DCS VR Performance Optimization

[Ready]

8 minutes ago, muzica9 said:

With you specs, can not get stable 60 fps with Frame Rate Throttling in open xr? You still have to apply that junk settings of 60hz and destroy your eyes??!! Why??!! 

 

 

Yes unfortunately. Without reprojection this is the best I can get out of it.

[muzica9]

i run at much, much higher settings. clouds at ultra(can't accept less for vr), open xr tlk turbo mode+Frame Rate Throttling 61 fps+Foveated rendering - quality /wide

all stable at 60 fps no matter scenario(heavy or low), in ST or MT, at 90hz for G2

only difference in specs: CPU I9 10850k and 48 gb ram

btw: i keep my 4090 at 80% power limit as i see no difference in terms of perfomance for DCS or in general.

[A.F.]

If I set it to 90 Hz and then throttle the frames to 60 Hz, I see slight double images in a curve during flight. The picture then appears blurred. This does not happen when I set the G2 to 60 Hz.

[muzica9]

you can not throttle the frames to 60 Hz. maybe 60 fps. i only see similar if motion reprojection is on, wich i don't use. 

[Pikey]

I do beleive that the MT patch made most of this thread dated, i used a very heavy mission to test and get 55FPS with non flat shadows but otherwise the same as above on a 4080. The only config I use outside of mostly high settings is of course the OpenXR tools NIS SS set to 120%. Bottom line is that I don't feel the need to change it much whereas when the thread was created, VR was way less optimised and it probably was a harder chase.

 

[Ready]

2 hours ago, muzica9 said:

i run at much, much higher settings. clouds at ultra(can't accept less for vr), open xr tlk turbo mode+Frame Rate Throttling 61 fps+Foveated rendering - quality /wide

all stable at 60 fps no matter scenario(heavy or low), in ST or MT, at 90hz for G2

only difference in specs: CPU I9 10850k and 48 gb ram

btw: i keep my 4090 at 80% power limit as i see no difference in terms of perfomance for DCS or in general.

90Hz in the headset and frames limited at 60fps gives me bad stutters. Don't understand why this could ever be smooth. Because whenever below headset frequency there is some stutter. @edmuss could you please maybe inject some knowledge?

My reference is our Multiplayer server, where I am CPU-limited.

If you have a trackfile for me to run and compare, I am interested to see the actual difference in performance. I ran at full settings as well (cloud, visibility, shadows all at max), but turned things all the way down to slowly start upping them again where I think I need them. I am not using foveated rendering because it looks crap on my videos. I am always recording in 4k as well, which takes up some resources. I see you have clutter/grass enabled, visibly I see no difference between ON/OFF.

Will see if I can limit my GPU power. Haven't done that yet. 

Edited March 19, 2023 by Ready

[Ready]

1 hour ago, Pikey said:

I do beleive that the MT patch made most of this thread dated, i used a very heavy mission to test and get 55FPS with non flat shadows but otherwise the same as above on a 4080. The only config I use outside of mostly high settings is of course the OpenXR tools NIS SS set to 120%. Bottom line is that I don't feel the need to change it much whereas when the thread was created, VR was way less optimised and it probably was a harder chase.

 

I agree things got better with the MT patch and the shadow fix in the latest OB-update. I will be tinkering with settings everytime a new update comes around there is stuff to work on. Be it a driver update, DCS update etc. I do notice a shift in balance between tinkering and actual flying, which is fantastic.

 

[edmuss]

2 hours ago, muzica9 said:

i run at much, much higher settings. clouds at ultra(can't accept less for vr), open xr tlk turbo mode+Frame Rate Throttling 61 fps+Foveated rendering - quality /wide

all stable at 60 fps no matter scenario(heavy or low), in ST or MT, at 90hz for G2

You have completely missed the point of running the G2 at 60Hz, it means that you can easily achieve refresh rate VR (as it's meant to be) which is perfectly smooth with zero ghosting, artifacts or stutter.

90Hz with a 60fps framerate limit is below refresh rate so you will have the stutters and ghosting associated with it.

Refresh rate flicker aside, 60Hz is identical image to 90Hz.

[Tusk.V]

Yes unfortunately. Without reprojection this is the best I can get out of it.

I’m shocked that you have to run settings this low with 60hz. I’m at 60hz and I have everything cranked all the way up. I wonder if there’s some underlying issue….

[Ready]

10 minutes ago, Tusk.V said:

I’m shocked that you have to run settings this low with 60hz. I’m at 60hz and I have everything cranked all the way up. I wonder if there’s some underlying issue….

Yes could be. That's what my wife also says all the time.  

I will run my track file and post the outcome. Haven't done that in a while.

[muzica9]

41 minutes ago, edmuss said:

You have completely missed the point of running the G2 at 60Hz, it means that you can easily achieve refresh rate VR (as it's meant to be) which is perfectly smooth with zero ghosting, artifacts or stutter.

90Hz with a 60fps framerate limit is below refresh rate so you will have the stutters and ghosting associated with it.

Refresh rate flicker aside, 60Hz is identical image to 90Hz.

no stutters and no ghosting in my case. at least i don't see something wrong or something to bother me. but, 60hz for g2 is strobe disco club. with or without virtual sunglasses

Edited March 19, 2023 by muzica9

[Ready]

Ran a couple of track files again. 

Tomorrow will run it with my previous settings at 90Hz. 

 

Edited March 19, 2023 by Ready

[Pikey]

Whilst its a good checklist, when I did this type of activity a couple of years back the setting that made the most difference to performance, was by far,
DRUMROLL:
 

Spoiler

the trackfile.

Two people can't have the same discussion until they use the same one.

[Ready]

2 hours ago, Pikey said:

Whilst its a good checklist, when I did this type of activity a couple of years back the setting that made the most difference to performance, was by far,
DRUMROLL:
 

  Reveal hidden contents

the trackfile.

Two people can't have the same discussion until they use the same one.

Agreed. I have been using the same track file for 100+ tests so far. My difficulty is running a trackfile that accounts for our server's busy stuff. This is currently messing with my performance on the ground, but this clears up when airborne

[Pikey]

Aye, and MP is weirdly different because the CPU usage pattern changes when the units are all pushed down the pipe, but it can impact differently depending on what your CPU and network is like and how many units are being exported. I remember discussing with other users the ability to have "network tracks" but such a thing doesnt exist and is likely not a useful metric so we have to guess how a large mission is handled as a client most of the time.

The track we used offline gave me 22 FPS (average) in single threading, 42 in Multithreading and that was on a 2080. This is using some low textures and medium view distance, but otherwise reasonable with 130% SS.  I upgraded and spent a fortune on a 4080 and I can turn everything up to max and still get 55+ fps on this track, which is a monster. So in VR my experiences have gone from borderline unplayable to butter smooth with a new card and multithreading. Its basically a new start, other baselines are out of the window now.

And like you said, get out there and enjoy it before some new feature swallows the headroom
 

[Ready]

10 minutes ago, Pikey said:

Aye, and MP is weirdly different because the CPU usage pattern changes when the units are all pushed down the pipe, but it can impact differently depending on what your CPU and network is like and how many units are being exported. I remember discussing with other users the ability to have "network tracks" but such a thing doesnt exist and is likely not a useful metric so we have to guess how a large mission is handled as a client most of the time.

The track we used offline gave me 22 FPS (average) in single threading, 42 in Multithreading and that was on a 2080. This is using some low textures and medium view distance, but otherwise reasonable with 130% SS.  I upgraded and spent a fortune on a 4080 and I can turn everything up to max and still get 55+ fps on this track, which is a monster. So in VR my experiences have gone from borderline unplayable to butter smooth with a new card and multithreading. Its basically a new start, other baselines are out of the window now.

And like you said, get out there and enjoy it before some new feature swallows the headroom
 

Mind sharing your track file? Can run it to see how my system compares.

[Pikey]

Sure! We've been permitted to share, its nothing 'special'. Do you have a way of creating a performance graph? I've seen other folks come up with these but I didn't look into it. One day maybe there will be a common benchmarking system, eh?

Casacus test track 1.trk

[Ready]

36 minutes ago, Pikey said:

Sure! We've been permitted to share, its nothing 'special'. Do you have a way of creating a performance graph? I've seen other folks come up with these but I didn't look into it. One day maybe there will be a common benchmarking system, eh?

Casacus test track 1.trk 1.3 MB · 0 downloads

When I run it I sit in a hornet going straight and level, overflying a bunch of boats and then some land. Nothing happens. Is that the plan of this track file?

[Pikey]

When you say 'nothing happens', you mean why is the subject plane not making sudden movements or shooting or being hit by the weapons?
So perhaps the limitations of tracks and benchmarking aren't obvious, but you cannot standardise where there is too much movement else desync means one patch to the next the plane is in the ground or a missile misses. Any change in the FM or weapons data would not affect the outcome. So the limitations are that you generally make only very slight movements and don't have any missile strikes else rng makes it pointless. The extent of the track file is that the unit variation and numbers and types and what is flying around is still going on and everyone will get the same (as best as you can within the limitations of track replay). There are mixtures of many things being crammed in,as would be in a typically complex mission (pause the track and check f10). Hope that helps.

I downloaded FRAPS free edition (god bless it last update a decade ago) The benchmarking tool allows csv generation for frametimes - https://fraps.com/download.php

i then downloaded https://sourceforge.net/projects/frafsbenchview/ (free GNU license) which allows you to drag and drop those csv's and show two graphs. I attached those. I didn't do it clean I was downloading but just an idea. of how to standardise. My settings pulled direct from my OB logs are:

 

2023-03-20 23:06:27.535 INFO    APP (Main): options.graphics =
{
    ['messagesFontScale'] = 1.5;
    ['forestDetailsFactor'] = 1;
    ['rainDroplets'] = true;
    ['LensEffects'] = 0;
    ['box_mouse_cursor'] = false;
    ['anisotropy'] = 4;
    ['water'] = 2;
    ['motionBlurAmount'] = 1;
    ['BlurFlatShadows'] = 1;
    ['outputGamma'] = 1.6;
    ['aspect'] = 1.7777777777778;
    ['lights'] = 2;
    ['shadows'] = 3;
    ['MSAA'] = 1;
    ['SSAA'] = 0;
    ['ScreenshotExt'] = 'jpg';
    ['clutterMaxDistance'] = 1470;
    ['cockpitGI'] = 1;
    ['terrainTextures'] = 'max';
    ['multiMonitorSetup'] = '1camera';
    ['defaultFOV'] = 78;
    ['shadowTree'] = false;
    ['chimneySmokeDensity'] = 0;
    ['fullScreen'] = true;
    ['preloadRadius'] = 49900;
    ['width'] = 3840;
    ['visibRange'] = 'Ultra';
    ['DOF'] = 0;
    ['clouds'] = 3;
    ['sceneryDetailsFactor'] = 1;
    ['useDeferredShading'] = 1;
    ['flatTerrainShadows'] = 0;
    ['textures'] = 2;
    ['secondaryShadows'] = 1;
    ['SSLR'] = 0;
    ['effects'] = 3;
    ['SSAO'] = 0;
    ['scaleGui'] = 1.5;
    ['sync'] = false;
    ['motionBlur'] = 0;
    ['height'] = 2160;
    ['heatBlr'] = 0;
    ['forestDistanceFactor'] = 1;
    ['civTraffic'] = '';
};

 

Edited March 21, 2023 by Pikey

[Ready]

16 hours ago, Pikey said:

When you say 'nothing happens', you mean why is the subject plane not making sudden movements or shooting or being hit by the weapons?

Yes. No idea why nothing happens.

Thanks for the information, will give those apps a try. 

[Ready]

OpenXR Toolkit Frame Rate Throttling

So I did a couple of tests yesterday. I put the hog in a wheel, overflying a city in Syria just over the rooftops, looking over the wing towards the houses a couple hundred feet below. G2 set to 60Hz (I know), OBS set to 60FPS. the DCS status bar is spikey but Inside the headset visuals are smooth; I see the houses flashing by crisp and clear. When I enable frame throttling and bring it down to 60FPS, the DCS status bar smooothens out, but the houses start stuttering. So throttling has a negative effect in this case. Turbo mode has no effect in either case.

Stutter in OBS

The strange thing is that the OBS recording of these tests stutter, no matter what I do. I tried OBS recording in simple mode on both high quality and indistinguishable quality. I have no idea how to get rid of the stutters. OBS is running as admin. I have no render or encoding lag. It seems that there is some kind of data loss between the headset and OBS. could that be the case? This was a very clear test case to see the difference in visual quality between the headset and the recording. Still to try some further messing around with OBS: Advanced OBS recording settings; CBR - bitrate 60.000 - encoder NVENC H.264 instead of HEVC, preset Quality.

I will look into this some more. I ordered another SSD for my OBS recordings. So I have 3 SSD's, one for Windows, one for DCS, one for OBS. I don't know if this helps. ¯\_(ツ)_/¯

Edited March 26, 2023 by Ready

[edmuss]

Turbo mode will only help to recover FPS on WMR when close to or below refresh rate appGPU (16.6ms). If you're below this frametime it will make no noticeable difference.

Try locking the framerate to 61 rather than 60, certainly my reverb sits at 61fps when above refresh.

Can't help on the the OBS stuff though.

[Ready]

1 hour ago, edmuss said:

Turbo mode will only help to recover FPS on WMR when close to or below refresh rate appGPU (16.6ms). If you're below this frametime it will make no noticeable difference.

Try locking the framerate to 61 rather than 60, certainly my reverb sits at 61fps when above refresh.

Can't help on the the OBS stuff though.

Tnx will try

[muzica9]

try also this mod. i got good results

 

[Wali763]

Maybe a stupid question:

Are many with a G2 (or other 90-Hz-HMDs) running DCS in 90Hz? Especially with MT and considering 4090-owners for sure have fast CPUs, this should be easily doable.

Yet when browsing this thread it seems, many people even with 4090s are below 90fps.

Do they prefer to use higher settings instead of fluid 90fps?