Hi,
You might be wondering about the WMR Dev Tools and the resolution the WMR headset is using. This article is trying to explain all this as well as the caveats you might run into, and some strategies to get the most of your hardware.
Introduction
In short, WMR OpenXR is automatically scaling down the HMD recommended resolution based on your video card VRam amount. Not only you’re not rendering into the HMD at its optimum resolution, but you don’t have any indication WMR OpenXR is degrading your experience.
For those unfamiliar with this, please understand because of the nature of the lenses and the way the VR system is pre-distorting the image rendered, you have to render at a different resolution than the HMD panels.
Peter Peterson, working at HP VR, is saying it is about 1.5x larger for the G2 (3164x3092 in WMR OpenXR):
Definitive Answer for the 100% Resolution Discussion [reddit]
Now look at what the OpenXR runtime is changing for your FS2020 experience with the Reverb G2:
Using WMR OXR 103.2011.17003
| FS2020 | Render Res. | Post Process Res. | Comments |
|---|---|---|---|
| TAA 60% | 1898x1855 | 3164x3092 | breathtaking visuals and crisp enough EFIS at ~30 fps |
Using WMR OXR 104.2012.9003 (Auto Scale)
| FS2020 | Render Res. | Post Process Res. | Comments |
|---|---|---|---|
| TAA 70% | 1733x1694 | 2476x2420 | smaller render res., fuzzy visuals at 30 fps |
| TAA 80% | 1980x1936 | 2476x2420 | higher render res. yet fuzzy visuals at 28 fps |
With WMR SS Auto (default), even if you’re rendering more pixels, because they are displaying at a non-optimal HMD resolution, the end result as seen in the HMD is fuzzier (I can assure you it is and those of you reading “My VR Settings” discussion can see all the rationale and explanation I’ve shared about my experimentations).
You can you see with the lower HMD recommended resolution due to WMR OpenXR SS Auto (Post Process Res. in FS2020) and the 10-by-10 stepping for FS2020 Render Res, you can never match the “ideal” render resolution you would have configured your settings for.
Using OpenXR Dev Tools
With the latest updates, WMR is now giving the equivalent setting to SteamVR Super Sampling, and because the G2 in particular is requiring 1.5 times as many pixels as the panel for optimum rendering quality, you can take advantage of the CAS Shader vs Rendering Resolution optimal balance for your hardware.
The performance gain (per displayed pixels) using TAA 60 is higher with a higher Post-Process Res because the ratio between the CAS Shader process pixels vs the Rendering Shader pixels is higher and quadratic. In effect, it takes a tiny amount of pixel shader time to upscale the image than to render it at the upscale res. For example in terms of pixel count with the Reverb G2:
| WMR OpenXR | Render Res. | Post Process Res. | Comments |
|---|---|---|---|
| TAA 60 + SS 100% | 3520790 pixels | 9783088 pixels | 36% pixels rendered, 62% pixels extrapolated |
| TAA 80 + SS Auto | 3833280 pixels | 5991920 pixels | 64% pixels rendered, 36% pixels extrapolated |
In “My VR Settings” discussion, I’ve posted the rationale and the practical use of choosing a combination of both TAA Render Scaling (Render Res.) and SteamVR SS (Post Process Res.) for the Valve Index. In effect, because FS2020 is using the CAS Shader to upscale the rendered image into the displayed view, and because this CAS Shader takes less resources than the rendering Shader(s) cost, this is helping getting more fps, while keeping sharp visuals both up-close (EFIS) and in the distance (scenery). I find TAA 60% upscaled to 3164x3092 with CAS is giving better details in the outside view than TAA 80% upscaled to 2476x2420 for example.
What is affecting legibility (far and close)
With FS2020, there are mostly 2 factors only:
- EFIS screens resolution depends on the actual render resolution and you must balance the Render Res. and the Post Process Res. between EFIS legibility, outside view details and fps.
- CAS Shader savings are optimal at 60 to 70%. Below this it is loosing details and above this it is not saving as much fps.
For example, Valve Index TAA 60 + SteamVR SS 220 is the best I could get for both EFIS legibility and outside view at a distance. Reverb G2 TAA 60 + WMR 100 gives near equivalent resolution but thanks to the finer pixels in the G2 panels you can read better while leaning a little farther back, and gives finer details at a distance (I can see the PHNL runways texture and shape 25 miles out in the G2 for example, not details, but the different texture color, whereas it is more a pixel soup with the Index until much closer).
In general there is no need to increase beyond TAA 70 in both Airliners and GA. With anything above you’re rendering more than the native G2 panels and although the VR processing is counter distorting the rendered view in order to compensate for the lenses distortion, therefore transforming the view into a fisheye view requiring more pixels in the center in theory, I don’t find raising TAA above 70 worth the fps cost when comparing the legibility difference in the headset in practice.
I hope this will help you better understanding this subject and how to turn this to your advantage!
