mbucchia

Beware about the term "standalone" which in the industry mostly refers to "app running on the headset (as opposed to PC)". If you are using Quad-Views-Foveated, then I assume you are using PimaxXR - which is an OpenXR implementation that "ditches SteamVR". If you want to keep using SteamVR, you may simply not use PimaxXR (toggle back to SteamVR in the PimaxXR Control Center app). However if you do that, you will need to install an additional program in order to do eye tracking with SteamVR (because SteamVR does not support eye tracking): Home · mbucchia/OpenXR-Eye-Trackers Wiki (github.com) So for short, if you want to keep using SteamVR AND also benefit from Quad-Views-Foveated, you need to: 1) Disable PimaxXR and toggle back to SteamVR 2) Have Quad-Views-Foveated installed 3) Have OpenXR-Eye-Trackers installed

Heads up for folks using Quad-Views-Foveated on Pro - there is an upcoming update that will enable Virtual Desktop brand new eye tracking support. I'll post about it when it's out.

That's just a backend driver, every headset has that too. Some get installed via Steam and are more visible (WMR, Vive, Pico), some are installed as part of the relevant platform (Oculus, Varjo, Pimax) so it looks like they don't exist, but they do!! In any case, it's a mandatory driver layer, SteamVR cannot do anything without it, regardless of whether it's WMR or not. In the case of WMR, that WMR for SteamVR driver is a simple bridge to WMR lower-level interface (aka Windows.Graphics.Holographic). At its best, its pretty lightweight (if the game does not submit any 2D UI for example) however at its worst it has to perform additional composition (if there are 2D UI or if using motion reprojection). But that is unavoidable, even the OpenXR for WMR layer has to do that.

The CPU performance issues can only be solved by ED in their game engine, for example through the use of quad instanced rendering (submit geometry for all 4 views at once). This isn't an easy undertaking, so I doubt it will happen any time soon, or ever.

You do realize that the "additional program" in your story is SteamVR. WMR, Oculus, Varjo... can all be used without SteamVR though the use of OpenXR, and that cuts off the "additional program". But you still need to have the base system running (Mixed Reality Portal, Oculus Link, Varjo Base). There isn't really a way around that. Some base systems are more lightweight than others, but in all situations, SteamVR is the extra overhead (though the definition of "extra" is actually up to interpretation, since it provides some of the features that the base system doesn't offer).

The experiment you described only proves that the PimaxXR setting is toggling your OpenXR runtime, which is the goal. I can guarantee you that OpenComposite settings makes no difference in your situation, because your game isn't trying to use OpenVR anyway. You seem to have no grasp of what OpenComposite actually does, so it's a good thing you're deleting it...

That's all you need really: this will remove the reference to SteamVR in your `%LocalAppData%\openvr\openvrpaths.vrpath` which is the only leftover. Just to be clear the toggle in PimaxXR does something 100% different than what OpenComposite does. They're not at all the same thing.

That doesn't change my statement. Make sure you are using the latest version of everything, the initial version of Quad-Views-Foveated (months ago) had lower performance with OBSMirror (but even then it wasn't anywhere near what you described). I've had at least 5 reports of Quad-Views-Foveated + OpenXR-OBSMirror working impeccably with DCS since.

This happens if you don't properly set it up via PowerShell (depends on Windows settings). Definitely a problem on your end. It's been working fine for many users, including with Quad Views.

Use https://github.com/Jabbah/OpenXR-Layer-OBSMirror which will capture the output of Quad-Views-Foveated and therefore the same exact image you see in the headset.

And the ONLY WAY to have it PROPERLY done is by having the game implementing it. So please redirect your requests to ED. That hand-to-motion-controller hack is terrible, impossible to setup, very uncomfortable to use, and that's why I abandoned it over 18 months ago...

Great to hear this is solved.

The MT version only supports OpenXR, so if you need `--force_OpenXR` then it further proves that somehow your game is running ST. The cursor isses only happen with ST, this has been confirmed many times, and every time users having issues eventually found out they run ST. You can be sure by looking for the watermark in the menu (see below). Unfortunately there is not much I can do, this is a game issue. The method I gave you has worked for everyone (you must run the game from your Steam library and not from a shortcut like you said yoi are doing). The only consistent way I know today go use MT is to uninstall the Steam version and to use the standalone OpenBeta, which does not have these weird issues

Nope, nothing missing. This bit down here is not obvious and you have to read carefully! (Doing this solved the problem for the last ~5 people who had this issue, so I'm confident it will work for you too) BTW @BIGNEWY this caveat above has confused a lot of users. Not sure why things are that way, but apparently with the Steam edition, if you run the executable in bin-mt outside of Steam it still runs ST. This is very very confusing and could be improved.

Double cursor means you are not running DCS multithreaded. It's explained in the instructions: https://github.com/mbucchia/Varjo-Foveated/wiki#dcs-world. null

Yes those multipliers in the configuration file will take the resolution DCS was about to use. When you use OpenXR Toolkit Resolution Override, that setting sets that final resolution directly. One thing I never checked is what happened if you set Pixel Density to something other than 1.0 in DCS. Idk if anybody knows.

@BIGNEWY, any opportunity to pin this post so it keeps near the top? Thanks.

Hey all - There seems to be a lot of confusion still around Dynamic Foveated Rendering support and information is spread over multiple threads specific to each headset, with the same questions asked over and over again. So hopefully this will make things easier. I use a Varjo and I don't know/want to use OpenXR Toolkit or OpenXR OBSMirror You want to install a tool called Varjo-Foveated, FOLLOW ALL THE INSTRUCTIONS HERE: https://github.com/mbucchia/Varjo-Foveated/wiki I use a Pimax Crystal, Meta Quest Pro, Vive Pro Eye or HP Reverb G2 Omnicept or I use a Varjo and I want to also use OpenXR Toolkit or OpenXR OBSMirror You want to install a tool called Quad-Views-Foveated, FOLLOW ALL THE INSTRUCTIONS HERE: https://github.com/mbucchia/Quad-Views-Foveated/wiki/DCS-World I am already using Foveated Rendering but something does not work Go and read the instructions again. There is a 99% chance the answer to your question/issue is in the instructions. In the unlikely case you are in the 1% with a real issue that is not a misconfiguration or something overlooked while following the instructions, you may post a log file (from Quad-Views-Foveated) for support. I will ignore all requests made without a log file. I want to use Foveated Rendering but I do not know how to follow instructions Do not try using Foveated Rendering. There is some reading/understanding to do and if you cannot do that, you will not be successful. Enjoy the game and keep things simple. Please let's keep this thread on topic. No headset trolling. No unnecessary discussions about your CPU, GPU or your MSAA settings. No posts off-topic. Thanks.

Can someone report the issue to TallyMouse please? It's a very common bug with many C#/.NET! Thanks. It looks like TallyMouse would be @GayCavalier on this forum(?)

'Square withing Square" if because you did not follow the instructions. The website clearly says to disable Lens Effects in DCS https://github.com/mbucchia/Quad-Views-Foveated/wiki#dcs-world Please people RTFM!!

This update is HoloLens only and has no effect on PCVR: https://github.com/microsoft/OpenXR-MixedReality/releases/tag/113.2309.7002

Your CPU (not) being maxed out is not an indication of (not) being CPU-bound. I think I've been told that MT in DCS is just basic stuff like rendering being on its own thread. Now that thread has to do 2x more work for the 2 extra views. The Oculus software isn't involved at all in the process of rendering and doing quad views, so it's quite doubtful it would make a difference. Especially if you have Turbo Mode, the only place where Oculus tends to waste cycles is bypassed. The overhead of QVFR was measured (even in those situations where "performance gets worse") and confirmed to be negligible (under 200 microseconds/frame on both CPU and GPU). So the issue isn't there either. I've mostly heard these performance issues happen with specific aircrafts(?), probably they have more complex geometry? Or maybe some bogus vertex shaders? The only way to tell would be to do a precise performance analysis, which is quite a complex task involving a GPUView capture then analyzing each payload one by one and trying to map them to what the game engine does.

I see. I will say, you can't really compare PCVR to standalone though. Red Matter 2 does VRS foveated rendering (same thing OpenXR Toolkit does) but it is in standalone, with a much weaker GPU, so there are a lot more bottlenecks to address. I think several have confirmed that VRS on fast GPUs like 4090 is just not helping that much compared to lower-end RTX cards. I'm aware that a few users are only seeing their GPU bottleneck replaced by a CPU or vertex shading bottleneck with quad views, however it seems to be a very small minority of users. The only solution to this problem would be a better implementation of the rendering in the engine, such as use of instanced rendering (DCS does not do instanced rendering at all even for stereo AFAIK), usage of geometry shaders or better command list handling (requires D3D12 or Vulkan). It sounds however that ED has little interest in improving quad views support, as shown by the 5 different bugs that have been brought up to them since January, and none of them have been acknowledged or addressed, even the most basic ones like the blatant violation of OpenXR spec and the incorrect submission of FOV data in MT... these are like, 30 minutes each to produce a fix

Quad views foveated is giving up to double-digit frame times boost in GPU-limited scenarios by saving at several stages of the rendering pipeline (raster, pixel shading, overall VRAM bandwidth). The downside is increased CPU usage due to redundant geometry submission. Very confused by your post. What is your suggestion for this new technology? VRS foveated rendering is also possible but the gains are much much lower since only saving at the pixel shading stage (see OpenXR Toolkit foveated rendering, low single-digit frame time gains). It has only very little downside on the CPU side however. Dynamic LOD foveated rendering would be the only option to help with CPU and vertex shading, but can only be implemented directly in the engine. I don't think anyone has ever attempted it. There aren't really other techs available (or please suggest for my review).

Not clear if "hangs on the first splash screen" means in VR or not in VR (as in: did DCS successfully enter VR?). If in VR, then do you have Motion Reprojection enabled In OpenXR Tools for WMR? If so either: - Turn it off. Or - Toggle Enable Preview runtime.