My 2070 SUPER VR settings and suggestion (Reverb G2 - WMR)

Agreed, my performance went to ■■■■ with the new drivers, back to 457.30

2 Likes

If you can stomach it, I recommend you try setting your G2 down to 60Hz. This will allow Motion Reprojection to work all the way down to 20FPS allowing to bump other sliders up.

For me, it is so immersive, that after a couple minutes, I no longer notice the flicker.

Here are the setting that work best for me at 60Hz (my system specs are very similar to CptLuck8):

Render Scale: OXR65,TAA70 (OXR70 works, but motion reprojection turns off occasionally)
Terrain LOD: 70
Terrain Vector Data: Ultra
Buildings: Medium (big hit w/ motion reprojection on Ultra)
Trees: Medium
Grass and Bushes: Low
Objects LOD: 200
Volumetric Clouds: I usually fly in clear skies, so haven’t tested
Texture Resolution: Ultra
Anisotropic Filtering: 4x
Texture Supersampling 4x4
Texture Synthesis: Ultra
Water Waves: Medium
Shadow Maps: 768
Terrain Shadows: 128
Contact Shadows: Low
Windshield Effects: Ultra
Ambient Occlusion: Off
Reflections: Off (even on LOW makes motion reprojection unreliable)
Light Shafts: Low
Bloom: Off
Glass Cockpit Refresh Rate: Low

1 Like

I should try your setting, @MinnMonkey . Might be even better than mine. I’m just blundering around, not quite knowing what I’m doing. My system specs are i9-9900K, 32 Gb 2666 Hz RAM, 2070S, 1 Gb non-system 970 EVO Plus.

Taking off from @CptLucky8 's recommendation, I have my refresh rate at 90 Hz in WMR settings, Automatic Reprojection and 100% scaling in OpenXR, and the following MSFS Graphics settings under the VR settings choice. The settings give a noticeable amount of “juddering” flying the Zlin Shock Ultra in wilderness areas of low or no photogrammetric detail but I really like the scenery and 3D I’m seeing.

Try flying Machu Picchu, Peru with these settings. GPS location: 13°09’48.6"S 72°32’43.4"W


You shouldn’t get juddering with motion smoothing.

  1. You might want to enable WMR OpenXR “Display Frame Timing Overlay” and check you’re delivering the minimum 30fps needed for reprojection.

  2. You might want to lower the OpenXR Scale down from 100% too, because motion smoothing is doing its pixel interpolation based on the OXR resolution, therefore the bigger, the slower. You might want to try another approach with your video card and motion smoothing: TAA70 + OXR65 (or TAA80 depending on aircraft/location)

What is interesting to me is that when I’ve written the Index and the G2 topics, the G2 + WMR was running better than the Index and I could push more rendering pixels in the pipe with the G2 as well. But with SteamVR 1.16.4, I can now fly the Valve Index at TAA100% always and I’m just varying SteamVR SS between 124% and 128% and with these, I get nearly as good legibility on the EFIS and excellent outside view competing with the G2, with a higher fps than the G2 and in rendering more pixels too!

3 Likes

I am wondering what is the benefit of choosing automatic reprojection over “always on”?

If I’m not mistaken, always on will for example render 1 frame then extrapolate 2 more, then repeat, whereas automatic will render as much frame as possible and “motion smooth” only if it can’t render enough fps.

The later is desirable if there are times you can render more fps because in this case you won’t have the motion smoothing artifacts. The former is useful to make it more “stable” because WMR doesn’t have to adjust in real time the prediction ahead at the risk of failing and causing more artifacts.

It really depends on your hardware and personal preference but automatic is working quite well with WMR.

1 Like

Thanks for your pointers, @CptLucky8. I read a bunch of your previous advice in this thread and others and some of your wisdom is slowly sinking in, probably not as deeply and thoroughly as it should, though! :slightly_smiling_face:

I don’t know how much you’d disagreed but one thing that I decided is that you don’t really have to fly anywhere to see a good deal about how a change in graphics settings is going to affect your visual VR display. You can just change settings in Options, General, then go back to the cockpit in a Ready to Fly situation spawned in the air, look around in the cockpit, and out at the scenery and get an idea of what the changes did to your view. It makes the point of comparison fast and 100% reproducible since it’s easy to save and reload a .FLT file.

I noticed this worked if I turned on “Display Frame Timing Overlay” in WMR OpenXR. As I changed the settings the predicted frame rate (Hz) would change as I looked at various parts of the cockpit and scenery and if I cranked up Render Scaling in MSFS, the frame timing would really go to hell, etc. (testing censor!). Neat that the frame rate background turns to red when one is violating ~30 fps limit but is blue when above and if the background is flickering between red and blue, I guess you’ve just about used up your GPU performance budget. But just as running MSFS Developer Mode in 2D, I think running the OXR timing overlay screws up the sim considerably. When I went to the MSFS graphics settings while the overlay was running, the Options display was messed up big time and if I tried flying with the overlay, I got terrible performance, had to shut off the overlay display and restart MSFS to get decent performance back with the OXR overlay off. Still the overlay is great just in checking things out in a static cockpit in VR to see what sort of relative performance hit you’re going to get tweaking various settings. Since I’m running the 461.09 driver and GeForce Experience, maybe I could do better if I killed GFE and went back to 457.30…

I liked your budgeting explanation of how you can spend GPU cycles consuming various resources. It’s kinda like only having so much allowance and being a kid in a candy store and wondering what candy flavors you’d like to spend your allowance on.

I think my allowance goes extra far for my choice of the Zlin Shock Ultra as my VR plane. It flies slow, so perhaps less of the scenery has to be updated per GPU cycle. And it has such a bare bones cockpit, if you fly in the cockpit and don’t go outside you spend less of your allowance having to update a whole bunch more scenery.

I ~understand what you’re saying about the point of diminishing returns going above TAA60/70. What I find in very limited experimentation is that things look the sharpest if I go for TAA100 and OXR100, both the Zlin cockpit panels and in the scenery distance but actually things look most real if I leave MSFS rendering at TAA70 and go for Texture Supersampling at 4x4 and Texture Synthesis on High. The panels are a little blurrier and so is scenery in the distance but as you’ve pointed out somewhere, that distant blurriness fits right in with air turbidity (and cloudy haze) in real life. So I decided with whatever extra allowance I get in the candy store for relatively hovering the Zlin Shock Ultra in the air, I’d rather spend it on TEXTURE eye candy! :slightly_smiling_face:

Undoubtedly my candy preferences are going to change with time.

2 Likes

Does anyone have an explanation why it looks downscaled or blurred when I use Motion Repro in OXR either to Always On or Automatic? When disabled, image is very good and quite smooth (I just wanted a little extra smoothing) but using MP in OXR somehow seems to kind of downscale? Looks really bad. (RTX3090, 64GB RAM, i9 9900K@4.9)

1 Like

I’m seeing the same thing (and was actually about to post about it). Hitting/exceeding 30fps, stationary on the runway. As you say, looks downscaled or perhaps more like anti-aliasing is turned off. Jagged lines, unclear instrument and menu text. Perfectly sharp when reprojection disabled.

Tried various combinations of capped fps, low latency & pre rendered frames to free up MainThread,. No impact.

Also tried different combos of OpenXR and in game resolutions. No impact.

(RTX 3080, 3600X@4.4, 32G@3200)

2 Likes

Hi,

Well, I’m not a technician, and someone will probably explain this better. Motion repro is a simulation of 90 FPS. Instead of looking at 90 single different pictures per second, you’re looking at 3x each frames x 30 on full repro. So, it is expected to have some form of degraded performance.

This picture is not scientifically accurate :), sorry, but it pretty much illustrates it. ( 90 fps is above, and below, each letter is a third of the size of the original, multiply 3 times to fill up the same space/time. )

Hi all! A quick update. I went from a 3900x to a 5900x last night — and I am absolutely shocked by the improvement.

System:
5900x (SMT on so far, but if same holds true SMT off should be preferred)
32GB 3200
2070S
Reverb G2

Now, if you’re looking for solid, objective testing, I don’t have that yet. I deliberately decided to focus first on experience, not on chasing numbers. My test was to replicate my last flight in VR: Xcub from KHAF, up the coast to Downtown SFO. Weather was Spring Sunshine (SoFly Preset.)

Previously I had been running OXR65/TAA60 with reprojection on. I loaded in to the sim and even before switching to VR I noticed a huge improvement in the speed of the menus, especially loading my mutiple mod/add on AC. Going in to the game, activating VR, the difference was immediately noticeable. Immensely smoother when looking around, no stutters, just natural. I decided to take TAA to 70 and while I could tell there was a small decrease in frame rate, repro still did wonderfully.

Performance stayed strong even when flying around downtown, and as the sun set and began casting rays.

I finally turned on the FPS counter and found that I was getting consistently 6-10 FPS better than before.

Now, I think that the 5900 is excessive, but it’s what I happened to snag. One would expect similar results going from any 3x00 to the equivalent 5x00.

I have more testing to do — I suspect the gains will be less noticeable with repro off - because as @CptLucky8 says, repro is mainly taxing the CPU.

I’m planning to test my old 70/70 without repro and see how it compares to before in both seat of the pants and FPS testing.

It seems that, unsurprisingly, single core clock speed — and most importantly IPC — is hugely important for us in VR.

1 Like

@JALxml
Thank you for your feedback! I don’t notice any particular decrease in fps with WMR fps overlay (actually I didn’t deactivate it for a long time now) but I’m also using 457.30 and this might be the reason (I don’t want to try 461.xx anymore until they release the version which is supposed to cure the VR stutters, and until we also get the next FS2020 update).

Nevertheless, I’m glad this is helping you and others, at least, this tells me I’m not insane this much :crazy_face: But overall this is really it: you have a frame time allowance and as long as what the CPU and GPU are not over-demanding during this frame time, you’re good to go. This is the reason you have no other choice than lowering some of the settings in order to give room for motion smoothing computations (it takes up to 25% of my CPU/GPU budget on my test system).

@Incubus5604 @MickT1701
If you leave OpenXR Scaling to AUTO, it is possible it is adjusting this behind your back. If you also use GeForce Experience, it will also adjust settings behind your back. This is why I prefer a controlled environment, let alone these systems are adjusting based on a GPU point of view, whereas I prefer adjusting from a Pilot point of view (they’ll reduce texture resolution and antialiasing for example, whereas I prefer reducing ambient occlusion and shadows).

@AIRBORNE4795
This is only one of the aspect: there is reprojection and motion smoothing both contributing and working hand-in-hand. I’ll try explaining (but it is not easy):

The app is rendering a view, then re-uses the same view 2 times more. But unlike 2D monitor, it will also re-project the view, that is it will display it in a coherent 3D location regardless of your head movement so that what is displaying stays at the same place in 3D. In addition, with motion smoothing, it can “project the pixels” too using a similar technique as “Sports Mode” on the TV giving the illusion of 240Hz with 60Hz signal.

In turn: reprojection makes sure the image is displaying at the correct 3D location regardless if this is a new image or a duplicate, motion smoothing helps creating new images from the existing ones in order to fill the gaps. Reprojection is also intrinsic to VR and is mandatory to compensate render-to-photon delay.

NB: it is best explained and detailed in Alex Valchos videos I’ve linked in the OP.

@Trowels
Thank you for your feedback! All simulators I know of are single core bounded usually, except XP11 which has a very effective use of multi-core compared to its competitors and is less prone to this, unless you’re using an add-on which is mostly single-core limited (anything SASL seems to fall into this from my experience so far). Choosing a CPU with single-core perf is therefore usually recommended and is one of the reason I’ve chosen the 9700K instead of a AMD equivalent at the time (The 9900K price/benefit was not good enough to me)

1 Like

Hey, Cap — hard to tell where you landed on this: I’m about to try a full VR flight in the CJ4 for the first time. Did you decide what settings you like for that? 60/80 without repro? Or are you still doing repro on jets?

I nearly don’t use the G2 anymore… mostly the Index instead.

I did try yesterday the G2 with SteamVR and WMR and tracking was giving me a awful time compared to the razor sharp outside in tracking of the Index. The cockpit was floating around me and lagging every move I made, WMR was sometimes shifting the entire position all of a sudden.

Sure the visuals were nice in the very narrow center I would appreciate them, but comparing again with the Index where not only I can see wider both horizontally and vertically, but more than this I can see clearly from edge to edge in just moving the eyes, was really revealing to me the G2 is just a deceiving product in many aspects you can probably only see once you’ve used a Valve Index. Really night and day an experience and I only wish the Index panels were refreshed with higher res, that’s all that would be needed to me.

Now It might be my G2 has a specific problem but I doubt it now since the latest HP Q&A. The “official” answer is clearly telling I’m right since the beginning and they have compromised the optics in the G2 in ways they didn’t need to (see below*).

Here is for the Index settings I’m using right now (with latest SteamVR 1.16.5):

My 2070 SUPER VR settings and suggestions (Index - SteamVR) 🟢 - #449 by CptLucky8

Otherwise with the G2 I’m still finding the settings I’ve posted when using motion smoothing are working good on my test system:

My 2070 SUPER VR settings and suggestion (Reverb G2 - WMR) - #2 by CptLucky8

That is I can’t use less than OXR 65% (because of distortions bug if below) and no more either because of performance on the 2070S. From there, I adjust TAA more or less 70 and clouds/terrain depending.

*I mean both G1 and G2 panels are the same res and size, they could have kept the G1 optics now that it shows they can anti-CA it effectively (as reported by many G1 users). Choosing the Valve optics for less CA also comes with less refraction and the need to curve the lens more therefore lowering the sweet spot cone and therefore making this headset optics even more questionable without any eye relief system and a so much recessed face mask. This is what I’m thinking about in terms of “compromised optics in the G2”. And let’s not forget they said initial production was slightly delayed because of a last minute change with the lenses… what last minute change is this??!

here is the HP comment

We have found that a minority of user’s eyes are farther from the optics after making the correct adjustments than optimal. Reducing FOV and optical performance. You are right that there is a careful balance we must hit between accommodating users with glasses, but also ensuring user’s eyes are close enough to the lenses to remain in the sweet spot. I cannot comment on future plans but this is something that is definitely top of mind as we want everyone to have the same experience.

https://www.reddit.com/r/HPReverb/comments/l76cke/hp_microsoft_here/gl54msi/?utm_source=reddit&utm_medium=web2x&context=3

In other words, I wouldn’t hold my breath about “what to see if the headset is calibrated properly” because even the HP person behind the G2 hardware is not providing a clear answer to the specific calibration measures and questions I’ve raised. But he is suggesting the distance to the lens is somewhat narrow and because of the shape design of the G2 it is very possible many people can’t get it right at all.

I believe I’ll process a RMA if not requiring a plain simple refund (even if post 30 days) because they’re selling a product they know can’t work with many people, unless they’re wearing their glasses which is not supposed to.

2 Likes

Thanks @CptLucky8 - assuming when you said ‘OpenXR Scaling to AUTO’ you meant the Motion Reprojection option, I switched to Always On. It seems Automatic Motion Reprojection does some other stuff automatically too. A little embrarassed that’s the one variable I didn’t consider despite all the other changes I was making and testing to try and isolate the issue.

Appreciate the effort and patience you keep affording all of us - I do hope you get to spend at least a bit of time enjoying some flying after you’ve taken care of the rest of us.

Thanks again.

2 Likes

@CptLucky8 I also have a HP Reverb G2 and having the same head tracking issues in MSFS as I use SteamVR because I want to use OVR Toolkit when I stream.

I also use X-Plane and I use my G2 with my Index controllers and do not have any of the head tracking issues so not sure what the issue is but might also try to use my Index in MSFS to see how that will work.

Thanks for your feedback while we work thru the current VR issues.

Yeah,I have a love and hate relation ship with this headset ( G2 ), it’s so close to being great, buy yet so far, which makes it even more frustrating. I regret selling my G1 so early, I didn’t get to compare. But from what I can remember, I could just glance at my VS and airspeed without feeling “bothered”.

I’m thinking about trying the Quest 2, but I’m not a big Zuk’ fan, it pains me to give him money.

I’m even thinking about getting the Index again, but the glare was a problem for me in XP…

ugh… money…technology…

1 Like

Thank you for your kind words! No I really meant Scaling to Auto, where OXR will then automatically adjust the scaling based on your VRam amount (and other things we don’t know).

You might want to read this then:
My 2070 SUPER VR settings and suggestions (Index - SteamVR)

and the latest post about SteamVR 1.16.5:
My 2070 SUPER VR settings and suggestions (Index - SteamVR) 🟢 - #2 by CptLucky8

Thank you for all your feedback as well, this is helping all of us in turn!

I regret having returned the G1 when they’ve announced opening the G2 preorders soon (I was in the 30 days return window).

It is so amazing how I can just grab the Index, wear it, and it perfect right away: no sweet spot to adjust, no clarity problem, no tracking issue, feels solid and robust, the handy in-headset button to change your SteamVR settings live in FS2020 session without the need for a controller, etc… so well thought out product for VR. The G2 was promised to be the killer headset for simmers and I’m glad many are enjoying it, but the Index is so much superior, and when I got the chance maybe one day to find a 3080/90, pushing the Index to SS200 is nearly the same legibility in the cockpit than the G2, with the bonus of edge to edge clarity over nearly 85% of the view for me…

I guess I may have found something before you for once then :slight_smile:

I’ve got scaling set to 100% manually, and tried other values too. Also, no Geforce experience. If there was something that far down the chain like that causing issues, I’d expect it to have an impact regardless of reprojection options, but there are so many variablles and interdependencies going on here, you can never know for sure.

[Edit]
Looks like that might have been some placebo or temporary effect. Anyway, I took some screenshots and they all came out looking the same regardless of disabled/auto/always, so it’s definitely a frame-to-frame issue, and this is despite the headset not moving (it’s stationary on the desk). I don’t get a constant 30.0 fps counter - bounces between 29.x and 30, so not sure if it’s a boundary issue. It’s being throttled in main thread by the reprojection though - I’ll get 40fps in same scene when disabled. Anyway, I don’t want to get too bogged down in this as it’s actually fine disabled, would have just been nice to get the extra smoothness. Or maybe it’s because I just can’t stop tinkering…

Thanks again

1 Like

What would help me a lot is to find some technical posts that explain how to calculate the effect on pixel resolution of changing MSFS and OXR scaling. Such as when you post the following, how did you actually calculate the Render Resolution and the Post-Processing Resolution? - sorry if I missed an explanation somewhere but you have almost published a full-length textbook on settings and results (for which everyone is very grateful!). Also, I’ve seen posts by others on Reddit or avsim where someone has calculated with particular settings how many pixels are real and how many synthesized.
Was wondering, too, if that calculation is real and meaningful and how is it done?
WMR Scaling and Dev Tools - Some Explanations

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

Also, it would be great to have in one place a description of the techniques, including the test scenes or flights, that you use to compare settings. Again, sorry if it’s already in one place in one of your discussion threads and I’ve missed it.

Another unrelated question is in the Windows Mixed-Reality Enthusiast’s Guide, which seems to be out-of-date and perhaps initially oriented mostly towards HoloLens developers, there is emphasis on going into Windows Developer Mode and running the Windows Device Portal and looking at the Performance tab. I was wondering if this was of any use and whether by doing that, I’m just going to consume CPU/GPU and make my PC more insecure to boot!