Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,376
Here is Nvidia explaining it:
https://developer.nvidia.com/sites/...works/vr/GameWorks_VR_2015_Final_handouts.pdf
Remember,this is only for VR headsets,and is taking advantage of the fact our periphery vision is not as good.
Its got nothing to do with our peripheral vision not being as good - thats foveated rendering which is different.
Imagine you want to take a flat piece of paper and make a dome shape from it, you would have to cut big chunks out of it, those chunks are completely wasted, they are not used at all. In VR, the GPU has to render the entire flat piece of paper at a higher resolution than the final "dome" and then lots of pixels are wasted, they arent down sampled like in antialiasing, they are completely cut from the final image.
Multiprojection (incorporating mulit resolution rendering) allows you to pre-distort the image so that some of those lost pixels are never rendered in the first place. The render target for the Oculus Rift is something like 2880x1550, about 1.6-1.7 times the number of pixels the actual displays have. Using this it can be cut down to about 1.25 times. Saving you about 30% performance cost.
Multiprojection also allows you to render both eyes simultaneously instead of rendering each eye sequentially and then stitchibg them together, which acounts for the other 20% they showed on stage.