Via a tweet, CapFrameX reveals they saw some footage of FSR 2.0. The new technology does not require AI acceleration, it's said to be a temporal upscaling technique with optimized anti-aliasing. It supports GPUs from all vendors and allegedly delivers impressive performance and image quality. It will be interesting to see if it's indeed "better than native."
The announcement of FSR 2.0 was not expected till at least RDNA3 architecture release, now planned for Q3/Q4 based on the leaks. However, with no requirement for AI acceleration, the technology might be supported by shader cores, possibly using similar instruction like Intel XeSS (DP4a).
FSR 2.0 to come/announced soon. Saw some footage today. Very impressive. ????— CapFrameX (@CapFrameX) March 11, 2022
?? Temporal upscaling + optimized AA
?? Does not need AI
?? Runs on GPUs from all vendors
?? Impressive performance and image quality
AMD even claims that it can be better than native. ^^#FSRGen2