AMD FSR antialiasing discussion

[BacBeyond]

Not sure what your problem is. Try reading the PCGH review ... they set UE4's TAAU for a comparison test. They also did not mention in their review that it broke DOF.

So they didn't test the same game that the reddit post is about (KingsHunt) which clearly shows how DOF is broken in it. Have you tested Terminator to see if they use DOF and if it makes a difference?

Yes or no, did Digital Foundry mess up their comparison pictures by having TAAU w/ DOF disabled vs Native 4k?

 

[digitalwanderer]

WHY WOULD ANYONE TEST TERMINATOR?!?!

It was a bad game, it IS a bad game, and no one likes it.

Why, why, WHY?!?

 

[DegustatoR]

WHY WOULD ANYONE TEST TERMINATOR?!?!

It was a bad game, it IS a bad game, and no one likes it.

Why, why, WHY?!?

There aren't exactly a wide choice of games to test FSR in at the moment...

 

[digitalwanderer]

Sometimes no choice is better than a bad one, although I get it since I was playing with Riftbreaker and it's not what I'd play for fun.

 

[troyan]

Why would it bother me? Does it bother you that DLSS has sharpening as well? Its part of how it works and is the expected functionality. Why would it bother me?

FSRs upscaling doesnt improve sharpness. It is the additional independent RCAS pass which creates the illusion of "native like image quality". And yes, it bothers me not to have the option to get the same sharpness level of the "native" picture in most games.

In Terminator for example FSR is so bad that even the UQ mode introduces flickering and pixel crawling. That isnt even close to the the native rendering...

 

[Blooper]

There aren't exactly a wide choice of games to test FSR in at the moment...

I don't get this. For the last two days people have been going on and on about them not being sure FSR will get widespread adoption, that people are just going to forget about it and what not. And comments like this saying stuff like how there are barely any games at launch.

We're two days in, and there's 9 titles. +2 came the day after the official launch. There's another 10 still "coming soon" and a total list of 44 developers on board with implementing the tech was even before the source code has gone public. After that we usually see implementations that AMD hasn't (to our knowledge anyway) even played a major role in, such as CP2077 implementing CAS sharpening as an option in game.

The first DLSS 1.0 title was 5 months after the technology was "launched" by Nvidia and there were only ever 7 games. 9 if you include Control and Mechwarrior 5, which both got DLSS 2.0 updates.

The first DLSS 2.0 game was 15 months ago. Since then we've gotten a grand total of 36 DLSS 2.0 titles and another 19 "coming soon".

Under what basis are people making claims about there being barely any games on launch, or expressing concerns about whether or not the tech will even be implemented? I can understand concerns about the launch lineup being rather bland in terms of the actual games chosen, but there was a much wider selection than when any versions of DLSS launched.

 

[PSman1700]

The comparison to DLSS is a strange one though.

 

[BacBeyond]

FSRs upscaling doesnt improve sharpness. It is the additional independent RCAS pass which creates the illusion of "native like image quality". And yes, it bothers me not to have the option to get the same sharpness level of the "native" picture in most games.

Its part of it. Its not separate.

How it works
FidelityFX Super Resolution is a spatial upscaler: it works by taking the current anti-aliased frame and upscaling it to display resolution without relying on other data such as frame history or motion vectors.

At the heart of FSR is a cutting-edge algorithm that detects and recreates high-resolution edges from the source image. Those high-resolution edges are a critical element required for turning the current frame into a “super resolution” image.

FSR provides consistent upscaling quality regardless of whether the frame is in movement, which can provide quality advantages compared to other types of upscalers.
FSR is composed of two main passes:

• An upscaling pass called EASU (Edge-Adaptive Spatial Upsampling) that also performs edge reconstruction. In this pass the input frame is analyzed and the main part of the algorithm detects gradient reversals – essentially looking at how neighboring gradients differ – from a set of input pixels. The intensity of the gradient reversals defines the weights to apply to the reconstructed pixels at display resolution.
• A sharpening pass called RCAS (Robust Contrast-Adaptive Sharpening) that extracts pixel detail in the upscaled image.

FSR also comes with helper functions for color space conversions, dithering, and tone mapping to assist with integrating it into common rendering pipelines used with today’s games.

https://gpuopen.com/fidelityfx-superresolution/

Do you also hate the fact that DLSS includes sharpening built in as well?

A: DLSS is powered by NVIDIA RTX Tensor Cores. By tapping into a deep learning neural network, DLSS is able to combine anti-aliasing, feature enhancement, image sharpening and display scaling which traditional anti-aliasing solutions cannot. With this approach, DLSS delivers up to 2X greater performance with comparable image quality to full resolution native rendering.

https://developer.nvidia.com/dlss

No clue what you are complaining at honestly but going way off the subject that was Digital Foundry using invalid TAAU results which are missing DOF claiming it looks better because its disabled.

In Terminator for example FSR is so bad that even the UQ mode introduces flickering and pixel crawling. That isnt even close to the the native rendering...

@ 1080p, there is slight flickering in UQ. Funny how you linked to a performance timestamp where it (obviously) looks much worse. They have a 25% performance increase for tiny amount of flickering on a door texture.

If you don't like it, turn it off? Its optional, its not forced. You can use the old render scale which (according to PCGH) is TAAU anyway. They didn't force it but used the built in scaler. (

) to see options screen from Jun 2020 showing its been there before FSR.

 

[digitalwanderer]

Under what basis are people making claims about there being barely any games on launch,

There's barely any games on launch that some of us don't play or have never heard of. That's my basis.

or expressing concerns about whether or not the tech will even be implemented?

I have not nor do I feel this way! I do have a feeling games will be adopting it and a great number of them as quickly as possible, I'm just terribly impatient and want to play with the new shiny toy NOW!

 

[DegustatoR]

I don't get this.

You don't get that there aren't a lot of games to test FSR in at the moment?

 

[Blooper]

There's barely any games on launch that some of us don't play or have never heard of. That's my basis.


I have not nor do I feel this way! I do have a feeling games will be adopting it and a great number of them as quickly as possible, I'm just terribly impatient and want to play with the new shiny toy NOW!

My post wasn't specifically targetted at you, it was more of general frustration given how common this was as a talking point in many reviews but also excessively commonly afterwards as commentary regarding the tech.

Like I said in that long-winded post though, I totally understand concerns regarding the quality of the games in the launch lineup - they were undoubtedly quite bland. But I do not understand those comments claiming that there's few games on launch or that they do not expect many games to adopt the technology.

 

[Blooper]

You don't get that there aren't a lot of games to test FSR in at the moment?

Regardless of if we're talking about the 7 on actual launch day or 9 as of two days later, why don't you try and bring some proof that it's not enough. It's simple - post multiple reviews that go in-depth into looking into all 7 games on launch day, and we'll see how many of those few reviews get updated in a few days time with the new additions because there are so few games.

Or we can even just leave out that last bit and instead try to tally up the average number of games tested in all reviews of FSR. I'd guarantee it'd be about 3-4 games on average - well below that 7 games mark. Hell, did anybody test out all 7 games at all?

 

[Deleted member 13524]

And FSR is using additional sharpening unlike the native rendering to create the illusion it looks as good or even better.

Wow, you actually think that using "illusions" in games that use rasterization is a reason to complain?

Ever heard of bump mapping, normal maps, post processing, everything screen space and idk every rasterization technique ever implemented?

 

[DegustatoR]

What I don't get is why some people are so defensive about a tech which isn't bringing anything new or interesting to the market even. It's just another upscaler, okay? Not the best, not the worst.

There were 7 games to test it in, that's a fact. About 3/4 of these are trash and also run on toasters without any updcaling. So people aren't that interested in checking the tech out by themselves. That's all there is to it. No hidden agenda about the support width or anything.

 

[pjbliverpool]

Riftbreaker demo.

Thanks, I gave that a try. I think I can see a difference in UQ but it's not massively obvious, at least not in this game. Performance mode was obvious, but what I found interesting is the fact this game offers a (presumably dumb) resolution scaler as well. It offers a little more performance that FSR from what I could tell but clearly looks worse. Overall I'm very happy to have the option on my 1070 and would certainly use it if I were performance constrained in a title. Naturally I'd use DLSS first if it were available but since it isn't I consider this a definite positive.

 

[PSman1700]

What I don't get is why some people are so defensive about a tech which isn't bringing anything new or interesting to the market even. It's just another upscaler, okay? Not the best, not the worst.

There were 7 games to test it in, that's a fact. About 3/4 of these are trash and also run on toasters without any updcaling. So people aren't that interested in checking the tech out by themselves. That's all there is to it. No hidden agenda about the support width or anything.

It probably wont be seeing all that much of action on PS5 anyways, the implementations already present there offer almost the same performance/quality, atleast for the bigger studios.
FSR is another upscaler for the PC market, which isnt engine agnostic, which is one of its strongest (if not most) features.

 

[Dictator]

Its part of it. Its not separate.



https://gpuopen.com/fidelityfx-superresolution/

Do you also hate the fact that DLSS includes sharpening built in as well?



https://developer.nvidia.com/dlss

No clue what you are complaining at honestly but going way off the subject that was Digital Foundry using invalid TAAU results which are missing DOF claiming it looks better because its disabled.



@ 1080p, there is slight flickering in UQ. Funny how you linked to a performance timestamp where it (obviously) looks much worse. They have a 25% performance increase for tiny amount of flickering on a door texture.

If you don't like it, turn it off? Its optional, its not forced. You can use the old render scale which (according to PCGH) is TAAU anyway. They didn't force it but used the built in scaler. (

) to see options screen from Jun 2020 showing its been there before FSR.

PCGH has this wrong as I posted earlier in the thread - Terminator does not turn on TAA U when you use the res scaler at all. You would have to enable it in the .ini separately or use unreal engine unlocker - that is why their comparison screens were actually with it off and just standard default UE4 upscale.

I don't get this. For the last two days people have been going on and on about them not being sure FSR will get widespread adoption, that people are just going to forget about it and what not. And comments like this saying stuff like how there are barely any games at launch.

We're two days in, and there's 9 titles. +2 came the day after the official launch. There's another 10 still "coming soon" and a total list of 44 developers on board with implementing the tech was even before the source code has gone public. After that we usually see implementations that AMD hasn't (to our knowledge anyway) even played a major role in, such as CP2077 implementing CAS sharpening as an option in game.

The first DLSS 1.0 title was 5 months after the technology was "launched" by Nvidia and there were only ever 7 games. 9 if you include Control and Mechwarrior 5, which both got DLSS 2.0 updates.

The first DLSS 2.0 game was 15 months ago. Since then we've gotten a grand total of 36 DLSS 2.0 titles and another 19 "coming soon".

Under what basis are people making claims about there being barely any games on launch, or expressing concerns about whether or not the tech will even be implemented? I can understand concerns about the launch lineup being rather bland in terms of the actual games chosen, but there was a much wider selection than when any versions of DLSS launched.

DLSS is completely different technology - its integration being different and more labour intensive is because it is doing something primarily different.

If you were to go to a siggraph conference or GDC, no developer would put FSR on the same band of computation as an image reconstruction technique in its totality because it is not doing that - yet the base of internet opinions is doing that. \FSRs weighted edge upscale would maybe be one part of a full image reconstruction technique that is much more complex in its entirety. Just thinking about FSRs characteristics I think can allow one to come to that conclusion. Why it is so much easier, simpler to integrate, costs so little in performance, and is slotted in after anti-aliasing? Why does AMD describe it as weighted edge upscaling with image sharpening? Instead of looking at those facets which allow one to conclude easily it is radically different than image reconstruction, the internet opinion is to want it to be image reconstruction.

If people keep wanting FSR to be something it is not, they are setting themselves up for the most intense disappointments or grand delusions that will need to be maintained through cognitive dissonance - as the first game that has DLSS in its latest iteration and FSR at the same time will be a bloodbath. It should not be a bloodbath rationally, as a rational person would say "FSR is not even doing the same thing at all, they are not real comparison points", yet prevailing opinion wants this to be AMDs DLSS competitor.

 

[BacBeyond]

PCGH has this wrong as I posted earlier in the thread - Terminator does not turn on TAA U when you use the res scaler at all. You would have to enable it in the .ini separately or use unreal engine unlocker - that is why their comparison screens were actually with it off and just standard default UE4 upscale.
DLSS is completely different technology - its integration being different and more labour intensive is because it is doing something primarily different.

If you were to go to a siggraph conference or GDC, no developer would put FSR on the same band of computation as an image reconstruction technique in its totality because it is not doing that - yet the base of internet opinions is doing that. \FSRs weighted edge upscale would maybe be one part of a full image reconstruction technique that is much more complex in its entirety. Just thinking about FSRs characteristics I think can allow one to come to that conclusion. Why it is so much easier, simpler to integrate, costs so little in performance, and is slotted in after anti-aliasing? Why does AMD describe it as weighted edge upscaling with image sharpening? Instead of looking at those facets which allow one to conclude easily it is radically different than image reconstruction, the internet opinion is to want it to be image reconstruction.

If people keep wanting FSR to be something it is not, they are setting themselves up for the most intense disappointments or grand delusions that will need to be maintained through cognitive dissonance - as the first game that has DLSS in its latest iteration and FSR at the same time will be a bloodbath. It should not be a bloodbath rationally, as a rational person would say "FSR is not even doing the same thing at all, they are not real comparison points", yet prevailing opinion wants this to be AMDs DLSS competitor.

I'm not sure why but you seem to have missed the context of my posts and just want to bash on FSR.

Someone else claimed that PCGH tested TAAU and were wondering why the reddit post didn't address that. Well like you just pointed out, PCGH didn't have the same issue that you did where the image wasn't being modified with broken DOF by TAAU.


At the end of the day, all people are about is the final image quality, and performance for that quality.

If Temporal Upscaling gives the best IQ and performance, people will want it.
If Spatial Upscaling gives the best IQ and performance, people will want it.

https://www.reddit.com/r/Amd/comments/o859le

I see that there is a new post showing off each level, and looking at it FSR looks very good. In a blind test I would pick it over TAAU in all except performance mode, which TAAU has better lace while FSR has better detail everywhere else from the sharpening.

Honestly from your comments here and before FSR released it seems like no matter how good the actual quality is, you want it to be worse.

Your testing and failure to update your article and re-do the video with new KingsHunt images that show an actual one to one comparison with depth of field disabled for Native, FSR and TAAU is telling. You have pushed out false comparison images and even after it was pointed out you haven't retracted them.

 

[Deleted member 13524]

DLSS is completely different technology - its integration being different and more labour intensive is because it is doing something primarily different.

If you were to go to a siggraph conference or GDC, no developer would put FSR on the same band of computation as an image reconstruction technique in its totality because it is not doing that - yet the base of internet opinions is doing that. Just thinking about it can allow one to come to that conclusion. Why it is so much easier, simpler to integrate, costs so little in performance, and is slotted in after anti-aliasing? Why does AMD describe it as weighted edge upscaling with image sharpening? Instead of looking at those facets which allow one to conclude easily it is radically different than image reconstruction, the internet opinion is to want it to be image reconstruction.

If people keep wanting FSR to be something it is not, they are setting themselves up for the most intense disappointments or grand delusions that will need to be maintained through cognitive dissonance - as the first game that has DLSS in its latest iteration and FSR at the same time will be a bloodbath. It should not be a bloodbath rationally, as a rational person would say "FSR is not even doing the same thing at all, they are not real comparison points", yet prevailing opinion wants this to be AMDs DLSS competitor.

How about focusing a little bit on what FSR is instead of what isn't?

FSR is:

1 - Free;
2 - Working on all hardware, not just a subset of the portfolio of one IHV;
3 - Brings a substantial performance uplift on a number of resolutions without discernable image degradation to most on Ultra + high res;
4 - Super useful for low power hardware where potato mode settings aren't enough to run the game;
5 - Free IQ upgrade with no performance penalty when used with DSR/VSR.
6 - Apparently easier to implement than TAAU because it doesn't break post processing effects.
7 - On its way to get quick adoption considering the number of high profile studios and engine creators involved.

Yes we get that DLSS2 is technically superior (first gen implementation, we can see where AMD is going from their patents), and that TAAU resolves better detail when it works well.
But I don't get why that must be the sole focus of the conversation. Nor how using terms like "delusions" or "bloodbath" makes said conversaton more productive.

For 99.9% of PC gamers, FSR will indeed be seen as a competitor to DLSS, regardless of the underlying technology.
It won't look nearly as good on lower resolutions, but it'll work on their intel and AMD iGPUs (it might be huge for SteamPal), on their Intel and AMD dGPUs and on nvidia GTX and MX GPUs. There are pros and cons for each, not just cons for FSR.

 

[Dictator]

I'm not sure why but you seem to have missed the context of my posts and just want to bash on FSR.

Someone else claimed that PCGH tested TAAU and were wondering why the reddit post didn't address that. Well like you just pointed out, PCGH didn't have the same issue that you did where the image wasn't being modified with broken DOF by TAAU.


At the end of the day, all people are about is the final image quality, and performance for that quality.

If Temporal Upscaling gives the best IQ and performance, people will want it.
If Spatial Upscaling gives the best IQ and performance, people will want it.

https://www.reddit.com/r/Amd/comments/o859le

I see that there is a new post showing off each level, and looking at it FSR looks very good. In a blind test I would pick it over TAAU in all except performance mode, which TAAU has better lace while FSR has better detail everywhere else from the sharpening.

Honestly from your comments here and before FSR released it seems like no matter how good the actual quality is, you want it to be worse.

Your testing and failure to update your article and re-do the video with new KingsHunt images that show an actual one to one comparison with depth of field disabled for Native, FSR and TAAU is telling. You have pushed out false comparison images and even after it was pointed out you haven't retracted them.

we do not delete the Kings Hunt images because it is rule number 1 on the eurogamer Website that we leave a geneaology of corrections. We do not amend the og article body or Text or images, but add to it and clarify it. That is how all corrections are done journalistically so people can understand where corrections came in context and we cannot be accused of Altering our content in a secretive way after the fact. Another example of that Praxis is in the World of academia where a paper with an error requiring clarificarion is not deleted from the Account, but an addendum or corrections or Update Is issued.