AMD FSR 4.1.2 Modded to Run on Intel Arc GPUs Leveraging XMX Cores

erek

8-D
2FA
Joined
Dec 19, 2005
Messages
18,728
"DarkerThanLpDark on Reddit put FSR 4.1.2 Lite on Arc B580 to the test in "Clair Obscur: Expedition 33," comparing its performance with the game's native XeSS implementation, and FSR's standard DP4a implementation on Arc GPUs. The latest OptiScaler mod was found offering almost the same levels of performance as XeSS (SR). FSR 4.1.2 Lite logged an upscaling time of 0.59 ms compared to 0.58 ms of XeSS (SR)."

1792846807640.png


Source: https://www.techpowerup.com/354099/...to-run-on-intel-arc-gpus-leveraging-xmx-cores
 
Also worth noting (from the Reddit post linked by TechPowerUp): B580 1080P, all Upscalers on Balanced Mode in Clair Obscur Expedition 33

I was wondering how times on AMD hardware compared so I tried digging around the net a bit. There are pretty big differences between AMD generations. RX 6000 doesn't have AI accelerator cores. RX 7000 adds them with INT8 support, but no FP8. RX 9000 adds FP8 and AMD claims 4x AI compute. Of course that doesn't apply to memory bound jobs. So far this is what I've dug up. Looks like RDNA4 is a lot faster than 3 at running FSR4, at least comparing the performance of the 9070XT and 7900XTX upscaling to 4k. I'm not sure if AMD's numbers are at all comparable, but if they are a B580 is a lot quicker at int8 FSR4 upscaling than an RX 7600. The 9070XT numbers suggest linear scaling, so a B580 might out-upscale a 7800XT as well. Then if I just wave my hands and pretend a 9060XT is half as fast as a 9070XT because it has half the cores & mem bus the 9060XT would beat the 7900XTX. I guess this kind of explains why AMD made 3 different versions of FSR4.

I think this mod is going to get me to finally try out Optiscaler. I have a 5090 in my main rig and don't have much of anything that I care about running on the B580 that needs upscaling and doesn't support XeSS. FSR4 and the resulting image quality upgrade changes that equation. Next I guess we'll see what Intel answers with. I have this hunch that XeSS is intentionally built to be faster and lower quality than FSR4 & DLSS. For Intel it's all about the iGPUs. They're weaker than dGPUs, so Intel needs fast more than quality. FSR4 Lite is the old card version, but it's also the iGPU version.
 
Back
Top