Jump to content

OpenXR Toolkit Tuning Guide (updated 21/02/23)


edmuss

Recommended Posts

3 minutes ago, TED said:

@nikoelgood demo videos of how it looks when all is tuned. Very similar to my experience lately. Out of interest what fps are you getting? Are u using fsr and  msaa? 
At the moment I’m using fsr and it seems to work well allowing me to fairly easily adjust depending on whether I’m going to do mp or sp. I’ve had no negative issues with it at all. 

 

I like to think we get good performance to compensate for us being helicopter pilots 😅

I'll be honest with you TED. I have no idea, after I saw the counter above 45FPS I forgot about it and started tuning to a very scientific "good feeling" (smooth, whilst also being realistic and accepting DCS is going to drop a frame or two every now and then). I believe apache is at 45, F14 is at around 70ish, F18 will be close to 90

I am using MSAAx2 (begrudgingly - as you already know the implementation really pumps the breaks on performance even on best available cards, like tying Usain Bolt to a ball and a chain), I am still running low shadows otherwise everything else pretty maxed out (expect for the couple of the sliders)

I am not using any sharpening or FSR/NIS or any shader mods. I think they are broken and they give me worse image. This is also why I don't give advice on them as I am poorly positioned to do so. I am glad it's working for you. The only thing I run is Saturation inside ToolKit to get the image nice (around 58ish on the slider) as I found it a little desaturated for my linking 

If anything I find the image in the G2 still a little "too sharp" and edgy. But hell, I can see the scratches in the AR coating of the flight instruments and I am really enjoying my free upgrade to the "G3" 

  • Like 1
Link to comment
Share on other sites

5 hours ago, nikoel said:

Affectionately known as Bitchin' Betty - she is part of the Apache Situational Awareness suite

The designers thought that the pilots were more likely to listen to a female voice

Having heard the Garmin, Airbus and Boeing dude equivalent, I have to agree

 

Mom voice 🙂

Ryzen 7 5800X3D | 64GB DDR4 3600| MSI RTX 4080 16GB Ventus 3X OC  | Samsung 970 Evo 2TB NVME | HP Reverb G2 | DIY Head Tracker Cap | Logitech X-56 throttle | VKB NXT Premium |  Win 11

"Any sufficiently advanced technology is indistinguishable from magic."

--Arthur C Clark

Link to comment
Share on other sites

5 hours ago, nikoel said:

Affectionately known as Bitchin' Betty - she is part of the Apache Situational Awareness suite

The designers thought that the pilots were more likely to listen to a female voice

Having heard the Garmin, Airbus and Boeing dude equivalent, I have to agree

 

Nah they just wanted the Star Trek experience. But if she said red alert, they'd be sued by paramount. 

i7 13700k @5.2ghz, GTX 3090, 64Gig ram 4800mhz DDR5, M2 drive.

Link to comment
Share on other sites

56 minutes ago, Gunfreak said:

Nah they just wanted the Star Trek experience. But if she said red alert, they'd be sued by paramount. 

I doubt it. They wouldn't have had any monetary loss to claim damages for.

i9-13900k (stock), ASUS TUF 4090 OC, Gigabyte Z790 UD AX, Corsair Vengeance Black 64GB 5200MHz DDR5, 1 x 1TB WD_BLACK SN850x GEN4 NVMe, 1 x 2TB WD_BLACK SN850x GEN4 NVMe, 1 x SanDisk SSD PLUS 1 TB SATA III 2.5 Inch Internal SSD, Thermaltake Toughpower GF3 1650W PCIe GEN 5.0 ATX3.0 PSU, Corsair AIO iLink cooler, Corsair Graphite 780t Full Tower ATX Case (White), Meta Quest Pro VR headset, Samsung Odyssey Neo G9 49" UltraWide Dual Quad HD 240Hz Curved Gaming Monitor, Thrustmaster Viper Mission Pack, Thrustmaster Cougar F-16 throttle, Thrustmaster Warthog throttle and stick, Thrustmaster F-18 stick, Thrustmaster TPR rudder, Winwing ICP, Edifer e25HD Luna HD speakers. 1 Gbps uplink and downlink FTTP. Windows 11. Hot tub, hot wife. Living the dream!

Link to comment
Share on other sites

4 hours ago, nikoel said:

I like to think we get good performance to compensate for us being helicopter pilots 😅

I'll be honest with you TED. I have no idea, after I saw the counter above 45FPS I forgot about it and started tuning to a very scientific "good feeling" (smooth, whilst also being realistic and accepting DCS is going to drop a frame or two every now and then). I believe apache is at 45, F14 is at around 70ish, F18 will be close to 90

I am using MSAAx2 (begrudgingly - as you already know the implementation really pumps the breaks on performance even on best available cards, like tying Usain Bolt to a ball and a chain), I am still running low shadows otherwise everything else pretty maxed out (expect for the couple of the sliders)

I am not using any sharpening or FSR/NIS or any shader mods. I think they are broken and they give me worse image. This is also why I don't give advice on them as I am poorly positioned to do so. I am glad it's working for you. The only thing I run is Saturation inside ToolKit to get the image nice (around 58ish on the slider) as I found it a little desaturated for my linking 

If anything I find the image in the G2 still a little "too sharp" and edgy. But hell, I can see the scratches in the AR coating of the flight instruments and I am really enjoying my free upgrade to the "G3" 

I've started using msaa again and it's still giving good performance. It just looks that little bit nicer although actually pumping up the resolution in openxr to over 100% it gets pretty close to not needing it at all. 

For what it's worth fsr seems to be working a treat for me with the latest drivers anyway. Definitely gives a little more headroom when set at 80% with little if any noticeable degradation in iq. 

A good setup for pushing and testing the tweaking is in apache  starting at Beirut with late afternoon sun and few clouds, high traffic. Takeoff and do a low level run up to the ridge east of the airport and sweep left around the city. If the setup is all working you'll easily get smooth 45fps until over the city where it may drop to 40 if u r low but pretty smooth. If the system manages that then it will manage anything really. 

This is generally my testing circuit. I get quite a bit of advantage with the fsr in this case. 

Link to comment
Share on other sites

I may have missed it somewhere, but will ask anyhow.  Will this work with the Valve Index or is it more limited to the WMR variety?  I have tried it with my G2 but prefer to use with Index, just cannot find any instructions to set up like the G2 had posted.  Thanks for the hard work on this.

System Specs: System Specs: MSI Tomahawk Z590 wifi, I9 11900K Processor, 64gb Corsair Vengance 3600mhz ram, 3080 Nvidia Video Card, Windows 11 Pro, 2 Samsung NVMe SSD (2tb and 1 tb), Valve Index VR system

Link to comment
Share on other sites

3 minutes ago, Firebird1955 said:

I may have missed it somewhere, but will ask anyhow.  Will this work with the Valve Index or is it more limited to the WMR variety?  I have tried it with my G2 but prefer to use with Index, just cannot find any instructions to set up like the G2 had posted.  Thanks for the hard work on this.

I'm just guessing but, considering the Index and Steam/SteamVR are both made by valve, I would imagine the Index requires SteamVR.

Ryzen 7 5800X3D | 64GB DDR4 3600| MSI RTX 4080 16GB Ventus 3X OC  | Samsung 970 Evo 2TB NVME | HP Reverb G2 | DIY Head Tracker Cap | Logitech X-56 throttle | VKB NXT Premium |  Win 11

"Any sufficiently advanced technology is indistinguishable from magic."

--Arthur C Clark

Link to comment
Share on other sites

36 minutes ago, Firebird1955 said:

I may have missed it somewhere, but will ask anyhow.  Will this work with the Valve Index or is it more limited to the WMR variety?  I have tried it with my G2 but prefer to use with Index, just cannot find any instructions to set up like the G2 had posted.  Thanks for the hard work on this.

Interested to know what your frametimes look like with the Index.  With the G2 you're primarily trying to get out from under WMR4SVR.  Seems like the Index shouldn't have that problem.  Should be a good match for your 1080ti no? 

Ryzen 5600X (stock), GBX570, 32Gb RAM, AMD 6900XT (reference), G2, WInwing Orion HOTAS, T-flight rudder

Link to comment
Share on other sites

11 minutes ago, Fiztex said:

Did anyone managed to get rid of distortions on Quest 2? Performance with OpenXR is great, but these distortions are kind of disturbing :crazy:

Is your PD set to 1.0 everywhere including DCS and are you using my .ini (at the bottom of the guide)

Link to comment
Share on other sites

43 минуты назад, nikoel сказал:

Is your PD set to 1.0 everywhere including DCS and are you using my .ini (at the bottom of the guide)

PD is 1.0 in ODT and DCS, I don't use OpenXR tool (OpenXR runtime remains Oculus, as otherwise it doesn't run in VR). I use OvGME version, I believe it has the same ini file built-in, right?

Actually I was able to remove the distortions by using this tool:

All I needed to do is to play around with L_TOP and L_BTM parameters while tilting my head left/right and up/down. Ended up with L_TOP equal to L_BTM and both are set to 1.14. Headset is indeed get recognized as G2 and thats the issue for Q2 as the projection shape should be different. FYI my in-game IPD is 52, IRL IPD set to 62 (middle between 1 and 2 positions of the Quest 2 IPD slider).


Edited by Fiztex
Link to comment
Share on other sites

Firstly, thanks for all your posts. I've been digging around in here daily since I recently got a G2. 

Interested in some opinions on what I'm seeing here with the Apache re: frametimes

MR on: 20ish fps, not nice, high frametimes
MR off, no frame cap in RTSS: 50-60 fps, low teens on frametime, feels ok except for that common ghosting/judder in the 3-9 line or during fast turns that we're aware of.
MR off, 45fps frame cap in RTSS: feels best out of the three but I'm curious as to why my frametimes sit above 20 in this instance?

Regardless since option 3 feels best it's my go but I'm interested in understanding why frametimes sit that way. 

Conversely I have nice low frametimes in the F-14 when MR is on 

Link to comment
Share on other sites

45fps = 22ms, if you can the framerate to 45fps your frametimes will always be >=22.

When running MR on, the frametime displayed by appGPU is what your gpu is currently rendering, the frametime displayed by postGPU is the overhead for MR etc.  Add app and post together to get total GPU time, if the total is: -

  • <11ms then MR will disengage and it will show at 90fps
  • >11ms & <22ms then MR will engage at 45Hz and it will show at 45fps
  • >22ms & <33ms then MR will engage at 30Hz and it will show at 30fps
  • >33ms & <45ms then MR will engage at 20Hz and it will show at 20fps

I've noticed that appCPU time is increased when using RTSS to cap framerates, whilst this can smooth things out a little it's not best practise.  I'm currently running 60Hz, 80% resolution, MR off, uncapped and pulling >75fps - I've overclocked the GPU to do so :D

Ryzen7 7800X3D / RTX3080ti / 64GB DDR5 4800 / Varjo Aero / Leap Motion / Kinect Headtracking
TM 28" Warthog Deltasim Hotas / DIY Pendular Rudders / DIY Cyclic Maglock Trimmer / DIY Abris / TM TX 599 evo wheel / TM T3PA pro / DIY 7+1+Sequential Shifter / DIY Handbrake / Cobra Clubman Seat
Shoehorned into a 43" x 43" cupboard.

Link to comment
Share on other sites

1 hour ago, edmuss said:

45fps = 22ms, if you can the framerate to 45fps your frametimes will always be >=22.

When running MR on, the frametime displayed by appGPU is what your gpu is currently rendering, the frametime displayed by postGPU is the overhead for MR etc.  Add app and post together to get total GPU time, if the total is: -

  • <11ms then MR will disengage and it will show at 90fps
  • >11ms & <22ms then MR will engage at 45Hz and it will show at 45fps
  • >22ms & <33ms then MR will engage at 30Hz and it will show at 30fps
  • >33ms & <45ms then MR will engage at 20Hz and it will show at 20fps

I've noticed that appCPU time is increased when using RTSS to cap framerates, whilst this can smooth things out a little it's not best practise.  I'm currently running 60Hz, 80% resolution, MR off, uncapped and pulling >75fps - I've overclocked the GPU to do so :D

Interesting, thanks very much for this. I'll give your settings a test too. Forever tweaking, barely playing haha

Link to comment
Share on other sites

On 4/12/2022 at 2:19 PM, DeltaMike said:

Interested to know what your frametimes look like with the Index.  With the G2 you're primarily trying to get out from under WMR4SVR.  Seems like the Index shouldn't have that problem.  Should be a good match for your 1080ti no? 

Sorry, Hadn't updated my specs in a while (corrected)............lol.   The 1080ti ranged in the 30-40fps range with the 1080 while the Index had similar fps, flyable but not perfect at Med-High settings.  With the 3080, and upgraded processor and M/B it is near the 60fps range with similar settings.

System Specs: System Specs: MSI Tomahawk Z590 wifi, I9 11900K Processor, 64gb Corsair Vengance 3600mhz ram, 3080 Nvidia Video Card, Windows 11 Pro, 2 Samsung NVMe SSD (2tb and 1 tb), Valve Index VR system

Link to comment
Share on other sites

^Interesting, I would have thought the Index would have run circles around the G2 in SVR

Your settings fall into three main categories:

  1. Those that affect your CPU times.  Shadows mainly but water and clouds also.  Tree visibility, not so much any more (used to be).  Civ traffic.  You're fine there; just remember, anything that affects your CPU time will affect your GPU time also and it's really easy to get yourself CPU limited in DCS so, don't go too crazy with that stuff.
  2. Those that affect your capacity to map textures.  I get the feeling AF and detail factors fall into this category.  You're fine there.
  3. Those that affect how many pixels you're rendering.  This would include pixel density, resolution settings, scale factor (if you're using FSR or NIS) and MSAA.  You're better off than you used to be. 

I'm not sure what I think about vis range any more.  Used to be a factor but not anymore, not with my new rig and probably not with yours either.  Crank it up.  

Optimal settings depends on the map and mission.  You can get away with stuff on the Caucasus map that you can't get away with on Syria or Marianas; and stuff in SP that you can't get away with in MP.

I'd suggest starting with a solo mission on the Caucasus map.  

  • You should be able to crank up textures, vis range, and AF all the way.
  • You can drive either HMD at native resolution.  PD 1.0, resolution (in OpenXR Developer Tools) 100%, and you don't have to use FSR or NIS if you don't want to.

With the Index, you might consider over-sampling.   I'd try 150 or 200% (using the ODT slider, leave PD alone), which would fool the eye into thinking your HMD is higher resolution, and will have a bit of an anti-aliasing effect.  Far as anti-aliasing is concerned, with the Index, I feel supersampling is gonna be superior to MSAA in pretty much every respect.  How much you supersample depends on how much you want in terms of shadows and clouds.  You should be able to add a fair amount of eye candy. Let us know what you come up with

With the G2, you have some tricky decisions to make. 

I think there's a strong argument for running it at native resolution -- the clarity is remarkable, makes it easy to read instruments and easy to spot things.  As you dial down the resolution, your "sweet spot" gets smaller and blurrier, and your sweet spot is small enough to begin with.  

The only problem is, you won't have much left over. 

You set water at whatever suits your needs; it'll be higher if you're primarily flying off the carrier, and lower if you're mainly over land.  Once you've done that, between shadows, clouds, and MSAA, you're only going to have room to crank up one.  

There's a great temptation to undersample the G2 to make room for at least two of those things, with MSAA being the greatest temptation of all, because the G2 image is pretty hard and shimmery.  This is where FSR and NIS come in.  Just be careful with that stuff.  You can get away with a scale factor of 77% (which is about 60% resolution) without affecting game play to any appreciable extent.  Go much lower than that and it'll hurt your spotting distance.  

Some of us strongly feel the G2 should be should be run at native resolution, and one should eschew scaling and learn to ignore the shimmering.  Others feel strongly FSR or NIS should be brought to bear.  Don't let people "should" all over ya.  Pick which one works for you.

Once you're tuned for the Caucasus map you should be OK on PG or NTTR (long as you're not there just to buzz Dubai or the Strip).  If you're flying on Syria, you'll have to make an adjustment.  I back off on textures, which makes sense for me because my VRAM bandwidth isn't as good as yours.  You might back of on FSR or NIS (if you're using it) because it's easier to adjust on the fly.  Pick your poison.  If there's an answer for Marianas, I'll be danged if I know.  

If you're in a complex single player mission, like the ones Liberation can crank up, frankly I think the answer is to run the mission on a server.  

If you're going in to multiplayer, who knows.  You're gonna be CPU (or RAM) limited  so I would personally look mainly at your shadows settings.  What used to work with me back in the Rift S days was to turn shadows on for SP, off for MP and that seemed to work for me.  I haven't been doing MP since I got my G2 and haven't seen a lot of discussion on that topic here.  


Edited by DeltaMike

Ryzen 5600X (stock), GBX570, 32Gb RAM, AMD 6900XT (reference), G2, WInwing Orion HOTAS, T-flight rudder

Link to comment
Share on other sites

Openxr Toolkit crashes my MSOPENXR TOOLS (dev kit) if I uninstall toolkit, it works fine.. When I have toolkit installed, I can't pull up the HMD options regardless....

Any fix for this?

I'm on a quest 2 with link, RTX2060, windows11, 32gbDDR4, i7-9700k

I have opencomposite, I can load in to DCS even running the OPENXR version on 3dmigoto, use those in game options, I KNOW openxr is working, just can't get toolkit to run with it regardless (I've tried with and without mods) 

Link to comment
Share on other sites

Have you got stable or beta branch of WMR4SVR?

Ryzen7 7800X3D / RTX3080ti / 64GB DDR5 4800 / Varjo Aero / Leap Motion / Kinect Headtracking
TM 28" Warthog Deltasim Hotas / DIY Pendular Rudders / DIY Cyclic Maglock Trimmer / DIY Abris / TM TX 599 evo wheel / TM T3PA pro / DIY 7+1+Sequential Shifter / DIY Handbrake / Cobra Clubman Seat
Shoehorned into a 43" x 43" cupboard.

Link to comment
Share on other sites

On 4/16/2022 at 8:21 AM, edmuss said:

Have you got stable or beta branch of WMR4SVR?

I'm on standalone beta. 

There is no guide for the quest and it doesn't work with WMR.. So using the DCS updater I have steamvr checked, opencomposite installed, I have devloper tools reading that quest is the set API and the demo works. 

Not running WMR portal either.

If I download and install toolkit, it crashes my developer tools when I go click on any of the tabs, and DCS won't start. It crashes on start up. 

For Quest users do we need to delete Dev tools and install toolkit instead or? 

It would be nice to have a step by step install for the Quest2 as so many people use them and they have their own software. 

 

I ended up downloading 3dmigoto for OPENXR and it runs, and I can access everything, So I'm pretty sure OPENXR is running,

I just can't use the toolkit, and it's what I want to use the most lol. 

Link to comment
Share on other sites

@HoBGoBLiNzx3, The OXR Toolkit is optional really. The OXR Dev Tools gives you the resolution options, MR on/off, which I use on always. And the OXR version of 3Dmigoto gives you sharpening, the haze reduction, color enhancing, and more. I am using the OXR Toolkit, but almost using none of its features now. I could even not use it at all since setting up 3Dmigoto...

Some others may weigh in on how they use both to achieve more, but perhaps you don't need it.


Edited by johnbowzer
Clarity

Intel i5-12600K | RTX4080 Super OC 16GB | 64GB 3600 MHz RAM | M.2 SSDs | Meta Quest 3

Link to comment
Share on other sites

I'm still (constantly) experimenting but I think I may have landed on something that suits me best. 
This is with a 3080ti and 12700K so I appreciate I'm pretty fortunate here but sharing in case anyone else with similar specs cares to try and finds it good for them

90hz
OXRT 110%
MSAA 2x
RTSS 45fps cap
Textures Medium
Ground Textures Low
Preload radius 80000
Shadows low

The RTSS cap seems to add 9-10ms to the CPU frametime, that takes me into low 20s but tbh it feels smooth.

I really liked the superior smoothness of MR, I'd use it more if it wasn't for the artifacting in the Apache blades. I thought I'd use MR for Tomcat flying but in testing I was all but certain I was seeing more shimmering in edges in the cockpit and more chromatic abberation in the peripheral on instruments. I can imagine going back to this as development continues.

Locking at 45fps isn't the ideal, doesn't feel as smooth as MR but comes close to eliminating the judder in the 3-9 line. 
I did like edmuss' 60hz solution but on a user personal level I'm happier using 45fps and increasing my resolution to 110%, the clarity is worth it for me. 

  • Like 1
Link to comment
Share on other sites

The RTSS hack to lock to 45fps does smooth things out quite a bit, however it does add to the CPU time.  I'm not 100% sure on the mechanics of this but I think the CPU time isn't accurately represented within the OXRDT performance overlay, it essentially locks it to 22ms (45fps) but I don't know if this is a bogus figure and any additional CPU load (due to mission AI etc.) will be on top of or absorbed by the extra overhead.

As a worked example: -

Without RTSS capping I might be getting 12ms appCPU with 18ms appGPU and getting 55fps.  If a 4ms appCPU overhead is added to give 16ms then I will still be getting 55fps because it's still faster than my 18ms appGPU.

With RTSS capping I will be getting 22ms (displayed) appCPU with 45 fps (I don't recall if it locks the appGPU to 22ms as well or displays what your GPU is actually doing).  With the additional 4ms appCPU overhead I don't know if the appCPU will then jump to 26ms which could in theory drag the GPU down to 38fps or whether it would just be absorbed within the RTSS overhead and in reality the GPU output would be unaffected.

Ryzen7 7800X3D / RTX3080ti / 64GB DDR5 4800 / Varjo Aero / Leap Motion / Kinect Headtracking
TM 28" Warthog Deltasim Hotas / DIY Pendular Rudders / DIY Cyclic Maglock Trimmer / DIY Abris / TM TX 599 evo wheel / TM T3PA pro / DIY 7+1+Sequential Shifter / DIY Handbrake / Cobra Clubman Seat
Shoehorned into a 43" x 43" cupboard.

Link to comment
Share on other sites

^Yeah I waded in to that discussion yesterday (it was that or do my taxes) and I thought my brain was going to melt.

What I took home was, basically when you cap frame rates you're telling the CPU to wait, which will make your CPU times look longer but as long as it isn't doing anything that shouldn't hurt you.  

Only problem is, CPU scheduling is a very complex matter in VR and there are potential problems if you mess with that timing, among them it can increase your latency considerably. 

SVR is such a mess to begin with, I'm not sure you could screw it up any more than it already is.  Back in the day (what, a month ago?) people were getting OK results in SVR by locking frames, to me it was meh but it definitely did something.  

In OXR, I dunno.  I do think we run into synch issues, which range from minor to highly annoying.  I get the feeling, those of us with lower power GPUs don't get it as bad, and that's because we are used to squeezing every last electron out of those things.  If you're really close to your frame-rate target I don't think you get as many synch issues and you don't need to lock your frame rates.  

For those of you with 3090 or 3080ti I imagine the synch issues can get dreadful at times.   Couple of ways to approach that.  Personally I think ya'll should try loading up your GPU.  Crank up settings, crank up resolution, whatever it takes to get your frame times up to like 20ms, not much more than that but not much less than that either.  That may be all you have to do, in other words you shouldn't need frame caps or motion smoothing if you're right in that golden range.  And you'll get an awesome image, shoot I'll take supersampling over MSAA every day of the week if I had that option. Which, I don't -- you do.  Shadows, clouds, textures -- go wild.  Bet my own money it works.  I got five cents right here

Ryzen 5600X (stock), GBX570, 32Gb RAM, AMD 6900XT (reference), G2, WInwing Orion HOTAS, T-flight rudder

Link to comment
Share on other sites

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...