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

How interesting this little experiment is:
5 times in a row I’ve been able to repro and solve motion smoothing not kicking in!

The last test flight was even more surprising to me: I wanted to compare the TAA60+OXR100 I was using with GA light aircraft with the FBW A320. Same spot, same conditions.

As soon as loaded, once again rubber banding showing up. I thought it worked already 4 times, I’ll try once more. So I raise TAA to a level I know it would disable MR entirely (that’s the trick) and in this case I chose 80, apply, close settings panel and as expected I lost MR. Then I open settings again, revert back to TAA60 apply, close settings panel, and voila: motion smoothing kicking in with the FBW 320 at TAA60 + OXR100.

Sure, with 60/100 it is not giving good EFIS either so I try raising this a little bit to 65 and this is starting to get somewhat noticeable better (it is not TAA100 though). I decide to take off like this and this went fine even if live weather was low ceiling and raining.

I keep climbing a little bit and I decide to check the view, so I open weather settings, choose scattered preset, and this starts to look quite good overall.

Then, I wanted to see up to which point I could raise TAA and keep motion smoothing with the FBW 320 and OXR100 (still using my latest base settings but without contact shadows, with medium clouds, and for those picking up this topic by search, still with a 9700K+2070S). And I couldn’t believe I was able to raise the resolution to TAA85+OXR100 while in air climbing over the country side in Pays Basque! I even have been able to raise Clouds from MED to HIGH, but at this resolution motion smoothing was quite often giving up causing an apparent stutter (but not rubber banding).

After trying a few different settings I’ve found TAA75+OXR100 almost stable with good EFIS legibility, and TAA70+OXR100 stable with slightly less good EFIS but just slightly (meaning moving my seating position 5 cm forward and this was looking the same as TAA75 but 5cm back).

I was sill surprised this kept motion smoothing running fine with such settings, so I decided to open OXR Dev Tools, change motion smoothing from ALWAYS to AUTO, switch from/to VR mode, and compare how it would go down to landing the aircraft back to LFBZ airport.

Well, it went perfect down to the runway, looking left the bing scenery was smooth, looking right the @thalixte photogrammetry Bayonne was smooth too! Only when on the ground and taxing it was again loosing motion smoothing from time to time but it never went back to rubber banding.

As a conclusion I’d say: try this!

  • if it is rubber banding, just dial a larger TAA so that it disables motion smoothiing, then dial TAA back and this should kick in motion smoothing for good.

  • my settings are really solid and they are high enough for a visually pleasant experience (more capable hardware of course would raise some of the them). I’ve tested them with:
    FBW320, TBM, GA like PA28, KA350, Longitude, C172 Analogue

  • you can play with Clouds first if you can’t keep up the fps (HIGH → MED is very effective on the spot), you can also play with Contact shadows OFF/LOW depending.

  • you can run FBW 320 with TAA70+OXR100 on this hardware. This is comparable to TAA100+OXR50 for the EFIS but it is OXR100 for the outside!

Quite an experiment today!

8 Likes

Very interesting!

I decided to run a few tests at OXR100 as a result.

I started at TAA75 and spawned on the runway at Laguardia in the Justflight Piper Arrow IV (all settings as before but reflections back down to low). Surprised initially to see awful performance , 12-16fps and constant red in the frametime overlay. Bumped up TAA to 120 and fps didn’t really change but then dropped TAA back to 75 and all of a sudden it settled down to mid-20s and by the time I was in the air it was a pretty constant 30fps maintaining 3:1 MR. I then tried to push it and found I could fairly happily maintain OXR100/TAA90. One downside - the whole MR weird propeller thing was way more evident in a plane with the prop right in front of you compared with the Baron!

Tried the Baron to get a direct comparison with previous tests. Similar story but no need to do the TAA dance. Locked in stable at OXR100/TAA75 and could again push it to TAA90 once in the air. EFIS very readable but still suffering from that slight MR shimmer.

Didn’t get quite so good results with the FBW A32nx. At OXR100 I had to drop TAA to about 80 to get a stable experience. Definitely prefer OXR85/TAA100 in this scenario and would probably still end up going with OXR100/TAA100 and MR off just for the extra bit of instrument clarity.

At the end of the day I feel spoiled for choice now!

1 Like

@NinjaCanOpnr
Thank you for your feedback!

The “TAA dance” trick is peculiar and cumbersome, but I was glad finding this trick yesterday*.

As for the propellers, I’ve added the link to the solution a few minutes ago directly in the “Bonus” chapter of my 2nd post here: My 2070 SUPER VR settings and suggestion (Reverb G2 - WMR) - #2 by CptLucky8


*I have no clue why this is happening specifically but I’d bet a bug of some sort in the WMR OXR driver pertaining to frame timing which could be not init properly if you start with MR activated (i.e. it would take non-MR → MR transition for the code to “init” properly). This is just speculation though.

2 Likes

Cap, thats great info on the TAA increase decrease trick. I did try with my barren test below, but it didnt help much, i think though in more dense areas it may be the trick, will try later.

Shame theres no easy way to record a video snippet of the rubber banding so everyone knows what that looks like. Maybe a through the eye (goggle) view perhaps.

So i did my newest test, i started on my basically clean win10 image (21H1). Still using the old trusty 457.30 nvidia driver… clean install of fs (so 111gb not the full size because didnt do the added content manager items yet), ran my same barren area test in b58, looking left tilting and forward with some up and down.

I tried oxr 50/ taa 85
Other settings:
TLOD 100
Vector Ultra
Buildings High
Trees High
Grass and Bushes Med
OLOD 100
VOL Clouds High
Text res ultra
Anis 16x
Text super 2x2
Text synth high
Water high
Shadow 1024
Terrain shad 512
Contact low
Windshield high
Ambient occ off
rfl low
Light low
Bloom on
Cockpit refresh high
(i’ve tried lower on most of the above in previous tests)

First run it was smooth, didnt notice any ground stutter or banding during the movements.

Then i switched to my full drive, used the same 111GB or so clean install folder (same settings, but the newer studio driver).
Did the same test, i could see some rubber banding (not a ton) but more than the clean.

So i decided to install the 457.30 driver instead of the newest studio driver… maybe slightly better (placebo)?

Now i went back to the clean windows… did the same test, this time i could see more stuttering than the first clean round, no idea why, nothing horrid though.

Its a toss up, not sure what the deal is, i still “think” the cleaner windows is slightly better, maybe even the nvidia older driver but unsure (at least with my 10900k at 4.9ghz, 3080ftw3, system).

1 Like

One way to describe it as well is looking at workers and airport traffic animations. They animate normally for a second or so, then slow down, almost freezing, then resume the animation at normal speed in a burst, then the cycle repeats. At least this is how it looks on my end, hence why I’m also assuming this is something with regard to the OXR API “time” reference and how FS2020 is using it, otherwise it is just that it takes too much time to compute and makes FS2020 skipping the frame.*

I was thinking about this yesterday as well! How would the trusty 457.30 run with SU5… so little time per day though! I’m not sure it will change much and I find the 471.xx working good now in VR (didn’t try the latest though)

I know your video card is more powerful than the 2070S but I’m certain there are still a few settings affecting too much the outcome regardless of the video card, only because they are affecting other areas of the rendering pipeline which are not directly bound by the pixel shader throughput alone. In addition, some settings are changing the pixels in such a way it makes it harder for motion smoothing sometimes between frame.

You might really want to reconsider these for example and compare:

  • Vector (OFF), Tex. Res. (High), Tex. SS (0), Vol. Light (Off), Bloom (Off)

*However I have a hard time thinking the game would base its internal update clock based on the rendering of the frames, because it is certainly not the way you’d code the main game loop update in any game, instead, you always code a frame rate independent update loop for obvious reasons. Having said this, I was already reporting the following a long long time ago:

Description
GUI transitions (fade in/out) timer appear to be dependent on the actual in-game FPS.

Expected
GUI transitions based on a stable/reference clock.

Comment
This introduces annoying delays when trying to switch option panels quickly, and makes the interface/sim feel un-snappy. Quicker transitions would provide a better feeling of ‘speed’ even on lower end hardware setup.

2 Likes

I will give reducing those settings a shot (pretty sure i had tried, well maybe not all of those at once).

I originally thought it didnt matter based on the results of this thread:

I’ve actually posted a question for the Q&A and in one of my post I’m explaining why this guide is not enough to me:

Having said this, we all know FS2020 has its own temper as well from time to time, and no matter what you try, sometimes it just fails miserably :slight_smile:

1 Like

Valid points on the VR, i figured, as that guide is more geared for 2D. Ill check out the other thread and bookmark :slight_smile:

Well with all the data analysis you have done which is awesome and thank you btw, i sure hope you have some pull with the devs to make things better for vr :slight_smile: Or maybe once they get to the next DX level, VR will improve and have better frames. It can behave nicely at times though like you say, then other times you bang your head off the wall trying to figure out what has gone wrong.

3 Likes

I thank you for your kind words!

And to illustrate this, I’ve just tried the same spot I’m doing for the last 3 days (LFBZ) and now I can’t get the FBW 320 running with MR at 70/100, only at 65/100… :frowning: As for the “pull with the dev”, I’ll send you a PM… In the interim you might want to read this: The marketplace desperatly needs quality control ASAP - #181 by CptLucky8

Ok I’ve forgotten yesterday when testing I’ve changed MR from ALWAYS ON to AUTO… Now back to ALWAYS ON and much better for me on the ground!

Good to know. Just trying these finds of yours now. I was able to get MR to lock in at OXR 100 / TAA 70 by raising to TAA 85 then back to 70.

Took off from Megeve in the 152 again. Quick loop around town and touch down, taxing and again CTD. Something about that airport…

Sim back up now. Going to continue playing around with this. But for me, clouds high is as low as I care to go. I’ll take a hit in resolution to accomdate that.

At least I was able to see simar results from raising and lowering TAA.

1 Like

And I’ve finally made a flight earlier with the FBW 320 from TFRR to TNCM with OXR100 + TAA70 + Clouds HIGH all fine, and of course now with the TAA dance trick to kick MR in :slight_smile:

I’ve had a small hiccup about ¼ the travel where I needed to do the TAA dance trick again, but I was also fiddling with the settings constantly switching interior/exterior view to see the changes and it might be I loaded too many things at once too fast. After that it settled again and all was fine for the rest of the flight, even if this was cheating because most of the flight is over the ocean :rofl:

Nevertheless MR stayed in during landing roll and taxing to the gate.

I don’t mind with an Airliner lowering the Clouds if needed on the ground on departure or even on the final approach if this helps keeping MR kicked in, especially if I trade clouds for TAA70 + OXR100 which is surprisingly quite good even for airliners EFIS in my opinion now!

1 Like

Yeah I ended up spawning in on approach to LFLJ in the fbw320 to test a complex plane. Ran great OXR 100 / TAA 65 clouds high. V smooth. Great to see the landscape so defined too.

Tickle me convinced. You’ve found something here.

1 Like

Something you might be running against.

Im only familiar with steam VR with valve index, so i don’t have an idea how its called in your VR software, but in steam VR its called “throttling behaviour”. Its where you set the parameters for motion smoothing. There are two sliders. One for the fixed fps and the other is called “Additional prediction time”.

I have figured when this value is set too low or too high, its causing stuttering and wobbling. Setting it properly resolves these issues. In example currently im using 120 Hz with fixed MR @ 20fps and 33ms additional prediction time. This almost completely eliminated wobbling and stuttering for me.

Also consider increasing max amount of prerendered VR frames in your NVIDIA settings. Its set to 1 by default. From my understanding it intereferes with MR, as it does not allow your CPU to prerender more then 1 predicted frame. Use 3 or 4 instead.

2 Likes

@AlexejGeldt
There is no such throttling and prediction settings with WMR/OXR. These are only in SteamVR and only if you’re using the StreamVR compositor (exit Vive for example).

As for NVidia VR Frames, since day 1 the overall consensus has been to using 1 to get the best performance and stable fps (less spikes). Some are reporting sometimes they have better results with 2,3,4 but most that I can read feedback from are finding it is better for them with 1, which I recommend too (both for Index and G2). It is possible it is video card dependant as well, or VRAM dependant?!

The most important setting for reducing “wobbles” and for stabilizing fps that I’ve identified is Multithreaded Optim. OFF (SteamVR and/or WMR).

NB: prior SU5 finely choosing VSync related settings where positively impacting the experience too in VR (see my older recommendations) but since SU5 they don’t seem to be having the same influence anymore (which is good and expected), and therefore I don’t recommend to change them anymore either.

The TAA scaling dance trick with WMR is a workaround to kick MR in, working for me, and now at least also working for @Jonny5Alive6618 (and I hope this helps many more people too!)

1 Like

I noticed this too, I didn’t understand really why but sometimes it would be very choppy in vr, and just adjusting TAA slightly has helped for me. It’s like everything gets re-drawn correctly after a small change and then settles in.

I really hope the devs can fix this temperamental VR, it shouldn’t be like this.

2 Likes

@CptLucky8 OXR 100 / TAA 65 is my number. I’ve always noted my res has been 5 to 10 below yours since the beginning. But everything else seems to scale well. The only difference is my i5 9600 at 5gh to your i7 9700. And my desire for high clouds.

Photogrammetry seems to account for about a 5 or 10 hit on the render scale, if you will.

Explain it like I’m 5;

Why is 100/50 - 75/75 - 50/100 a more efficient combination than any other combo?

1 Like

Please explain what everyone is referring to with, e.g., “TAA at 80”?
Don’t you all mean Terrain Detail? Am I not correct that TAA is not adjustable, other than selecting an antialiasing mode such as TAA, FXAA etc.?

They are talking render scale value, though frankly im not sure where the TAA part comes from :slight_smile: I just accept it

1 Like

I’d suggest you start with the 1st post for all the details. It is just a convenient way to “tag” each resolutions we’re talking about (there are several different to adjust).