Support NeoGAF

[ClosBSAS]

ctrl + f DOTA 2....
nothing found.

keeps on moving.

 

[Kazekage1981]

has anyone tried FSR 2.2 ?

 

[Kataploom]

Didn't know FSR quality had made this big of a jump, it basically makes my new 6700xt into a full fledge 4K card

 

[winjer]

has anyone tried FSR 2.2 ?

It has only been implemented in FH5 and the last NFS.
I only tried FH5. It works okish, but because it has so much movement, it has trouble reconstructing some detail. DLSS2 has a similar problem, but to a lesser point.
The best solution for this game is native+TAA.

 

[Kazekage1981]

Didn't know FSR quality had made this big of a jump, it basically makes my new 6700xt into a full fledge 4K card

 

[01011001]

The best solution for this game is native+TAA.

*native+DLAA

 

[Fafalada]

As far as I can tell, FSR 2.x is well optimized. So it runs as well on AMD, NVidia or Intel.

It's usually good - but I've run into cases where it falls apart on AMD APUs.
On 6800U FH5 with FSR2.x actually runs slower than native unless you go down to Performance. Amusingly the same chipset runs Scorn @ 60hz on Quality (same target res as FH5) without breaking a sweat(and yes, much faster than native res too) - maybe it's a game problem or some driver conflict, but it was odd to see.

 

[winjer]

It's usually good - but I've run into cases where it falls apart on AMD APUs.
On 6800U FH5 with FSR2.x actually runs slower than native unless you go down to Performance. Amusingly the same chipset runs Scorn @ 60hz on Quality (same target res as FH5) without breaking a sweat(and yes, much faster than native res too) - maybe it's a game problem or some driver conflict, but it was odd to see.

That's not an issue with FSR 2.0, neither because of AMD hardware.
These upscaling techs, require some processing power, and that means that in low end hardware the gains from rendering at a lower resolution are not offset by the time it takes to process the upscaling.
Like most reviews showed when FSR2.0 was released, on low-mid end GPUs, from AMD, NVidia and Intel, the gains from FSR2 are smaller and smaller. An when we go low enough, it starts to negate performance.

 

[twilo99]

GPUs like the RX 580 and GTX 1060 will remain relevant for modern games as long the latest games keep using this tech. High on life for example, should have been released with it.

No, those GPUs should burn in hell along with the xbox series s ... they are holding back game development. We would 100% have star citizen by now if it wasn't for those horrible things.

 

[ACESHIGH]

No, those GPUs should burn in hell along with the xbox series s ... they are holding back game development. We would 100% have star citizen by now if it wasn't for those horrible things.

We would have 100% star citizen if devs learnt to manage project scope.

 

[Fafalada]

These upscaling techs, require some processing power, and that means that in low end hardware the gains from rendering at a lower resolution are not offset by the time it takes to process the upscaling.

Sure but that's not what's happening in example I gave.
You have different games on same exact chipset with wildly different gains/loses with the same setup (FSR@Quality - Scorn is 30-40% faster, and
FH5 20% slower than native)- we're clearly not looking at a simple '+fixed cost' equation.

In fact majority of games I tried with FSR2 run much better on this chip - even cases where I tried the DLSS injection path. FH5 is one of the rare outliers that juts do something differently.

 

[winjer]

Sure but that's not what's happening in example I gave.
You have different games on same exact chipset with wildly different gains/loses with the same setup (FSR@Quality - Scorn is 30-40% faster, and
FH5 20% slower than native)- we're clearly not looking at a simple '+fixed cost' equation.

In fact majority of games I tried with FSR2 run much better on this chip - even cases where I tried the DLSS injection path. FH5 is one of the rare outliers that juts do something differently.

You are probably hitting a memory bandwidth limitation.
That is an issue with all SoCs.

 

[Fafalada]

That still doesn't quite explain it - for it to manifest in one game but not the other it would mean bandwidth contention with something else is a problem - not usage from FSR alone.
But when base-rendering completes the frame 5ms early (and frames are not CPU limited either) - for FSR to spend all that time to bring it back to the cost of native, it would be contending with... what exactly?
It seems more like a type of API issue that creates hard-sync points where there shouldn't be any, in some scenarios. Might be misuse of the API on that particular game - or driver problem, but it doesn't have the signs of a direct hardware bottleneck as such.

 

[ToTTenTranz]

And for fsr 2.1 guality is so big leap from 1.0 that im rly interested what 3.0 could bring .

FSR3 is probably just frame generation / reprojection like DLSS3. It should be a nice boost if you're starting at high framerates (above 45-50 FPS?), but if you're trying to generate frames from a 15-20FPS baseline then the input latency might be too big.
TLDR it's good for very expensive GPUs on >100Hz monitors (or VR headsets, which is where the tech originates from), but it won't do wonders on lower-power GPUs.

On 6800U FH5 with FSR2.x actually runs slower than native unless you go down to Performance.

This makes no practical sense and it's probably a bug.
FSR2 brings me significant framerate boosts in Cyberpunk and Dying Light 2 on my Steam Deck, which has significantly lower compute throughput and memory bandwidth than a 6800U. Not to mention the fact that the Deck's CPU is running Proton to translate DirectX routines into Vulkan.

 

[winjer]

FidelityFX Super Resolution 2.2 Available on GPUOpen

AMD has released AMD FidelityFX Super Resolution (FSR) 2.2, an advanced cross-platform temporal upscaling technology that is accessible on GPUOpen.com. Game developers can now access the open-source A...

www.guru3d.com

This means we'll get more games with support for it. And also the dll for modded games.