AMD FSR 2.0 Quality & Performance

[W1zzard]

We're reviewing AMD's new FSR 2.0 upscaling technology, which offers amazing image quality improvements that are able to match the visual finesse of NVIDIA's DLSS. The great thing is that the new FSR doesn't require any special hardware and works even on NVIDIA GeForce and Intel.

Show full review

 

[Denver]

Oh My God, it really looks great... Nvidia must be sweating blood seeing their solution that needs a dedicated ASIC, wasting precious space on the die, being matched via open source software. lol

 

[Al Chafai]

i love this part:
"I really have to applaud AMD for democratizing upscaling without additional hardware requirements, and all we need now is widespread game developer support."

 

[FeelinFroggy]

all we need now is widespread game developer support

This is the key. It's hard to get too excited about FSR or DLSS when so few games support it.

 

[harm9963]

Going AMD ,is compelling

 

[DeathtoGnomes]

Thank you for the review @W1zzard. It took a while for DLSS to gain traction with developers at first too. FSR1.0 had an easier time due to the ease of implementation and backwards compatibility if/when developers took the time to patch it in. I think FSR2.0 will be more quickly implemented than DLSS and FSR 1.0 was.

 

[Al Chafai]

This is the key. It's hard to get too excited about FSR or DLSS when so few games support it.

owh believe me this is going to be well supported for sure. especially when this technology can be used on current gen consoles like the PS5/Xbox SX/S as well.

 

[phanbuey]

DLSS 2.0 was basically the main reason I went with a 3080. This is huge.

 

[Bavor]

In some of the screenshots I noticed the same issue that FSR 1.0 had. Objects such as wires, fences, railings, handrails, etc... looked worse with FSR 2.0 than with DLSS. They looked like they were either missing pieces or had jagged edges with FSR 2.0 when the equivalent DLSS setting didn't have that issue. FSR 1.0 had the same issue.

 

[3211]

In some of the screenshots I noticed the same issue that FSR 1.0 had. Objects such as wires, fences, railings, handrails, etc... looked worse with FSR 2.0 than with DLSS. They looked like they were either missing pieces or had jagged edges with FSR 2.0 when the equivalent DLSS setting didn't have that issue. FSR 1.0 had the same issue.

Those are the limits. This is not as good as the article makes it out to be, from this one cherry picked game from amd. Once you will have fine objects in motion, hair on characters the artifacts will be even more apparent. Its good, but its no dlss killer of any kind. It looks worse, it runs worse and we'll see how the adoption is going to be with additional work required. DLSS is in like three times more games than the junk fsr 1.0 which should have been the easiest thing in the world to implement

 

[tabascosauz]

I'm all ears - but AMD needs to commit to this in the same way that Nvidia has been stubbornly stuck on DLSS like flies on a turd. As in, getting more devs onboard to expand the game support list (which Nvidia isn't done with either honestly), frequently releasing drop-in incremental improvements (e.g. DLSS 2.x .dll versions), and not treating this like a one-off science experiment as they have in the past (TressFX, standalone "FidelityFX sharpening").

For the longest time MW2019 was the only title but DLSS has been slowly creeping into my games library (War Thunder, Siege, No Man's Sky) - I hope FSR 2.0 will soon follow. Lazy/incompetent/overworked devs will still be the challenge - ie. DLSS "2.0" in War Thunder, sad abandoned excuse of DLSS in BFV, DLSS performance in BF2042, "FSR support" but Linux only in BL3...

 

[zlobby]

Another great, free, open technology by AMD.

TenZZor cores my aZZ!

 

[Xuper]

hmm , FSR 2.0 on 1080p is way too much sharp. I was surprised DLAA/DLSS quality has more blur detail than FSR 2.0 + sharpen ( 1080p )

Edit : this image slider is wrong.

 

[Al Chafai]

Those are the limits. This is not as good as the article makes it out to be, from this one cherry picked game from amd. Once you will have fine objects in motion, hair on characters the artifacts will be even more apparent. Its good, but its no dlss killer of any kind. It looks worse, it runs worse and we'll see how the adoption is going to be with additional work required. DLSS is in like three times more games than the junk fsr 1.0 which should have been the easiest thing in the world to implement

wooooooh so did you actually try this tech on other games? where did you come up with all these conclusion so fast? you sound like a hater not gonna lie....
during gameplay i am sure spotting the difference between the two will be so hard even for the most experienced. people are forgetting that these reconstruction techniques aren't perfect by any means, and for AMD to provide such improvements without the need for any proprietary hardware is a massive step forward.

 

[rutra80]

Well done. Now the only feature AMD is missing, is performant raytracing.

 

[ARF]

Well done. Now the only feature AMD is missing, is performant raytracing.

DLSS and RT are proprietary "features" by nvidia with no value for the user who can think.
The PS5 and new XBox do not support RT, so the gamers actually do not need it.

AMD's mistake is that it follows instead of thinking proactively about new unique features with real value.

 

[qlum]

Oh My God, it really looks great... Nvidia must be sweating blood seeing their solution that needs a dedicated ASIC, wasting precious space on the die, being matched via open source software. lol

IIt is kind of the other way around, the card had AI cores, DLSS was developed in an effort to find a use for them.

 

[AnarchoPrimitiv]

wooooooh so did you actually try this tech on other games? where did you come up with all these conclusion so fast? you sound like a hater not gonna lie....
during gameplay i am sure spotting the difference between the two will be so hard even for the most experienced. people are forgetting that these reconstruction techniques aren't perfect by any means, and for AMD to provide such improvements without the need for any proprietary hardware is a massive step forward.

Does sound like a hater, plus, everyone seems to forget that for all intents and purposes, Nvidia has limitless financial resources when compared to AMD, but expect AMD to not only compete, but to do better (while also not seeking profit in the same way as Nvidia....so many people think AMD should be a non-profit company and hold them to standards they hold nobody else to)....this is a great big step, and should only get better as long as Nvidia doesn't pull an intel and instead of innovating, just use vast amounts of money to box AMD out and get developers to be exclusive to Nvidia IP....which they will

 

[Nordic]

I really enjoyed the introduction to the technology. It was a great write up.

It is too bad that the minimum requirements are so high. I could really use this with my 1060 but it is below min spec for 1080p up scaling. Those who need it the most can't even use it.

 

[AnarchoPrimitiv]

DLSS and RT are proprietary "features" by nvidia with no value for the user who can think.
The PS5 and new XBox do not support RT, so the gamers actually do not need it.

AMD's mistake is that it follows instead of thinking proactively about new unique features with real value.

Again, compare the R&D budgets for Nvidia and AMD, Nvidia is at $5 billion, AMD is at $2 billion, and we KNOW more than half of that $2 billion is being spent on x86, so in reality, AMD has a fifth of the budget Nvidia has to compete....and still manages to do what they do, in terms of bang for the buck, AMD is the most impressive.

 

[ARF]

so many people think AMD should be a non-profit company

This is a radical opinion which doesn't reflect the reality. Probably you'd agree that there is a difference between "moderate" profit margin of 15%, and aggressive, crazy profit margin of 65%, yes or no?

 

[Chrispy_]

Still not sure if Tensor cores really do anything for DLSS; They were architected for AI workloads but none of the DLSS "AI" happens on your RTX card, it's performed by Nvidia on their large server farms and then baked into the game-ready driver as a preset.

 

[ARF]

Again, compare the R&D budgets for Nvidia and AMD, Nvidia is at $5 billion, AMD is at $2 billion, and we KNOW more than half of that $2 billion is being spent on x86, so in reality, AMD has a fifth of the budget Nvidia has to compete....and still manages to do what they do, in terms of bang for the buck, AMD is the most impressive.

AMD's mistake is that it answers these dirty initiatives by nvidia. Tessellation, and now RT... Do you remember when nvidia paid a game developer to REMOVE the DX 10.1 implementation (Assassin's Creed DX10.1) in which the Radeons were better?

 

[Prima.Vera]

We need a comparison test also with the upscaller that is now implemented by Epic in Unreal Engine 5.

 

[zx128k]

Basically as per the XeSS for Intel, XMX increases performance and allow more process which increases image quality. This is the same for DLSS and tensor cores. From Intel slides, XeSS on DP4 instructions runs on all GPUs but is both slower and has reduced image quality. One of the big issues with temporal upscaling is a lack of processing power leads to blurring. Also AI is better at removing artifacts and can make guess when there is a lack of samples further improving image quality.

In the Matrix Awakened demo DLSS quality mode appears to outperform TSR in UE5 by 10 fps. So I expect DLSS and XeSS to both outperform TSR and FSR 2 in both quality and performance with a 4k upscale. With DLSS having a fixed cost because the number of tensor cores. I expect the performance to be close at lower resolutions. The issue I would guess with FSR 2 is it wot upscale from 1080p to 4k because the quality wont match. Dynamic resolution scaling is going to be the best bet to keep quality close the nvidia DLSS. you samply increase the resolution or decrease it. Thus you can have better quality and maintain performance. With temporal upscaling there should be little need for sharpening.

In the article we can see that a 4k result requires a 1440p input for FSR 2 to upscale to 4k in quality mode. Like DLSS this is better than FSR 1. Internal resolution affects the number of rays needed in Ray Tracing games thus rendering at the lowest resolution posible is very important before upscaling. I can see that FSR 2 i about 5 fps behind, given that there is a performnce gap in RT games as well. This means AMD is cannot close the gap with NVidia.

Image quality is good in still shots, need to see raw video to see image quality while the image is moving. Thats when most of the artifacts happen. FSR 2 looks a little more blurred compared to DLSS but close. DLSS 2 does looks closer to native. Stills wont tell you much. Would not use FSR 1 but would use FSR 2. FSR 2 is the right balance, were FSR 1 was not.