No performance gain with FFR - how is that possible?
Hey there, I'm using Unity (2021.1.28f1, URP 11.0.0) to build (Vulkan, Multi-Pass) for Quest 2. I'm desperately trying to get more perfromance for the application I'm making, as it's not reaching a constant 72 fps, but am stuck around 60 fps. I switched to Vulkan to be able to use Fixed Foveated Rendering and was hoping to thus gain the ~20% framerate increase I need. Unfortunately even though I'm GPU bound, activating FFR does nothing in terms of framerate, it just leads to a (barely noticeable) reduction in visual fidelity. Even at the highest level 4, I do not notice any speed-up, yet according to the Oculus Quest documentation the device should only have to render less than half the pixels! In terms of performance, his should be equivalent to reducing the camera's render target scale quite a bit, which, as I tried (reducing from 1 to 0.9) leads to the desired framerate of 72 fps. Now, I wonder: what is FFR really doing? The graphics in the docs indicate that only 46% of pixels (I counted!) need to be rendered on FFR set to "High Top", so how can that not lead to a massive improvement in perfromance? Thankful for any information on how and when to use FFR. Cheers!
Fixed Foveated Rendering on Game Ports
I am somewhat involved in the Dr. Beef game ports (Quake, Quake 2, Half life, RTCW, Doom 3). I am trying to ascertain why it is not possible to switch on FFR, or rather it has no effect. For starters I have been setting the FFR level via ADB. I have been able to get this to work in Oculus home and other Apps (Unity etc) so know how to detect it (I use the highest setting). However in all of the ports there is no change at all. My understanding from reading the SDK documents is that this should just work. Is there something special in the rendering we need to enable for this to happen?
Decreasing the shimmering on the edges with fixed foveated rendering
The low resolution at the edges isn't distracting because of its low detail, but because of the shimmer it creates with movement. Are there any approaches to decrease the distinctiveness of this? Does the stereo mismatch between L/R views play a factor? Can it be decreased with an additional blur shader on device?
