News AMD Promises Open Source Multi-Platform FidelityFX Super Resolution

[Admin]

AMD Promises Open Source Multi-Platform FidelityFX Super Resolution. How will the alternative to DLSS look, when will it be available, and will game developers use it? These are all important questions.

AMD Promises Open Source Multi-Platform FidelityFX Super Resolution : Read more

 

[splashmaker]

Based on how deep learning works, I don't think you'd want to combine it with CAS. The network should have visibility on all aspects of the image generation, CAS would be too basic and generate lower quality outputs compared to a DL model. Overall I think you might be able to get good quality without tensor acceleration but at a lower resolution or frame rate than with acceleration.

 

[VforV]

The bigger question is: will it work on the RX 5000 series? Because if it does that's a huge performance gain for RDNA 1, expanding it's life... and make the RX 5700 XT even more attractive vs the 2070 Super, than already is.

 

[FlxDrv]

The bigger question is: will it work on the RX 5000 series? Because if it does that's a huge performance gain for RDNA 1, expanding it's life... and make the RX 5700 XT even more attractive vs the 2070 Super, than already is.

Yes it will. Super Resolution will use directml to work and scince then new rdna2 have no hardware accelerated machine learning cores there is no reason that the can't enable the feature on older cards. You'll just need a directx12 gpu for it to work, so it may work all the way back to GCN1 (hd 7xxx) but rdna2 and rdna1 maybe getting it first and then older cards will get it after, if they need to adapt the feature to each architecture.
I'm pretty confident that at least rdna1 and rdna2 gpu's will get it in the next big adrenaline update that comes out at every end of year, so in the next few days, to maybe a week or two hopefully.