Btw how the eye tracking gonna bring more performance boost with foveated rendering compared to more conventional foveated rendering in current psvr 1?
I was scratching my head over this as PSVR doesn't have eye tracing, but then I vaguely remembered some tech about three different zones of fidelity? Is that what you mean?
The human eye's actual area of high fidelity vision is about 31.5 arc seconds, half a degree. In a 90 degree FOV display, almost all of that cannot be perceived in any detail, only a tiny portion at the centre of your gaze. Ergo, you should be able to render to that tiny portion of the screen, a few hundred pixels square, and the rest can be rendered at super low res without the brain being able to notice, concentrating 90% of your GPU power to just the portion of the screen looked at instead of the whole screen. Without eye-tracking the best you can do is render parts of the display in different resolutions but if you limit yourself too much, the player can't move their gaze and would have to maintain a completely fixed forwards look for the whole of the game.
Best case comparison, PSVR fidelity zones might concentrate 75-85% of rendering power on 50% of the screen for a 1.5x improvement in graphics per viewed pixel as opposed to rendering uniformly across the TV. Foveated rendering can bring 90% GPU power to 1% of the display for a 90x improvement in graphics in the area viewed versus the same distributed over a full TV. Yes, 90x better! Depending on the quality of the tech at play that's an obtainable target in efficiency.
Why isn't that visible in the videos? Why aren't we seeing something akin to two generations' advance in graphics over PS5 in these PSVR2 clips? Several reasons. 1) No-one is writing games for this yet. 2) Some tech won't scale. eg. RT lighting BVH construction will maintain the same full-scene overhead and only the sampling will be faster as you are sampling a smaller area. At 120 fps target, you haven't much time to use RT so lighting will remain old-school. 3) Cost! Imagine a game made to target a conventional 50 TF of rendering power, squeezing the life out the SSD to get as much data in to make that look amazing. It's plain cost prohibitive. Content in VR games will be priced for the market which is pretty small. VR games at the moment are largely an enthusiast niche solution from indie devs at a lower price.
The tech is awesome, the efficiency gains spectacular, the potential staggering, but what we'll actually get is an unknown. However, putting the tech in the hands of the mainstream is a necessary move to get things going.