I had a thought this morning.
With current VR tools like ATW and ASW used to keep latency low, generate synthetic frames, and also used for things like sliced timewarps, etc. could we make use of old school scanline interleaving (the original 3DFX method of multi GPU support) without the negatives like microstutter and artifacting, but, most importantly, do so in a content agnostic way?
I was thinking, one GPU renders the even lines in a frame, the other GPU renders the odd lines in the frame, and at each step those halves of a frame are synced, post processed, etc. (with ATW and ASW) then sent to the IGPU (as a passthrough) to display on the HMD?
The HMD itself would be plugged into the IGPU, so there wouldn’t be a need to specifically implement multj GPU in an app. Could be done by Pimax at a driver level.