Someone in reddit linked to this:
Interesting new features on the Turing Architecture. (From Whitepaper)
It is an analysis of the RTX Turing features by a VR game developer “vblanco”.
Of special note is the part about Variable Rate shading. In it he says:
Variable Rate shading: This is also a huge feature, but its biggest use is for VR. It allows the developer to change the “resolution” of parts of the screen at will. The fun part is that the internal images are unchanged, so its not only extremelly easy to implement, but it could be done as a driver level toggle. With this, nvidia could make VR games automatically render at a lower resolution on the edges of the screen, giving you beetween 20% and 40% extra performance without quality loss. For every single VR game currently in the market, without the developer doing anything. Im not completely sure nvidia will actually do that, but given how the technique works, its definitely possible. If not, its still basically a “toggle” a developer could add with barely any code. Looking at the feature, it seems i could implement it for my VR games in barely a day. Page 43
He also adds at the bottom:
The variable shading will make the new cards extremelly efficient for VR, and ready for foveated rendering at the driver level, without the developer doing barely anything.
He also talks about many other great VR features of the RTX cards, but lets focus on this first.
So what if he says is true, then foveated rendering at driver level can already become a reality today!
First step is Pimax should be able to use it as option to reduce GPU power used at peripherals. Maybe add as a setting to enable it in PiTool.
Then when the eye tracking module is released, and an eye tracking module is present, they can use it to dynamically adjust which areas to use more/less GPU power for true foveated rendering!
This could greatly improve performance, which would then allow higher supersampling which would then increase image quality!
It seems Variable Rate Shading can even be used to increase quality at specified area.
Anyways, if this is really possible then I hope Pixax implements it at driver level.