09-30-2021 12:28 PM
In general for Quest/Quest2 development is it better to render to larger eye textures and let the layer compositor sample or to render to device (recommended) sized eye textures with multi-sampling in the render pass?
10-01-2021 05:51 AM - edited 10-01-2021 05:59 AM
I think there's no clear-cut answer to this. Both approaches have their tradeoffs. E.g. MSAA should be cheaper/more memory efficient, but only works on geometry edges while Super-Sampling can increase sharpness generally. So it also depends on the content.
I can recommended this reddit comment for further information (still correct information imho, even if the post is a bit older).
As "general advice", there's also this tweet from carmack, that's relevant:
"If you have spare GPU performance in your mobile VR app and want it to look better, first enable chromatic aberration correction and anisotropic filtering before trying to raise the eye buffer resolution. Always 4x MSAA. Go sRGB correct if you can, but a retrofit can be hard."
- https://twitter.com/id_aa_carmack/status/980931811684507649