a respectable "upscaler" that works somewhat good.

Never historically been the same demand on PC - if you wanted more performance within reason you had the option of buying a faster GPU.

It has only really changed because nVidia introduced DLSS for a mixture of reasons but largely because it was something that could use the Tensor cores and make real time ray tracing at decent performance levels more of a reality.

Checkerboard type reconstruction techniques are fine as far as they go but more suited to a situation where you are gaming on a large screen looking at it from a distance than typical PC monitor situation for gaming.

(In fact on a sort of related note I feel like half the people commenting in this thread are doing so after looking at scaled down comparisons on a phone or tablet type device and not seeing the difference when you have a full screen image).

 

[PrincessFrosty]

I really am astounded at the levels of cognitive dissonance on display here. If 9 out of 10 people say the sky is blue and one says the sky is green. The person saying the sky is green is by definition an outlier.

Uh-huh, so what did Alex say that makes him an outlier? And why is that specific thing wrong?

The DF review is the only mainstream review I have read or watched that is negative about FSR. That is why it's an outlier, people are asserting it as a fact because it is a fact by the very definition of the word.

But all you're doing is asserting it's negative without evidence. OK, I assert that it's positive without evidence. See how stupid this is?

How about if you believe that something is negative you give an example of why you think it's negative? I've asked this over and over, so far people have largely just stated objectively wrong things that I've literally just checked and verified are wrong.

So again this is just deflection and pivoting away from the point which is, provide some examples of why you think the review was "negative". I'm not interested in your personal emotional response and how you personally cope (or don't cope) with facts which you don't find palatable. That's just emotions occurring in your brain which have little or no relevance to the actual facts. You're confusing your own emotional reaction with the objective nature of the review. To be objectively negative in any way you'd have to have examples of where Alex was saying something verifiably incorrect, that makes the feature look bad where if he was honest it would look good. And so far no one has done that, people have tried but if what they're saying is wrong then the evidence doesn't support the claim.

I'm just asking for examples that support the claim, if what you're saying is true then this should not be hard.

This is the DF disclaimer taken straight from their RTX 3080 Early look article. Now try to tell me this doesn't make DF look like an Nvidia PR site.

ALL of the review sites regarding the 3080 launch (preview) were under strict NDAs about the specifics of what they could review, what games, what settings and what precise conditions. Those are conditions that Nvidia imposed on review sites if they wanted to get early access to the hardware in order to get reviews out for the launch day. DF are just being transparent about this to the end user so they have context for the review. It actually demonstrates open and honest engagement with users, they have zero control over the NDA Nvidia impose, they either agree or don't get early samples, the same as all other review sites.

I have read some reddit posts where DF has stated it is not them who are wrong, but every other tech site is wrong because they don't know how FSR really works.

Source?

 

[Grim5]

Wow Digital Shilleries sold their ass to nvidia over this one - they insulted the rest of the tech tubers and publications, fallout already happening - will be another PCPer with the fall.

Where is the proof I'll wait. Ya'll fanboys are rabid

 

[nvidiamd]

Let me ask you this:
Is it true that Control DLSS ran on Turing shaders in the beginning, instead of the tensor cores (i think it was DLSS 1.9)? And yet it provided the best DLSS implementation at that time and also good enough performance, similar to what the tensor cores are doing:
"Performance-wise, DLSS performs almost identically to the resolution which it is upscaled from. For example, 4K with DLSS set to 1440p runs almost identically to a native 1440p render. The same also applies to other resolutions."
Yet when we were talking in the past about the possibility for AMD to do their own DLSS like feature, everyone was saying that AMD can't do that because it doesn't have the "dedicated hardware".

 

[Rroff]

Let me ask you this:
Is it true that Control DLSS ran on Turing shaders in the beginning, instead of the tensor cores (i think it was DLSS 1.9)? And yet it provided the best DLSS implementation at that time and also good enough performance, similar to what the tensor cores are doing:
"Performance-wise, DLSS performs almost identically to the resolution which it is upscaled from. For example, 4K with DLSS set to 1440p runs almost identically to a native 1440p render. The same also applies to other resolutions."
Yet when we were talking in the past about the possibility for AMD to do their own DLSS like feature, everyone was saying that AMD can't do that because it doesn't have the "dedicated hardware".

Quite a bit of stuff in the DLSS and RT pipelines don't use Tensor cores (even though originally claimed was the case) even though in a best case scenario they can provide a significant performance boost over shaders due to the lack of flexibility and having to hand tune to avoid penalities i.e. Tensor processing is often last in the chain and can't be run simultaneous with other work.

When and where you can use Tensor cores though the speed up can be significant.

At lower resolutions you could probably get away with using Shaders but towards 4+K resolutions you would have to rely on Tensor cores to get decent performance gains.

Off the top of my head IIRC doing it on the shaders on Ampere takes about 4ms at 1080p compared to ~1ms on the Tensor cores at 1440p (output resolution) - I don't know the numbers for 1080p on Tensor but I'd assume sub 1ms.