OpenXR Guide - Deprecated - This time for real (▀̿Ĺ̯▀̿ ̿)

[Ghostman]

I tried this I was running NIS at 80% before Sharpening at 35 and 100% in openxr tools with MR off was getting 45-50fps with around +30 CPU headroom which I don't know what that means or gives lol.

So I tried NIS at 60% and Sharping at 10 and 200% in openxr tools with MR off was getting 50-55 fps with around +35 CPU headroom.

with the same with MR always on I was getting 43-50 fps with +50 to +55 CPU headrooom.  

all this on my 3090ti

[VR Flight Guy in PJ Pants]

I cannot help to feel the irony -- OpenCompsite supposed to allow games to be played without SteamVR, but now, for Pico4, the only way to run OpenComposite DCS is starting DCS inside SteamVR!

[plane_crazy242]

this is the best thing since the DCS implemented multi-thread processing!!!  Seriously, this SEEMS SO MUCH SMOTHER.
I'm getting 43-44  FPS with setting on High and it seems so smooth, 65-69 FPS on VR setting,  (5800X3D, 64GB, 3080),
Damn, this is worth it just for the non stuttering splash screen

Now for my newbie questions

1. can you dock the FPS and clock so it doesn't follow my view around?
2. should I uninstall steamVR or does it not matter?
3. not a DCS questions, but will openVR run with Il-2 and MSFS the same way with the same nice improvements?
4. how do you land?

 

[Splashtrick68]

46 minutes ago, plane_crazy242 said:

this is the best thing since the DCS implemented multi-thread processing!!!  Seriously, this SEEMS SO MUCH SMOTHER.
I'm getting 43-44  FPS with setting on High and it seems so smooth, 65-69 FPS on VR setting,  (5800X3D, 64GB, 3080),
Damn, this is worth it just for the non stuttering splash screen

Now for my newbie questions

1. can you dock the FPS and clock so it doesn't follow my view around?
2. should I uninstall steamVR or does it not matter?
3. not a DCS questions, but will openVR run with Il-2 and MSFS the same way with the same nice improvements?
4. how do you land?

 

If you mean the display for the companion tool.. once I dialed in my settings, I've turned it off. You can hold CTRL+ALT and the down arrow (by default/unless you've changed) and tab thru some options there. I also have kept my FPS locked to 60 for now as my experience is really smooth (once in awhile a jitter when viewing cockpit, but I'm happy). I have Reverb G2 v2 and 5800x 32gb and 4080 and only time I see framerate drop a bit is on the deck around other planes.. I have most all settings on high as well. You are right though, this is THE BEST thing I've experienced for VR so far.. it's dope! 

Once you set your selection in the composite app, you can permanently select Steam VR for IL2, I actually haven't and not sure what it runs on when opening in VR.. IL2 was always a bit more optimized imo for VR, I had a good experience with that in VR with a 3070.. not sure as to why. I don't know if you can use openXR/VR for it but I guess can try? I've landed the same so far, I actually am having a harder time now playing in 2d since this as my VR experience is so solid.

Not sure if I helped with anything, that's my experience so far. I hope others find this as well. I would also assume that when ED implements additional cores for gameplay, VR will come WAY up and this may be less of a thing or need?

[speed-of-heat]

OK some data for people to play with, all the tests were conducted with a stabilised head position and against the same performance track and settings otherwise, I am not advocating these particular settings, no sharpening was set, the numbers represent the average from the performance track using the in built tools, they should be used to understand the relative performance benefits of each case. 75% was used to get an exact match to the OXR Tools setting.

Test	FPS	a CPU(ms)	rCPU(ms)	GPU (ms)	Notes
27	75.89	8.75	6.17	12.50	Resolution Baseline 3168 x 3096
28	85.45	7.87	5.92	11.05	Resolution Test 1  OXR Tools Render Scale at 75%
29	88.17	7.60	5.94	9.83	Resolution Test 2  NIS per eye 2376 x 2322 aka 75%
30	87.37	7.70	5.96	9.89	Resolution Test 3  FSR per eye 2376 x 2322 aka 75%
31	90.01	7.13	5.84	9.79	Resolution Test 4 OXRTK Resolution override per eye 2376 x 2322

However it becomes clear if you don't want any sharpening , using the resolution override has the best overall performance, I am not making a qualitative judgement, as my head wasn't inside the headset when these tests were run.

Edited December 8, 2022 by speed-of-heat

[edmuss]

15 minutes ago, speed-of-heat said:

OK some data for people to play with, all the tests were conducted with a stabilised head position and against the same performance track and settings otherwise, I am not advocating these particular settings, no sharpening was set, the numbers represent the average from the performance track using the in built tools, they should be used to understand the relative performance benefits of each case. 75% was used to get an exact match to the OXR Tools setting.

Test	FPS	a CPU(ms)	rCPU(ms)	GPU (ms)	Notes
27	75.89	8.75	6.17	12.50	Resolution Baseline 3168 x 3096
28	85.45	7.87	5.92	11.05	Resolution Test 1  OXR Tools Render Scale at 75%
29	88.17	7.60	5.94	9.83	Resolution Test 2  NIS per eye 2376 x 2322 aka 75%
30	87.37	7.70	5.96	9.89	Resolution Test 3  FSR per eye 2376 x 2322 aka 75%
31	90.01	7.13	5.84	9.79	Resolution Test 4 OXRTK Resolution override per eye 2376 x 2322

However it becomes clear if you don't want any sharpening , using the resolution override has the best overall performance, I am not making a qualitative judgement, as my head wasn't inside the headset when these tests were run.

 

It might be worthwhile adding the sclGPU (I think it is) values on there, this emcompasses the overheads associated with upscaling; the higher the resolution the more overhead you need.  I've seen typically values between 0.8 and 1.5ms additional computational time.

 

[speed-of-heat]

3 minutes ago, edmuss said:

It might be worthwhile adding the sclGPU (I think it is) values on there, this emcompasses the overheads associated with upscaling; the higher the resolution the more overhead you need.  I've seen typically values between 0.8 and 1.5ms additional computational time.

 

It doesnt record to the log file as far as I can see, which means no reliable way of getting the data.

[edmuss]

Ah I didn't realise that one

It's possible that it might be incorporated into the GPU frametime?

[speed-of-heat]

I assume it is... Though the major difference seems to be the app cpu time vs the render cpu time

[Tshark]

SoH are you still using OXRTK reprojection for game play?  I too could not believe the smoothness flying the Mariana's in a UH-1 with it on.  If they could get the reprojection anomalies reduced just a little bit more it would be perfect.

[speed-of-heat]

19 minutes ago, Tshark said:

SoH are you still using OXRTK reprojection for game play?  I too could not believe the smoothness flying the Mariana's in a UH-1 with it on.  If they could get the reprojection anomalies reduced just a little bit more it would be perfect.

I go back and forth on this at the moment between using reprojection at 100-90% and 90hz and at 2650 x 2589 @ 60hz without reprojection, which works everywhere but the Mariana's, at the moment i'm in 60hz phase... they both have their ...defects... 

Edited December 8, 2022 by speed-of-heat

[mbucchia]

12 hours ago, edmuss said:

Ah I didn't realise that one

It's possible that it might be incorporated into the GPU frametime?

It is not. OpenXR Toolkit app stats do not include OpenXR Toolkit overhead. SCL GPU is typically very small and it explains your 2-3 FPS difference between NIS/FSR/Override.

The rest of your numbers are mostly "within noise band", 500us APP CPU difference is surprising but it is likely inaccuracies from test setup and test repeatability. There is no reason for NIS/FSR to interfere with APP stats. Also you are clearly GPU-bound so that difference in CPU numbers make no difference in terms of FPS. 

[Thomas_LOLW]

On 12/7/2022 at 4:25 AM, Glide said:

Make sure WMR is set to Best Visual Quality.

I think this settings is only relevant for the Windows Mixed Reality Home Screen...correct me if I am wrong.

[speed-of-heat]

2 hours ago, mbucchia said:

It is not. OpenXR Toolkit app stats do not include OpenXR Toolkit overhead. SCL GPU is typically very small and it explains your 2-3 FPS difference between NIS/FSR/Override.

The rest of your numbers are mostly "within noise band", 500us APP CPU difference is surprising but it is likely inaccuracies from test setup and test repeatability. There is no reason for NIS/FSR to interfere with APP stats. Also you are clearly GPU-bound so that difference in CPU numbers make no difference in terms of FPS. 

Thanks for the explanation

[speed-of-heat]

6 hours ago, mbucchia said:

It is not. OpenXR Toolkit app stats do not include OpenXR Toolkit overhead. SCL GPU is typically very small and it explains your 2-3 FPS difference between NIS/FSR/Override.

The rest of your numbers are mostly "within noise band", 500us APP CPU difference is surprising but it is likely inaccuracies from test setup and test repeatability. There is no reason for NIS/FSR to interfere with APP stats. Also you are clearly GPU-bound so that difference in CPU numbers make no difference in terms of FPS. 

I wanted to check the CPU FT time variance and it's by no means exhaustive but it is fairly replicable; FWIW:

Test	FPS	app CPU(ms)	render CPU(ms)	GPU (ms)	Notes
27	75.89	8.75	6.17	12.50	Resolution Baseline 3168 x 3096
28	85.45	7.87	5.92	11.05	Resolution Test 1  OXR Tools Render Scale at 75%
29	88.17	7.60	5.94	9.83	Resolution Test 2  NIS per eye 2376 x 2322 aka 75%
30	87.37	7.70	5.96	9.89	Resolution Test 3  FSR per eye 2376 x 2322 aka 75%
31	90.01	7.13	5.84	9.79	Resolution Test 4 OXRTK Resolution override per eye 2376 x 2322
32	90.02	7.04	5.77	9.88	Resolution Test 5 OXRTK Resolution override per eye 2376 x 2322 (RETEST)
33	89.91	6.83	5.65	9.92	Resolution Test 6 OXRTK Resolution override per eye 2376 x 2322 (RETEST)
34	88.08	7.50	5.91	9.83	Resolution Test 7 NIS per eye 2376 x 2322 aka 75% (RETEST)
35	88.13	7.41	5.85	9.85	Resolution Test 8 NIS per eye 2376 x 2322 aka 75% (RETEST)

the delta seems consistent about .5ms  between Scaling and Override, as this is across separate days (and there is no AI in the loop)...there is no head movement (the HMD is stabilised, I'm not wearing it) I don't believe (though i havent ruled it out) there is anything environmental that would trigger that delta. 

Edited December 9, 2022 by speed-of-heat

[Silver_]

I wanted to tell you about a personal experience that surely happens to more than one. When I activate MR I can't see certain objects clearly, such as the edges of the cockpit or the tips of the wings (distortion effect such as water, blur), the FPS meter is not clearly visible either (when I use it, of course) or the MFD's (in the case of the F16 Viper). All this has forced me to have to deactivate MR even in spite of having gained FPS's thanks to the trick of my partner @chichowalker (in a post above). Thanks Chico.

 

I noticed that if I leave MR off and on within the OpenXR toolkit in the Inputs tab, Shaking Reduction and set it to -30% (in my case) I get more smoothness when doing fast head turns inside the cockpit. I tried other smaller values like -20 or -10 but the one that fits the best, goes better in sync, with my system (CPU, GPU, etc) is -30%. For me it is a remedy for the small graphic jittering that occurs when you turn your head quickly and have MR disabled, as in my case. I hope this helps someone. All the best.

[D4n]

On 3/17/2022 at 10:33 AM, nikoel said:

The connection between Windows Mixed Reality portal and SteamVR is a performance bottleneck.

Then why would ED not make WMR use OpenXR for DCS instead of SteamVR? It's shocking to find out about this incredibly bad SteamVR bottleneck!!!

[TKu]

They simply used supported technology that was available at that time.

 

[dburne]

Pretty sure ED has mentioned they are looking at maybe adding native Open XR support. That is what needs to be done.

[markturner1960]

.

Edited December 9, 2022 by markturner1960

[diamond26]

11 hours ago, speed-of-heat said:

I wanted to check the CPU FT time variance and it's by no means exhaustive but it is fairly replicable; FWIW:

Test	FPS	app CPU(ms)	render CPU(ms)	GPU (ms)	Notes
27	75.89	8.75	6.17	12.50	Resolution Baseline 3168 x 3096
28	85.45	7.87	5.92	11.05	Resolution Test 1  OXR Tools Render Scale at 75%
29	88.17	7.60	5.94	9.83	Resolution Test 2  NIS per eye 2376 x 2322 aka 75%
30	87.37	7.70	5.96	9.89	Resolution Test 3  FSR per eye 2376 x 2322 aka 75%
31	90.01	7.13	5.84	9.79	Resolution Test 4 OXRTK Resolution override per eye 2376 x 2322
32	90.02	7.04	5.77	9.88	Resolution Test 5 OXRTK Resolution override per eye 2376 x 2322 (RETEST)
33	89.91	6.83	5.65	9.92	Resolution Test 6 OXRTK Resolution override per eye 2376 x 2322 (RETEST)
34	88.08	7.50	5.91	9.83	Resolution Test 7 NIS per eye 2376 x 2322 aka 75% (RETEST)
35	88.13	7.41	5.85	9.85	Resolution Test 8 NIS per eye 2376 x 2322 aka 75% (RETEST)

the delta seems consistent about .5ms  between Scaling and Override, as this is across separate days (and there is no AI in the loop)...there is no head movement (the HMD is stabilised, I'm not wearing it) I don't believe (though i havent ruled it out) there is anything environmental that would trigger that delta. 

 

For what is worth I have similar experience with Quest2 using OpenXR toolkit in the sense that override option behaves better than scaling for the exact same resolutions.

[mbucchia]

40 minutes ago, diamond26 said:

For what is worth I have similar experience with Quest2 using OpenXR toolkit in the sense that override option behaves better than scaling for the exact same resolutions.

That part I have to admit I am super confused about :D, not exactly sure how it could be different.

[dutchili]

Installing this week's release of nvidia driver worked wonders for performance with oc2 and rtx3070.

Running oc2 on openxr runtime without opencomposite, pd 1.0 and oc2 upscaling set to 1.5 (max). 36 fps @72hz in simple missions is near perfect.

 

never got the optimization mentioned here to work. Any short instruction good oc2?

my issue is that i don't get to see the menu after installing opencomposite and manually copying the dll to DCs/bin

 

[Antix70]

For the last few updates, I can't get OpenComposite to do any switching, DCS just forces SteamVR.

The ONLY thing that works is replacing the Openvr API dll.

[D4n]

How to close menus like these when they open while playing DCS? (without WMR controller, since I ordered my WMR goggles without controllers)