News XeSS SDK 2.1 release opens up Intel's framegen tech to compatible AMD and Nvidia GPUs — Xe Low Latency also goes cross-platform if framegen is enabled

Eventually everyone will settle on a standard that works everywhere. Both FSR and XeSS are getting more open with XeSS being the more "open" of the two, which is hilarious cause Intel.
Upscaling, frame-gen, and low-latency all need to be tightly woven into the graphics API's -- DirectX, Vulkan, and Metal. From there, if the vendors want to have some differentiated features based on their AI software and hardware, that's fine -- I say go for. For example, that might be to reduce overhead on shaders, so performance aspects could vary across vendors but quality levels would generally be the same across the board.

And yes, funny that Intel are so open on this technology, though it's out of needs: to encourage adoption by game devs.
 
Upscaling, frame-gen, and low-latency all need to be tightly woven into the graphics API's -- DirectX, Vulkan, and Metal. From there, if the vendors want to have some differentiated features based on their AI software and hardware, that's fine -- I say go for. For example, that might be to reduce overhead on shaders, so performance aspects could vary across vendors but quality levels would generally be the same across the board.

And yes, funny that Intel are so open on this technology, though it's out of needs: to encourage adoption by game devs.


Pretty much every "modern" graphics feature started off as some sort of vendor proprietary item that eventually morphed into an API standard. Modern texture compression started off as a proprietary feature of S3's MeTaL API that was present in the S3 Savage 3D cards. I even had one way back in the day. So I have no doubt that upscaling / etc.. techniques based on AI algorithms will be eventually have a vendor neutral standard.
 
Pretty much every "modern" graphics feature started off as some sort of vendor proprietary item that eventually morphed into an API standard. Modern texture compression started off as a proprietary feature of S3's MeTaL API that was present in the S3 Savage 3D cards. I even had one way back in the day. So I have no doubt that upscaling / etc.. techniques based on AI algorithms will be eventually have a vendor neutral standard.
Yes this transition is fundamental. Everything get off as vendor proprietary as it gives head start advantage that paid off RD investment. Then when the technology becomes saturated when everyone else has something of similar it will quickly become a hassle for users and developer alike, which prompts some vendors to consolidate their tech into standard and everyone quickly switch over to that.