Tiny Tina's Wonderlands: FSR 2.0

[maxus24]

Tiny Tina's Wonderlands has recently been updated with support for AMD FidelityFX Super Resolution 2.0 (FSR 2.0). In this mini-review, we take a look and compare the image quality and performance offered by FSR 2.0.

Show full review

 

[ModEl4]

Let's hope Unreal Engine 4 & 5 implementation to be better.
Joking aside, it seems the FSR 2.0 implementation procedure to require more effort vs DLSS 2.0 in order to hit optimal results but I may just reading the situation wrong.

 

[docnorth]

Slightly(?) out of topic. How do we pronounce Tiny Tina?

 

[Chane]

I don't think it's implemented correctly. I'm not seeing the typical benefits associated with FSR 2 in any of the screenshots. It would be nice to have a test of it done on a Radeon to see if the performance loss is still present under 4k.

 

[mechtech]

"In terms of performance, the FSR 2.0 performance uplift at 4K resolution is a great improvement to the game even in Quality mode, but 1440p and 1080p have issues."

Well since 4k is where you need it, I would say that one is the most important, followed by 1440 than 1080.

 

[konga]

The implementation here is messed up. Some of my friends have sent me some real awful looking screenshots full of ghosting and other artifacts. Image quality completely breaks down when there are lots of particles and other movement on screen, which is not uncommon in this game.

 

[PrettyKitten800]

I was so excited at first because I went from 60-80fps @ 4k to well over 100fps. The longer I played the game though, the worse the problems got. Tons of artifacts, chunks of screen going one solid color, flickering in and out… then whole objects like the vending machines or the vault stopped rendering.

The game is pretty buggy to begin with, but I never had graphical issues before.

 

[wolf]

RIP DLSS

 

[InVasMani]

I don't think it's implemented correctly. I'm not seeing the typical benefits associated with FSR 2 in any of the screenshots. It would be nice to have a test of it done on a Radeon to see if the performance loss is still present under 4k.

Possibly is or isn't or needs more adjusting. I do agree though FSR should be tested on Radeon hardware for obvious reasons. If it works at all on Nvidia that's great, but that's not how it should be tested ideally especially when DLSS just doesn't work period on Radeon hardware. It's not simply performance the image quality might be suffering as well with Nvidia hardware being utilized.

 

[Recus]

Possibly is or isn't or needs more adjusting. I do agree though FSR should be tested on Radeon hardware for obvious reasons. If it works at all on Nvidia that's great, but that's not how it should be tested ideally especially when DLSS just doesn't work period on Radeon hardware. It's not simply performance the image quality might be suffering as well with Nvidia hardware being utilized.

Despite pushing "FSR is better" narrative it's simply not true. And claiming that issues are because of Nvidia GPU is simple damage control. I thought AMD stuff works equally well on all brands. Those DLSS not capable users thanking AMD for FSR must feel pretty stupid because of very bad image quality. lol

 

[konga]

I do not believe there is anything about the FSR 2.0 algorithm that would negatively affect image quality on one GPU but not another. The way it looks is the way it looks, and the only real difference between GPUs will be execution speed.

Intel is claiming that XeSS may have better image quality on Intel GPUs because it will use a more optimized pipeline on those to take advantage of their special matrix whatever cores, but FSR 2.0 is run entirely on the shaders, with the same code across all GPUs.

 

[Sithaer]

I was so excited at first because I went from 60-80fps @ 4k to well over 100fps. The longer I played the game though, the worse the problems got. Tons of artifacts, chunks of screen going one solid color, flickering in and out… then whole objects like the vending machines or the vault stopped rendering.

The game is pretty buggy to begin with, but I never had graphical issues before.

Yup, also played it a bit since I noticed this update and ran into various issues.
For example this boss area has TONS of weird stuff going on with FSR Quality:

Native no FSR image:

FSR Quality:

That puple water thing does have a shiny surface by default but FSR makes it sparkle and artifact all over the place and its even worse when you move the camera around while looking at it.

Seems like a bad implementation or something.

 

[zlobby]

Slightly(?) out of topic. How do we pronounce Tiny Tina?

IDK but I always looks for the Feds over my shoulder after saying it.

 

[Punkenjoy]

I do not believe there is anything about the FSR 2.0 algorithm that would negatively affect image quality on one GPU but not another. The way it looks is the way it looks, and the only real difference between GPUs will be execution speed.

Intel is claiming that XeSS may have better image quality on Intel GPUs because it will use a more optimized pipeline on those to take advantage of their special matrix whatever cores, but FSR 2.0 is run entirely on the shaders, with the same code across all GPUs.

FSR 2.0 is similar to XeSS in the sense that RDNA GPU have a different code path than others cards. Many people complained that they did saw such bad image quality on AMD and after testing it, DF found that image is slightly better. They also acknowledged that they searched super hard to get to a place where this behavior happen (the fuzzy character moving.)


But there, i do not think that it's what causing FSR 2.0 issue here. I suspect they do not have set the Mipmap Bias correctly and the cable we see are texture instead of polygon.

Normally, you get a different level of Mipmap per resolution to get a smooth and clean texture. A texture with too high resolution too far away or in a too low resolution will look very pixelated.

When you want to upscale an image, you want to run the mipmap bias of the resolution you want to upscale or higher so you can extract the pixel you wouldn't see at lower resolution. This one of the way FSR 2.0 and DLSS extract sub pixel details.

You clearly see in the screenshot they selected that the black scratch is disappearing with FSR and that shouldn't happen if the Mipmap level would have been setup properly.

If the cable are texture instead of polygon, they would have similar effect.

And if they missed that, who know what they didn't do well too. Doing Temporal upscaling require more work than spatial and if you don't do the prerequisite first, you will have image quality problem.

 

[ModEl4]

FSR 2.0 is similar to XeSS in the sense that RDNA GPU have a different code path than others cards.

I had the impression that FSR 2.0 is shader based using HLSL, isn't?
Maybe you had in mind the 6.2 or 6.6 usage depending the hardware?
If you meant that, this isn't exactly in the same sense different use of code path as XeSS's case, they just use 6.6 on some GPUs that support 64-wide wavefronts and when it makes sense performance wise. (One of the speculations for RDNA3 is related to wavefronts throughput, suggesting that FSR2.0 will have less performance penalty (millisecond wise) vs RDNA2 which along with other potential improvements will result supposedly in RDNA3 to have less performance penalty (millisecond wise) vs at least Ampere/RDNA2 based GPUs regarding FSR 2.0 implementations.

 

[tancabean]

FSR 2.0 is similar to XeSS in the sense that RDNA GPU have a different code path than others cards.

Im pretty sure FSR runs the same DirectX shader on all hardware. There is no special hardware or special path for RDNA.

 

[PrettyKitten800]

Yup, also played it a bit since I noticed this update and ran into various issues.
For example this boss area has TONS of weird stuff going on with FSR Quality:

Native no FSR image:
View attachment 252313

FSR Quality:

View attachment 252314

That puple water thing does have a shiny surface by default but FSR makes it sparkle and artifact all over the place and its even worse when you move the camera around while looking at it.

Seems like a bad implementation or something.

It definitely needs some fixing. They implemented it really quickly, so I’m sure it’s going to get better overtime

 

[nguyen]

Oh so FSR2.0 is supposedly only work properly on RDNA GPU, would that make it more or less "proprietary" and "niche"? since RDNA1/2 take up like 1-2% of the GPU marketshare while RTX GPU takes up ~30% (according to Steam)

 

[Punkenjoy]

Im pretty sure FSR runs the same DirectX shader on all hardware. There is no special hardware or special path for RDNA.

After reading again the thread that stated the rendering difference, it was pointed out that it would be true in DX12 but the game would use vendor libraries in DX11 to run the same code and there could be a difference there. Still more digging need to be done but it look like at least on God of wars that there are slight difference between AMD and Nvidia.

Or maybe not, after some digging, people found that people actually compare the video stream and not the actual image. They were not able to reproduce the issue with screenshot but the timing is very hard to do.

 

[LiveOrDie]

Let's hope Unreal Engine 4 & 5 implementation to be better.
Joking aside, it seems the FSR 2.0 implementation procedure to require more effort vs DLSS 2.0 in order to hit optimal results but I may just reading the situation wrong.

I have FSR 2.0 in my UE5 game on steam and DLSS still looks better, Its just not worth using even looking at those images you can see the blurriness added.

FSR 2.0 is similar to XeSS in the sense that RDNA GPU have a different code path than others cards. Many people complained that they did saw such bad image quality on AMD and after testing it, DF found that image is slightly better. They also acknowledged that they searched super hard to get to a place where this behavior happen (the fuzzy character moving.)


But there, i do not think that it's what causing FSR 2.0 issue here. I suspect they do not have set the Mipmap Bias correctly and the cable we see are texture instead of polygon.

Normally, you get a different level of Mipmap per resolution to get a smooth and clean texture. A texture with too high resolution too far away or in a too low resolution will look very pixelated.

When you want to upscale an image, you want to run the mipmap bias of the resolution you want to upscale or higher so you can extract the pixel you wouldn't see at lower resolution. This one of the way FSR 2.0 and DLSS extract sub pixel details.

You clearly see in the screenshot they selected that the black scratch is disappearing with FSR and that shouldn't happen if the Mipmap level would have been setup properly.
View attachment 252349

If the cable are texture instead of polygon, they would have similar effect.

And if they missed that, who know what they didn't do well too. Doing Temporal upscaling require more work than spatial and if you don't do the prerequisite first, you will have image quality problem.

After testing FSR 2.0 on my own game it has nothing to do with Mips, FSR is forcing lower screen percentage and failing to upscale it end of story.

 

[Dredi]

end of story.

Sure, we can’t prove anything about your game. Though in the case of tiny tina there are clear indicators for wrong mipmap bias (or some other) setup (pop in happens at different distance compared to native).

 

[Bomby569]

"In terms of performance, the FSR 2.0 performance uplift at 4K resolution is a great improvement to the game even in Quality mode, but 1440p and 1080p have issues."

Well since 4k is where you need it, I would say that one is the most important, followed by 1440 than 1080.

hey, there is a lot of people with very tight budgets and poor rigs that are the ones that benefit the most from this. Play in 1440p you can barely fell the difference. Now imagine people that have can leave 720p behind for 1080p over this, the improvement is much bigger

 

[mechtech]

hey, there is a lot of people with very tight budgets and poor rigs that are the ones that benefit the most from this. Play in 1440p you can barely fell the difference. Now imagine people that have can leave 720p behind for 1080p over this, the improvement is much bigger

Possibly, but tight budget and/or poor rigs probably wouldn't have RDNA cards. But one can also adjust graphics settings to gain fps. Some settings like fog, etc. can drastically improve fps while only having minimal impact on graphics quality. I had to "optimize" the graphics settings on Borderlands 3 to get 60 fps on 1440p.

 

[Bomby569]

Possibly, but tight budget and/or poor rigs probably wouldn't have RDNA cards. But one can also adjust graphics settings to gain fps. Some settings like fog, etc. can drastically improve fps while only having minimal impact on graphics quality. I had to "optimize" the graphics settings on Borderlands 3 to get 60 fps on 1440p.

FSR is not for RDNA cards only.

 

[mechtech]

FSR is not for RDNA cards only.

Correct. But most of the cards I would not classify as poor or tight budget. (well maybe that RX 590 )

FSR/DLSS are a good thing though, and hopefully, game implementation and card support will grow and get more optimized, so like you said, people with lesser gaming PC's will be able to get better quality graphics with addition fps.