@fulatoro These are always resolutions for 1 eye usually. The issue is the G2 is reporting too high a value for the OpenXR ārecommended view configuration sizeā. People have been complaining about this when using the G2 in SteamVR and HP is supposedly investigating it (these are values stored in the HMD non-volatile memory most likely and can be changed)
NB: Iām further explaining the reasons for the G2 higher recommended resolution here:
WMR Scaling and Dev Tools - Some Explanations
Nevertheless, what counts with my experiments with the Valve Index is looking at the W x H numbers final image will be viewed at, and this can differ from the W x H the image is rendered to. It can differ because it is changed in SteamVR SS resolution, and it can also differ because of the FS2020 render scaling:
| . | Reverb G2 | Valve Index | Notes |
|---|---|---|---|
| FS2020 Render | 2214 x 2162 (TAA 70%) | 1800 x 2004 (TAA 100%) | Steam SS 80% for Index |
| Send to OpenXR | 3164 x 3092 | 2016 x 2240 | Enlarge to OXR/SteanVR 100% (Recommended res.) |
| Process Image | 3164 x 3092 | 2016 x 2240 | Distort image for HMD lenses |
| Display Image | 2160 x 2160 | 1440 x 1600 | HMD native screen res. |
NB: It is expected rendering to a larger resolution because the image is distorted to compensate the lens curvature and in doing so, the center region is enlarged and the outer regions shrink (like a fisheye effect). If the rendering was to the native resolution, center would appear blurrier in the HMD. At 80% SS in SteamVR it is still good enough to me in the Index but it is clearer at SteamVR 100% in any case. Also, the simulator has to render to a higher resolution than native, especially with the Valve Index because not all the panel pixels are in view due to the lens shape and the position of the screens in the HMD. This is why headsets also have a ālens maskā which is used to save drawing pixels in these areas (the mask you see reported displaying 1/4 of the screen size - a bug - for people with Oculus headsets with FS2020). You can see these masks in the most comprehensive database Iāve found about this here: HMD Geometry Database | Collected geometry data from some commercially available VR headsets.
Example Reverb G1 (0.04 coverage):
Example Valve Index (24.88% coverage):
Having said this, what Iāve noted and experimented with is using TAA is causing the CPU to work a lot more and it is inducing as a consequence CPU/GPU sync problems. It is amazing because I get more fps at TAA 100% than TAA 80% for example just because of this overhead. But this effect is most likely due to something else Iāll try documenting today along with my update: my settings are setting FS2020 near the vertex processing limit and this is what is responsible for most cases of stutters when trying to make the CPU doing more. I believe I found 1 major opportunity to make a better VR experience (not necessarily more fps but less to no stutters) now that Iāve identified this bottleneck (or at least what appears to be a bottleneck). More about this later!