You’re forcing motion reprojection twice separately in WMR and MSFS - those are separate. MSFS reprojection isn’t very good at all. I suggest you turn it of, especially if you’re using WMR MR anyway. On the surface, reducing resolution and then upsampling with DLSS and then increasing WMR resolution doesn’t make much sense, it should be about the same (only with added motion artifacts) as not reducing resolution at all, and then not increasing it, like using 100% TAA + 100% WMR, which is what I do. Using “let Windows Decide” is also strange - you’re just not controlling “quality” or “performance” and in my experience “performance” really affected the visuals. If for some reason you prefer “quality” you should set it specifically. But, MSFS works in misterious ways, so if it all works for you - great…
I didn’t think there was upsampling in DLAA. And, I thought WMR MR and Ingame MR complimented each other, not either or.
I am going to test the OXR 200% setting with OpenXRTK disabled. I think using DLSS Performance with OXR 200 has a similar affect.
DLSS = Deep Learning Super Sampling, it is reducing resolution (and GPU load) and upsampling by definition. “Performance” and “Quality” are reducing resolution by different amounts. DLAA = Deep Learning Anti-Aliasing and can work together with DLSS to do a more convincing/sharper edges with less shimmering, to reduce upsampling artifacts. DLSS+DLAA doesn’t mean it’s not reducing reduction and upsampling, it shoudl still do that. I’m not sure why MSFS hase performance/qualiry/DLAA grouped in one selector, they are controlling different things. So by using DLSS you are reducing resolution (and GPU load) and upsampling, but with OXR 200% you are increasing resolution (and GPU load). Those are just two actions working against each other. Not sure of the exact math and if they turn out to cancel each other completely or only partially, but they are working one against another, both introducing artifacts in the process. But if it works for you - it works. I hate DLSS motion artifacts on numbers, glass cockpit and moving objhects outside, and if the result is similar to just using TAA 100% / OXR 100% that makes much more sense, and no added artifacts or resampling.
Motion Reprojection in OXR and MSFS are two different engines. There’s no sense to use both. MSFS MR is generally not good. Depth is not supported anyway, I think. OXR MR is actually very good, it’s the only way I can fly, but it intriduces motion artifacts - wobbles. But makes everyrhing very smooth. And lately it seems to generate more wobbles, maybe due to the latest OXR runtime. But the great mbuccia has hinted an upcoming major improvement in that area, hopefully soon. Anyway, no sense useing them both, it could only make artifacts worse. If you like MR (like me) - use OXR, and disable MSFS one.
The two methods, with and without OXRTK, are very similar. OXRTK method gives a “windowed” or “large print” view of the same quality. The non-OXRTK method is superior because you get the native resolution without the window affect. From what I read, both methods may use the same API that allows the solution to substitute a fake buffer for the one the sim thinks it’s going to draw to with the buffer that will be used for scaling, AA, etc. I think OXR 200 just tells the sim to create a “really big buffer”, which gets handed off to DLSS (or OXRTK NIS Scaling). Something about that “RBB” makes the pipeline run so sweet.
So, the Openxr 200% idea was complete placebo. I made the mistake of testing two settings at once. No openxr changes are needed to get smooth results in the G2.
If you’re replying to me, I’m not talking about AXRTK at all in my last reply. I do not mention it either. OXR I refer to is Open XR (Developer Tool), not the Toolkit (OXRTK). OXRTK adds NIS/FSR upscaling which I’m not talking about here, and I’m not using it since I got the RTX 4090. I do use FFR and removing the black mask. I also have Turbo enabled and forse MR to 45 to compensate.
Totally agree and verified, by disabling hyperthreading the performance in VR goes up a lot. I am using the Process Lasso tool to disable it only in MFS every time it starts, so you avoid disabling it for all windows processes. Everything runs much smoother and I can turn almost all settings up to ultra. I thought I’d have to buy the next generation rtx 4000 to cabar with my frustration with this simulator in VR, but after doing some tweaking and following some advice on the forum with my rtx 3090 everything is running smooth as silk. It’s a big relief because I still see people with good computers with performance issues and I think this thread solves a lot of them. I also agree as many comment that the performance improves by increasing the resolution in VR, something paradoxical because many think it should be the other way around, but with powerful graphics cards you are no longer limited by the processor and everything works much better.
Which CPU do you have?
I have 13900kf and I’m guessing if disabling Hyperthreading would help.
This CPU has 24 cores, 8 are Performance, 16 are Efficiency.
Hyperthreading words on the Performance cores only (total threads is 32), so in theory it should work, should it?
May I also ask how you disable HT via Project Lasso, for the thing to be FS2020-related?
My computer is a Ryzen 5900x (12 cores- 24 Hyperthreading), rtx 3090 graphics card, Nvme 2Tb gigabyte hard drive and 32 Gb ram 3800 Mhz (14-14-16-36-52-1T). All standard, I do not use overclock on any component. My VR viewer is the Htc Vive Pro 2, I use ultra quality at 90 Hz, motion compensation in the Vive Console options, that same option within the simulator options does nothing for me, but in the Vive Console options it practically doubles the fps. (helps a lot in the fps and unlike other people, apart from the menus, I have a small “water dance” that is barely noticeable, I have no problem within the simulation) and 150% resolution in SteamVR settings, 4048x4048 per eye, a barbarity that nevertheless goes smooth as silk, apparently the bottleneck disappears completely by demanding much more to the graphics card. Now I am really enjoying the simulator, even in heavy aircraft like the Fenix A320. I know there are many things that influence performance and many different rigs, so what works for me may not work for others, but until a week ago I was truly frustrated and seeing some comments even from people with bad performance and a rtx 4090, I thought I should wait for the rtx 5000 series to truly enjoy VR in the simulator. Everything has changed and only with tweaking a few options. Raising the resolution of the VR viewer (maybe that makes you not so CPU limited) , disabling HT and enabling motion compensation in Vive Console (the latter is what has helped the most) has made a drastic change for me at the performance and smoothness level. By the way, all set to ultra in the simulator and use DLSS mode in quality mode, even TAA has very good performance with these parameters. In my case there is not much difference between using DX11 or DX12, maybe in DX11 everything goes a little smoother. If instead of ultra I use the high mode everything goes much smoother, even at ground level and there is not really a big appreciable graphical difference in VR with that change. Sorry, my VR viewer is not a Reverb G2, but I saw the thread and I found it very interesting…I do not know in detail the difference between Reverb G2 and Htc VIVE Pro 2 or the options that manages a viewer with respect to the other, although I think they have similar performances but at least I hope to have helped with my contribution to some people to try other parameters and can improve their experience…
@RomanDesign I have killed them all. And not just the men windows processes, but the women and children windows processes too… :))))
Anyway…I would like to add my 2 cents to the stuttering issues. There’s 2 things above all that make the most difference. At least with WMR headsets.
-
Use high OXR resolutions. I personally go with OXR 150% so i end up with around 3800x3800. I have almost no stuttering at these resolutions with rtx 3080ti / i7 9700k. However at 100% OXR i stutter heavily. No doubt because at 100% oxr i get cpu bound while at 150% i’m gpu bound. This game absolutely HATES being cpu limited. And from what i’ve seen the CPU optimization has actually gotten worse since SU11.
-
I can’t emphasize this enough. I can’t give you links because it’s prohibited to use them here, but please try the google maps mod. For some reason it doesn’t only look better than bing maps, but also runs better. Must be something to do with the servers msfs is using to stream map tiles that google maps servers just do better. It doesn’t exactly lead to more fps, but it does produce much less stutters. I’m sorry to moderators, i hope it’s not wrong to talk about this mod (without posting links to it).
-
I think msfs will never really be 100% smooth no matter what gpu you use. There’s just somethinga about the code that’s just complicated beyond repair. At least, that’s what i believe from experience with other games that have suffered from the same problems. Such as No man’s sky. It’s not that it hasn’t gotten better. It’s not that you can’t make it VERY VERY playable. But it’ll always have a stutter here and there, a sudden fps dive for a second, something… So yea. That’s what i’ve accepted with msfs and i will keep my expectations low in that regard. The reason i’m saying this is… I’ve ALMOST gone and bought a 13900k because of msfs and i changed my mind at the last second. There’s literally not a single VR or non VR game that would have ANY kind of problem with my cpu/gpu combo. I can run everything at 100% satisfactory levels. MSFS is basically the only thing out there that bottlenecks my hardware. I’m not gonna go and spend nearly 1000 euros on a new plate + cpu and make life easier for Asobo by throwing money at this thing. As you can see people with that kind of hardware stutter a bit as well. If they ever manage to optimise cpu usage in this game as much as it would be necessary, you can be sure it won’t matter much if you have a 13900k, a 2 year old i5 or whatever amd, it’ll be fluid either way. I will agree the latest cpus do make sense for the rtx 4090 owners, but not to anyone else.
Cheers mateys
What CPU are you using, and what are your typical FPS with these settings, and typical CPU and GPU frametimes if you have them?
I was also experiencing stutters both in PC and VR mode. Every few seconds. The culprit was ASUS GPU Tweak III which I used to OC my GPU. When I exited out of that program the stutters are gone. Even when I used the ‘default’ setting in ASUS GPU Tweak III I had stutters. Good luck!
I’m getting a solid 45 fps with this setting. I also tested this on DCS with OpenXRTK, and the performance improvement is the same. They don’t have DX12 so no DLSS for them, but NIS Scaling in OpenXRTK seems to benefit from this setting regardless of sim.
I have the i7 11gen cpu.
I think this is placebo effect now. I tested without any Openxr changes, and performance is excellent in VR with both DLSS Performance and OpenXRTK NIS scaling.
I am having stenciling and stuttering issues and poor framerates lately with PMDG B737, but all other “stock” aircraft, I seem to have doubled my framerates by turning off hyperthreading. (Using a RTX2080 Super)I went from 30 to over 70fps. I just bought the Fenix A320 and I am going to make a flight now and see how that goes. Also the USB disconnect I kept getting in my Oculus Rift S seem to have stopped, so far. Not sure if hyperthreading has anything to do with VR headset performance or USB power issues.
Since going to 130% resolution in OXR with DLSS and turning off photogrammetry I have experienced only 2 tiny stutters over 3 x 90min flights.
The resolution increase is dramatic in its improvement, the sim doesnt feel like it is struggling any more.
I also think the stutters and occasional fps drops are related to networking - the cloud streaming of scenery is one of those things that you just know is going to be slightly dodgy and intermittent and ‘just about working and no more’.
However, that OXR resolution increase has so far been a one and done tweak that I have not had to go in and change any other settings. Nvidia DLSS combined with 130% OXR resolution has delivered MSFS to the VR promised land as far as I’m concerned.
This is the first time I’m more interested in spending money on the marketplace than on hardware to run the sim.
EDIT specs and info for anyone else struggling - RTX 3070, AMD 5600x at 4.78ghz, Win10 (went back from Win11 which is an abomination)
OXR 130%, DLSS performance set in sim
No OXRTK
No Motion Smoothing
I have also now killed all Windows updates using the Chris Titus powershell debloat scripts to lock this system down as much as possible, I dont want anything changing.
happy for you. I know how good it feels once you get it right. However, i absolutely can’t understand why you’re not using OXR Toolkit. 
There’s not a single downside to using that app, beside incredible gains. If for nothing else, at least to calibrate brightness, colors, etc…I strongly suggest you reconsider this choice. I personally use its FVR function, because it produces very significant framerate gains, i use CAS which goes along really nice with DLSS and produces an even bigger clarity and i just couldn’t live without its brigthness and other corrections. Also, the remove mask feature is incredible as you gain more FOV at (seemingly) zero cost in performance. And i also set world scale in OXRTK since that way at least it doesn’t get reset like the ingame setting (unless that was fixed). I do however use MR i should add so that part comes in handy as well. Since you don’t, you might explore its new experimental TURBO function (doesn’t work well with MR) but some users suggest it gave them massive performance boosts.
I do have a reason for not using it anymore and it’s this - it’s another layer on top of an application that is changing underneath it, on top of an operating system that is changing underneath that.
I can mitigate the OS changes to the extent that anyone can, I can do nothing about updates to the sim, but I don’t want anything else that is likely to be sensitive to any updates in ways that are unpredictable.
I’m delighted that the toolkit is working for folks.
This made a big difference for me as well with my 5950x. I have 32 cores so I can spare 16 lol. But I still had a lot of stutters, particularly around real busy airports… so I tried something yesterday and it totally surprised me. I went in to my BIOS and did an all-core clock frequency lock at 4.5GHZ instead of PBO boosting randomly at 5GHZ. SMTT/HTT is still disabled as well, I can’t believe how big of a difference this made. It really seems like MSFS performs a lot better with stable clock frequencies and not the ups and downs with PBO.
Anyway, it wont matter soon… because I have a 5800x3D on the way. I’m frustrated I didn’t try locking all cores months ago, but maybe it will help somebody here.
That would have been a valid argument if that layer would tend to negatively impact a large share of users. Since, however it doesn’t, there’s very little reasoning behind your arguments. Of course, it’s your choice, i just feel you’re missing out for no reason. Also, if you ever need to make sure what the updates to the windows/drivers/msfs have changed and want OXRTK out of the equation, you just turn it off. Literally with a single mouse click. Cheers
I think my reasoning is sound, there will always be a number of users negatively affected by a change to the sim or OS, I want to reduce the odds of me being one of them - now that I have it running well and looking good.
With VR being as sensitive as it is already, and given that I have it running superbly now, it follows that anything not native to the sim or OS operating in the domain of VR that can be removed should be removed.
i’m sorry, but that’s horrible reasoning. If that was true, no one would ever run a single mod in any game. It’s not how it works either in this particular case. As mbucchia always tests all releases in regards to latest game updates. Another argument that is important and is very specific in this case is …the person adding this “additional layer” is actually an official microsoft OXR developer. If he doesn’t know what he’s doing, then no one is 
I am terribly sorry if i come out a bit “intense”. I really don’t mean anything bad, i am just strongly convinced in this case that you are completely mistaken. Did you even try OXRTK yet?