Thanks. I’ll keep motion prediction off for now, as it’s unacceptably distracting to me.
To each his own! after a while I’m getting used to it and don’t pay much attention to these artifacts, but sometimes I still do!
Yep, the nice things about preferences is – we all get to set them according to our own. <3
One thing I do note though – it’s definitely sending the depth buffer for reprojection, because head tracking seems to be smooth at a much higher framerate than anything MSFS puts out. Even when MSFS freezes for a few seconds during loading, I’m able to move my head around and it’s not until I move/rotate a significant portion that it dims the screen and puts up the logo image in front.
If there was no depth buffer sent to SteamVR, this would be impossible.
I’m not sure of this at all, but I didn’t dig enough into this in particular either.
The other problem is that even if there is a depth buffer sent to the OXR API, is the SteamVR OXR Driver using it?
I’m looking at the Optical Flow SDK right now and it has provisions for hints:
Flag to enable external hints:
NVOF API provides clients the flexibility for feeding external flow vectors which will be used as hints by NVOFA while performing the motion search.
Pass good quality flow vectors as hints.
NVOF API provides the option for passing external hints. The external hints are given highest importance while performing the motion search by the hardware. Hence it is recommended to pass good quality flow vectors as hints, and not use the feature in case the hints are not of good quality.
It happens FS2020 is already computing pixel motion vector texture for TAA and this would be a good hint. Should they be able to provide this texture through OXR, and SteamVR uses it to pass along to the Optical Flow lib, this would help a lot. But is there an OXR API for this? Or even an OXR Extension for this? And what about SteamVR?
This can only be solved if Asobo, Valve and NVidia sit around the same table and decide to implement this. Given what they seem to have told about Pimax though… Pimax answered my question😢
Well this interesting. I am a Quest 2 user but used a lot of your settings and suggestions in the past month from this thread. After lots of experimentation I settled on using the SteamVR runtime instead of the openXR that ships with Oculus.
Few reasons for this choice:
- The SteamVR image didn’t look so washed out as the one that Oculus produced.
- SteamVR had less microstutters.
- I love the World scale option.
Within that choice I had been experimenting a lot with supersampling. Oculus users have this great OTT tool, but Steam can do it also. I tried many combo’s but in the end I settled on OTT doing the framerate lock on 18 Hz with ASW and the supersampling. Just leaving SteamVR on 100%. It just was more fluid that way.
Along came SU5 and as many users I was able to upscale resolution and settings without stutters. But after a few flights it became clear that the image quality had changed. I could not lay my finger on what the difference exactly was but the outside world looked more cartoonish and unsharp.
After reading your settings I tried to use the Steam supersampling again and Lo and behold! the sim looked like it was before SU5. Less washed out, more definition in objects and sharper…
I don’t know how this is possible but the change is visible. Only downside is that I now have some minor microstuttering again.
I’m glad this is helping!
One of the reasons behind 126% and 152% (+ FOV 95%) is that with the Index native panel resolution and lens distortion, it seems to me these are slightly enhancing aliasing compared for example to 125 and 150, compensating a little bit for the blurriness due to motion smoothing and/or FS2020 rendering. It is really subjective though but it seems it is striking a balance to me, enough to make barely legible middle EFIS small letters slightly more legible (but aliased).
That’s interesting, I’m also a quest 2 user but never saw any impact at all from increasing the SteamVR super sampling… Is there anything special you need to do for it to kick in?
I don’t know if I understand your question well. Do you mean how to enable supersampling in SteamVR? Just use the slider for resolution scale.
Yes, moving the slider in the steam menu never had any impact for me in visuals or performance even if I put it outrageously high or low. I had thought this was a limitation of virtual desktop but I tried it through link and had the same. Maybe I just needed to restart the sim or something, I’ll have another experiment.
I am wary not to derail this thread which is about the Index, sometimes I am tempted to gift a quest 2 to Captain lucky just to get his input on that headset
I’d be pleased but only if you also gift the computer that goes with it. I don’t want to have anything to do with the company backing this headset up on any of my electronic devices…
One thing to check is whether you’re setting this on the global setting or the per-application setting, which can override the global setting.
Thank you, I was using the per application but maybe will experiment further, especially given the good quest results mentioned above.
I’m afraid I need the PC, but I’ll grant use of it via virtual desktop nightly between 2300 and 0700 hours UK time. No crypto mining and stay out of My Documents
I’ve done a few more tests yesterday about this. Motion smoothing is taking more resources and usually you can balance this out with fewer frames per sec (and artifacts of course).
With the 9700K+2070S I can barely push beyond 152% and 18fps with the Index and FS2020 with aircraft with lots of EFIS for example, like the FBW A320 (for this one I must go 15fps on ground but can switch back to 18fps in the air) (NB: I can push much higher with XP11 though)
What stroke me is how much FS2020 has a struggle when on ground, to the point I have to disable motion smoothing otherwise FS2020 is like rendering pairs of frames in reverse order (displaying frame 2 first, then frame 1 next, then frame 4, then frame 3 etc… this is what it looks like).
Yesterday I’ve tried the FBW A320 latest dev version but this time with motion smoothing off. I’ve been able to keep 30fps (SS 152% and all the rest unchanged) on the ground, and when up in the air 45fps, and this was not bad at all to me.
If you read this topic from the beginning and when VR was initially available, I was explaining how 30fps constant is still useable in practice with the A320 (not FBW back then) and after a while you get used to the juddering. It has been months though that I’m only using motion smoothing at 18fps so I go used to the artefacts instead! Switching back to motion smoothing off was slightly uncomfortable at first because of the juddering but at 45fps it was not so bad after a short while, especially if you develop the habit of rotating the head (reprojection kicking it) instead of translating it (juddering is unavoidable in this case).
This is a short feedback post to share the experience!
PS: about “FS2020 on ground taking too many resources while on ground to the point I have to disable motion smoothing”: this one is puzzling me still because in 2D I can see there is still enough room on the CPU/GPU. I’ve been wondering whether FS2020 wouldn’t be using the GPU in a way which is interfering with the Turing silicon the NVidia Optical Flow lib is using for motion smoothing. What I mean by this is that it is has if both FS2020 and Optical Flow are fighting for the card silicon, which could be the case if for example FS2020 would be using compute shaders for example and if those would take a longer processing time when on the ground (handling much more things). This is one area I’d really like to figuring out, at least just by mere curiosity for these sort of things!
After further testing, Win10 updates, NVidia updates etc… this is what I’m starting to consider my best settings right now for the Index.
Of course, if your hardware is more powerful you will be able to raise a few settings there and there, but the following should be a solid base for most users to get started with.
NB: I’ve been adjusting these flying the FBW A320 which is a demanding aircraft.
Win10 21H1 (with KB5005033)
HAGS OFF, Game Mode OFF
NVidia 471.68 SD (Studio Driver - latest)
SteamVR 1.18.7 (latest non-beta)
FS2020 Graphics Settings (same in 2D and VR)
|Render Scaling||100%||Best for EFIS|
|Terrain LOD||100%||Go-To value all around, raise to 200% while up above|
|Terrain Vector Data||Off||This is controversial to me, prefer off for now|
|Grass And Bushes||High|
|Volumetric Clouds||High||you can save some fps and keep nearly same visuals with Medium instead if you’re flying mostly overcast or with non isolated cumulus clouds in the sky|
|Texture Supersampling||Off||saves a few fps and this helps when on ground (with aniso 16x it is ok)|
|Terrain Shadows||256||I don’t see much impact with 512 but not a large visual diff either|
|Contact Shadows||Off||eats fps and I prefer raising other things instead|
|Ambient Occlusion||Off/Low||Further testing makes me believing it is using a compute shader which is taking a hit on the motion smoothing compute shader. Set this to Off if your system is struggling on the ground for example.|
|Light Shafts||Off||I didn’t really measure the impact of this one completely yet, but it might be affecting motion smoothing for the same reasons as Ambient Occlusion|
|Bloom||Off||This is reducing the legibility of integral lighting in the cockpit sometimes to me|
|Depth of Field||Off|
|Glass Cockpit Refresh Rate||High|
FS2020 Traffic Settings (same in 2D and VR)
|- Aircraft Traffic Type||Off||I don’t need any (flying MP only)|
|- Show Nameplates||.||There is no good or bad choice here!|
|Airport Life||I mostly balance between my taste and perfs|
|- Vehicle Density||15%|
|- Aircraft Density||0%||I don’t need any (flying MP only)|
|- Worker Density||5%|
|Land & Sea Traffic|
|- Leisure Boats||50%|
|- Road Vehicles||50%||Boats are usually lightweight on perfs but cars aren’t|
|- Ships and Ferries||80%|
|- Fauna Density||0%||this was source of fps loss in the past and I didn’t try it lately|
|AI & MP Traffic Detail|
|- Generic Models (AI)||On|
|- Generic Models (MP)||Off||this would be better if only…|
|- Traffic Variety||High|
|Refresh Rate||90Hz||I find this one the best both with and without motion smoothing|
|resolutionScale||152%||tradeoff between legibility and computing power - this is about the maximum render resolution I can push with TAA100 and motion smoothing on the 2070S|
|FOV||95% or 90%||adjust this based on your eye relief setting (with the shorter eye relief position I still can’t view the sides at 95% but I can at 90%)|
|Motion Smoothing||Enabled||I mostly use motion smoothing (see “Tips” for more info)|
|Fixed Rate||18fps||15fps on ground with FBW 320 on complex airports|
|Prediction||33ms||strikes the best balance to me with 18fps or 15fps|
- The only SS settings worth the legibility differences to me are 126%, 152%, 220%, 300%
- When on final, you can quickly change the FOV to 50% to get a very legible runway in the distance and crisp cockpit lettering/EFIS.
- If your hardware permits, you can try SteamVR SS 300% and this gives you the same legibility as SS126%+FOV50%.
- If the aircraft/airport is really too heavy, disable motion smoothing and set a fixed rate of 30fps or 45fps
After testing many different drivers and settings, I find these the best with the “System” specs detailed above.
NB: most settings are using default values in my screenshot. They are inheriting from the Global Settings. If your global settings are different, they will also reflect in your FS2020 custom settings. Besides “Monitor Technology” all the others should match whether inherited or overridden.
NB: I’m now using NVidia Studio Drivers. They seem to be providing more consistent frame time in VR to me, and they are visually free of any artifacts that others are reporting with SU5 like textures shimmering (buildings edges jagged highlights and aliased popup menus are still a problem due to FS2020 itself though).
From now on I’ll use Maximum Performance (unlike the screenshot above)
I can’t really differentiate any particular difference between Optimum Pow. and Maximum Perf.
The latter may be less performant in practice (hitting temp and/or power caps) so I’d tend to lean toward Optimum Pow in general, but I’m yet to hit any protection with Max. Perf in practice.
I get smoother experience and no stuttering with Threaded Optimizations Off. This is very consistent in my experience with the Index and the G2 (along with HAGS Off and Game Mode Off).
I don’t find any need to further tweak Low Latency and VSync mode anymore.
Here is how to make your own NVidia Slim drivers:
1. Unzip the NVidia driver .exe (with 7z for example)...
Then run setup.exe, don’t select GEForce Experience then select the components (I only install Video driver and Physx, and unselect USB and Sound)
Credits: several posts at the guru3d forums
Do you have any specific reasoning behind “threaded optimisation” being off in your drivers?
Is testing both a “reason” enough?
NB: I believe I’ve already commented on this in my different topics or at least somewhere on this forum.
There are technical reasons too, but too complex and in an ever changing driver/system/game environment. But suffice to say as long as the game is DX11 and with VR, Threaded Optimisation OFF should give better results for most in VR.
I found my issue, I had been launching steam along with some other apps via a batch file with admin permissions, while msfs was running as my standard user. This stopped the steam super sampling from working somehow.
I will be looking with quest 2 - Can I ask is there any mathematics to reach these numbers, or was it trial and error ? Thank you
These are for the Index only, based on perceptual measures first and foremost. This is highly dependent on the panel resolution, the lens deformation, the processing of the raw view pixels into the display view pixels etc… As far as logic goes, I rationalize this (and this is biased) that these are integer scale fractions + a little more in order to exploit aliasing at my advantage.
In practice though, between 126% and 140% there isn’t much visual change, the same between 150 and 200, this is why in the end and after having tested nearly all steps by 10 increments (see all my previous posts about this) I finally settle with these 4 for now, knowing 126% was what I used prior SU5 and 152% is what I can use now. The others will have to wait for a 30XX or a 40XX
I shall add than 126% and 152% are giving nearly the same legibility inside the cockpit (with EFIS) but the later is a little better for the outside. If I were using 126% instead of 152% with SU5 I might be able to push some other settings higher, or the fps higher.
NB: 152% and my settings above are a best-balance on this system for now though, well suited so that I don’t have to change settings in flight every time to adjust to the situation while keeping high levels of visuals overall. With these settings the only thing I change during a flight, depending on the situation, is in SteamVR only:
- Click the button on the HMD and
- Change 18fps/15fps (motion smoothing enabled) or
- Disable motion smoothing and set 30fps/45fps.
- On final I like sometimes reducing the FOV to 50% for the highest legibility of both the runway and the EFIS.
To be honest I am not sure even those cards will help! It is a different headset but I have been experimenting a bit now with additional Steam SS on Quest2, using my 3090. Starting with my standard 5408*2736 native resolution and 100 render scale, I see very little benefit from additional Steam super sampling, and certainly nowhere near enough to warrant the noticeable performance reduction. For my setup at least, my guess is any further legibility improvements on cockpit buttons or EFIS will need to come from fixes and improvements to the sharpening filter and TAA interaction (or even just removal of the sharpening filter at high res !? ) rather than additional super sampling on top.
I’ve updated my comments in my post above: