AMD Announces FidelityFX Super Resolution (FSR), its DLSS-rival

[btarunr]

AMD finally made a big announcement on its ambitious FidelityFX Super Resolution (FSR) technology, the company's rival to NVIDIA's popular DLSS. Much like it, FSR aims to significantly improve gaming performance with minimal loss in image quality, through a sophisticated supersampling algorithm. At this point, AMD did not detail the nuts and bolts of the feature, but mentioned how the feature could look to gamers.

There are four FSR presets typically available to a supported game—Ultra Quality, Quality, Balanced, and Performance, which AMD claims offer performance gains of 59% for "Ultra Quality," 102% for "Quality," 153% for "Balanced," and 206% for "Performance." These should come particularly handy when playing games with raytracing on; and were measured on "Godfall" with RX 6800 XT, with 4K "epic" preset, and raytracing enabled. As of now, the company is working with over 10 game studios and game engine developers to integrate FSR, and the technology is expected to support "over 100 CPUs and GPUs."



Update Jun 22nd: We have now posted our in-depth review of AMD Radeon FidelityFX Super Resolution (FSR).

The FSR slide-deck follows.


View at TechPowerUp Main Site

 

[wolf]

I am very eager to see and try this myself, AMD really needs this to succeed, and at face value, they seem fairly confident here in what they plan to deliver. I am also very keen to see a Digital Foundry analysis and comparison.

 

[Space Lynx]

I am very eager to see and try this myself, AMD really needs this to succeed, and at face value, they seem fairly confident here in what they plan to deliver. I am also very keen to see a Digital Foundry analysis and comparison.

look at this screenshot. FSR is dead on arrival, look how blurry the ground looks. DLSS 2.0 auto wins if this is all AMD has to showcase....

 

[wolf]

I also wonder about this portion..

and CPU's... is it because some AMD APU's will support it? or is it run (even in part) on the CPU, so only certain ones/above a certain threshold of cores/speed etc will support it?

I am really interested to understand what is going on under the hood.

 

[ViperXTR]

had to see real samples, stream videos cant show of real image quality

 

[nguyen]

look at this screenshot. FSR is dead on arrival, look how blurry the ground looks. DLSS 2.0 auto wins if this is all AMD has to showcase....

View attachment 202353

I was about to laugh at how badly FSR look on the 1060 too LOL, the right side looks like 1080p image with CAS, everything just lose so much details and look blurry AF.

 

[Tomorrow]

look at this screenshot. FSR is dead on arrival, look how blurry the ground looks. DLSS 2.0 auto wins if this is all AMD has to showcase....

View attachment 202353

Too early to say for sure. First the resolution is 1440p. This means that source image is likely 1080p. Also it uses the second best "Quality" mode not the highest "Ultra Quality" mode. And it's just one game so the implementation may not be the best.

The same way DLSS is far less impressive when running on 1440p with "Performance" mode when compared to 4K in "Quality" mode.
And as always it's best to wait for independent testing. Hope AMD provides early access to testers before the June 22 release date.

 

[Space Lynx]

I was about to laugh at how badly FSR look on the 1060 too LOL, the right side looks like 1080p image with CAS, everything just lose so much details and look blurry AF.

Nvidia wins this round, I'll still look at reviews when it launches and side by side images of other games, but that image alone being their showcase image does not inspire confidence...

that being said, I still think it's amazing AMD can compete with a 3080 at the $649 price range in the year 2020... with solid stable drivers. to be honest you got to give them credit for that, their gpu division almost died after the 5700 driver issue drama. I actually tried to get a 3080 on launch day but failed. that was only reason i got the 6800. and i have been pleasantly surprised how stable its been. and how many frames it gives me. etc

Too early to say for sure. First the resolution is 1440p. This means that source image is likely 1080p. Also it uses the second best "Quality" mode not the highest "Ultra Quality" mode. And it's just one game so the implementation may not be the best.

The same way DLSS is far less impressive when running on 1440p with "Performance" mode when compared to 4K in "Quality" mode.
And as always it's best to wait for independent testing. Hope AMD provides early access to testers before the June 22 release date.

ya I'm not tossing it out or anything, just not hopeful. will be interesting to see reviews June 22nd, not far away now.

 

[Aretak]

look at this screenshot. FSR is dead on arrival, look how blurry the ground looks. DLSS 2.0 auto wins if this is all AMD has to showcase....

View attachment 202353

Both sides look like a blurry mess. There's clearly motion blur in progress too (look at the foliage on the left and you'll think you need glasses), so I assume this is a screen capture from some low bitrate stream. And a low resolution capture at that. Let's see some actual high quality screen captures and footage.

 

[Space Lynx]

Both sides look like a blurry mess. There's clearly motion blur in progress too (look at the foliage on the left and you'll think you need glasses), so I assume this is a screen capture from some low bitrate stream. And a low resolution capture at that. It tells us absolutely nothing, and only a deranged fanboy would be calling it "dead on arrival" based on this nonsense. Let's see some actual high quality screen captures and footage.

are you suggesting a company as professional and rich as AMD does not know how to give a presentation properly? and to put forth their best showing at said presentation as reputable as computex is known for?

your logic is failing here. if it was a mistake then it was, but I highly doubt AMD would have given less than 100% effort in showcasing their much debated and talked about FSR/DLSS software. they knew the community was excited for it. they wouldn't have half assed their presentation. again it may have been a mistake in which case you may be right... we will know June 22nd when more reviews come out.

 

[95Viper]

Keep on topic; and, do not insult other members.

Thank You

 

[Liquid Cool]

I also wonder about this portion..

View attachment 202354

and CPU's... is it because some AMD APU's will support it? or is it run (even in part) on the CPU, so only certain ones/above a certain threshold of cores/speed etc will support it?

I am really interested to understand what is going on under the hood.

Wolf,

I'm sure someone will come along and correct me if I'm wrong, but I do believe APU's were mentioned as being supported in the presentation.

Best,

Liquid Cool

 

[Space Lynx]

I also wonder about this portion..

View attachment 202354

and CPU's... is it because some AMD APU's will support it? or is it run (even in part) on the CPU, so only certain ones/above a certain threshold of cores/speed etc will support it?

I am really interested to understand what is going on under the hood.

I watched the whole thing. APU's are supported yep.

 

[Camm]

The 1060 comparison image is not kind, and reminds me of DLSS 1.0 in someways….

This being said, its important to note that AMD’s solution isn’t AI, and isn’t Temporal, its Spatial. Upside, there’s no frame delay so image persistence should be excellent. Downside, spatial upsampling usually doesn’t accumulate frame date (unless the algorithm is adaptive), so image quality is likely to be lower. The effect of this can be seen in DLSS in when you update the entire image (by spinning) and when you stop there being a momentary ‘snap’ to full quality as DLSS accumulates previous frame data.

FSR being available everywhere is likely the biggest thing that will kill DLSS off however, no need to spend time integrating one technology for one platform when you can build and ship once.

 

[Space Lynx]

FSR being available everywhere is likely the biggest thing that will kill DLSS off however, no need to spend time integrating one technology for one platform when you can build and ship once.

especially since the consoles are selling at record never before seen paces... developers will be putting their effort on those consoles first for all the AAA ports, which means FSR focus.

 

[nguyen]

The 1060 comparison image is not kind, and reminds me of DLSS 1.0 in someways….

This being said, its important to note that AMD’s solution isn’t AI, and isn’t Temporal, its Spatial. Upside, there’s no frame delay so image persistence should be excellent. Downside, spatial upsampling usually doesn’t accumulate frame date (unless the algorithm is adaptive), so image quality is likely to be lower. The effect of this can be seen in DLSS in when you update the entire image (by spinning) and when you stop there being a momentary ‘snap’ to full quality as DLSS accumulates previous frame data.

FSR being available everywhere is likely the biggest thing that will kill DLSS off however, no need to spend time integrating one technology for one platform when you can build and ship once.

More like FSR exist concurrently to DLSS where RTX owners get the benefit of higher visual fidelity at the same performance, much like FidelityFX CAS atm.
Would be pretty funny if FSR can be applied on top of DLSS, DLSS Quality + FSR Ultra Quality for some insane visual/performance uplift LOL.

 

[Camm]

More like FSR exist concurrently to DLSS where RTX owners get the benefit of higher visual fidelity at the same performance, much like FidelityFX CAS atm.

Nvidia will milk it for Gameworks titles, but I wouldn't expect all that many titles to be getting DLSS post 2022.

 

[nguyen]

Nvidia will milk it for Gameworks titles, but I wouldn't expect all that many titles to be getting DLSS post 2022.

DLSS is evolving everyday so I wouldn't be discounting DLSS just because some inferior upscaling tech just happen to be more popular....That's like saying luxury car are going extinct because cheaper cars are more popular LMAO

 

[Mussels]

Look how DLSS launched - full expect this to be a blurry mess for a year or two, while helping out APU and weak GPU owners (laptops especially)


This is a huge win for all gamers, having options that let us get away with weaker hardware

 

[wolf]

This is what Anandtech article is saying.

And, to drop into op-ed mode, this is where AMD has me a bit worried. In our pre-briefing with AMD, the company did confirm that FSR is going to be a purely spatial upscaling technology; it will operate on a frame-by-frame basis, without taking into account motion data (motion vectors) from the game itself.

For GPU junkies, many of you will recognize this as a similar strategy to how NVIDIA designed DLSS 1.0, which was all about spatial upscaling by using pre-trained, game-specific neural network models. DLSS 1.0 was ultimately a failure – it couldn’t consistently produce acceptable results and temporal artifacting was all too common. It wasn’t until NVIDIA introduced DLSS 2.0, a significantly expanded version of the technology that integrated motion vector data (essentially creating Temporal AA on steroids), that they finally got DLSS as we know it in working order.

Given NVIDIA’s experience with spatial-only upscaling, I’m concerned that AMD is going to repeat NVIDIA’s early mistakes. Spatial is a lot easier to do on the backend – and requires a lot less work from developers – but the lack of motion vector data presents some challenges. In particular, motion vectors are the traditional solution to countering temporal artifacting in TAA/DLSS, which is what ensures that there are no frame-by-frame oddities or other rendering errors from moving objects. Which is not to say that spatial-only upscaling can’t work, only that, if it’s competitive in image quality with DLSS, that would be a big first for AMD.

 

[Space Lynx]

This is what Anandtech article is saying.

I feel like AMD would be smart enough to not make such a similar recent mistake, so maybe they have an ace up their sleeve? We will see June 22nd... just a waiting game now really. 3 week wait isn't bad at all though. I'm expecting DLSS 2.0 to win handily though. but hopefully AMD did just make a mistake with those Godfall screens...

 

[nguyen]

I feel like AMD would be smart enough to not make such a similar recent mistake, so maybe they have an ace up their sleeve? We will see June 22nd... just a waiting game now really. 3 week wait isn't bad at all though. I'm expecting DLSS 2.0 to win handily though. but hopefully AMD did just make a mistake with those Godfall screens...

Yeah let hope FSR exist in some other game beside the shitty Godfall ,if AC Valhalla support FSR then it makes sense but Godfall, who the heck play this game?

 

[Camm]

DLSS is evolving everyday so I wouldn't be discounting DLSS just because some inferior upscaling tech just happen to be more popular....That's like saying luxury car are going extinct because cheaper cars are more popular LMAO

I think you mistake how technologies have been historically used and integrated.

1\ Most widely implemented for targets across multiple platforms (think Speedtree, Havoc, and some aspects of Gameworks).
2\ Available by default (think UE extensions)
3\ Paid to implement.

The 'quality' of the solution is oddly enough, not usually the deciding factor.

If Nvidia can reduce the burden on devs to almost nothing, we may see it on PC ports, but that will limit it to Unreal Engine titles almost exclusively (as don't expect large publishing houses to maintain branch compatibility specifically for PC when cross target alternatives exist).

 

[nguyen]

I think you mistake how technologies have been historically used and integrated.

1\ Most widely implemented for targets across multiple platforms (think Speedtree, Havoc, and some aspects of Gameworks).
2\ Available by default (think UE extensions)
3\ Paid to implement.

The 'quality' of the solution is oddly enough, not usually the deciding factor.

If Nvidia can reduce the burden on devs to almost nothing, we may see it on PC ports, but that will limit it to Unreal Engine titles almost exclusively (as don't expect large publishing houses to maintain branch compatibility specifically for PC when cross target alternatives exist).

As long as DLSS offer perceivable superior image quality at the same performance as FSR, that's enough reason for devs to support DLSS.
If FSR can't distinguish itself IQ wise from just simple upscaling + CAS (which DLSS 2.0 already proved itself) then FSR will probably join Radeon Boost into oblivion.

 

[Camm]

If FSR can't distinguish itself IQ wise from just simple upscaling + CAS then FSR will probably join Radeon Boost into oblivion LOL.

Of course, and if AMD is dumb enough to launch it at that quality, that's exactly what will happen.

But you are delusional that devs will spend the hours to implement a one vendor only for one platform upscaling tech if something close enough is available that supports all of their targets unless they are being paid to implement it or its literally a tickbox.