WipeUout

Use the latest PimaxXR (ver 0.4.0), there is an option to switch compulsive smoothing on/off when you open the app. null

I guess you are using DCS-ST with SteamVR? If not, then Reshade will not work with OpenXR, it is not compatible.

If you are using DCS-MT, it does not require opencomposite anymore since DCS has native OpenXR support. Opencomposite also reduces FPS performance significantly. Have a look at the first two posts here:

After second thought, it makes more sense. I am using wide FOV (160 degrees H) and it would take a much more than 30% bigger image to match the projected image through the lens. Good thing that this operation to apply barrel distortion correction is done by shaders, rendering an image at 190% would seriously reduce the FPS to less than 25% of what it is in 2D.

I don't understand why the PimaxXR reported resolution is lower than my native? I am at "1" everywhere (Pimax client =1, OXRTK no override, DCS PD=1) I have a PimaxXR reported resolution of 5012x3160 (15.8 MP) which is a bit lower than the physical resolution of the 8kX 4k panels x 2 at 7680x2160 (16.6 MP). Unless the reported resolution is per eye? This would mean that the upscaled "for barrel distortion correction" is 190%?

Getting the same problem with my 8kX. Smart Smoothing has never been an option, it is not good and the transitions are just full of stutter, better off without it. Unfortunalely compulsive smoothing option has been removed from Pimax client but if you use PimaxXR 0.4.0, you can switch it on there. I did not try it but it was working very good at 1/2 back then. Pimax FFR does not work with the OpenXR runtime, it will only run with SteamVR. You can use the FFR in OXRTK instead. I have a PimaxXR reported resolution of 5012x3160 (15.8 MP) which is a bit lower than the physical resolution of the 4k panels x 2 at 7680x2160 (16.6 MP). Not sure what this number stands for. Are you super sampling some where?

Did the test and here are my results with OXRTK recording feature: DCS Setting: preset VR, PD=1.1 OXRTK: CAS 100%, FFR performance preset Average FPS : 90.48 -> This is does not translate the accurate GPU load as the HMD refresh rate is at 90hz. The reported FPS will stay at 90+/-1 FPS even if your frame time is less than 11 ms. What 90.48 just means is that my GPU has "headroom" and it can pump those frames faster than 90 FPS. Lowest 86 FPS, max 91 FPS. Average CPU Frame time (app + rdr): 4.27 ms -> this equates to a CPU average capability of 234 FPS. In other word, my CPU has lots of headroom as it can pump up to 234 FPS, if the GPU can swallow it... which is not entirely true as it is multithreading and the graphic pipeline can't use all threads. Let's just say that the CPU is not a bottleneck here. Average GPU Frame time : 8.76 ms -> this equates to a GPU average capability of 114.2 FPS. This translates in what is the real average FPS that the GPU can pump. Looking at your data you have a running test time of 108.641 seconds and it is a bit longer than the actual track, which last 106 seconds. This means that you are capturing performance outside the track run and thus your real performance might be better or worse depending on what is going on during those extra 2.6 seconds. I don't know about MSI afterburner to monitor but I trust OXRTK which is a VR application. Using OXRTK recording feature will generate a csv file that can be use to calculate your averages, you just need an spreadsheet program. Since I start the recording a bit before I press "FLY" (to open OXRTK menu, toggle recording on), and stops a bit after the end of the track (open OXRTK menu, toggle recording off...), I remove a few seconds (lines) in the csv file at the start and at the end to make sure I only use 106 seconds of data. The recorded data outside the track run is easy to spot. Rigor in testing is crucial to ensure meaningful and accurate results, it's my engineering background probably... Now analyzing your results and taking into account those extra seconds, you definitely have lower performance than expected. You system should be 20-30% faster than mine (GPU wise). You should get an average GPU frame time between 6.8ms and 7.3ms. I also noted that you wrote about the infamous "Tacview" which is know to slow down the sim considerably in some cases. I personally uninstalled it 2 years ago, you might want to double check this. Re-installing windows might help a lot. If you go this route, have a look at this before to ensure a "fully optimized for game" Win 11 installation:

I'll do a ''default DCS VR settings'' test once back home later and post results. It will be interestiing to compare since we have the same pixel count... My 4080 results should be lower than your 4090's... Do you have your GPU frametime results? Sometime the FPS does not give the real picture as far as GPU usage.

Syria is a good test for a high end setup. For me it is strictly and always the same track in Caucasus with the FA-18 (attached). It is time consuming enough to have to optimize my system, one track file is enough. I keep my results in an excel file and have over 250 recorded test results since OB 2.7.9 only, and probably twice that which did not make it in the spreadsheet. Too much testing, not enough flying! No eye tracking with the 8kX, I am limited to fixed foveated rendering (FFR) but it is nonetheless a huge gain. As for my settings: DCS. I use low/med textures whenever on the Syria/Mariana/Falkland's maps and high/high on Caucasus/NTTR/Persian Gulf. I use med water, flat or low shadows and low clouds everything else is off. I crank up the pixel density to 1.1 to help a bit more with aliasing since I don't use MSAA and my HMD output is very clear. 21% increase in resolution (PD=1.1) is not as good as MSAA 2X but much less taxing on the GPU. I lower the gamma to 1.1 on day mission to eliminate as much as possible the haze but to compensate, my backlit and brightness is cranked up quite a bit in Pimax client. Pimax client. Everything at normal except the brightness as stated above. I use 90hz refresh rate on less demanding maps and 75hz on the demanding ones. No smart smoothing, no FRR either as it does not work with OpenXR, only with SteamVR. Thank god we have FFR in OXRTK! OXRTK. I use CAS 100% to have the crispest possible image and of course FFR at performance setting. A bit of brightness and that is pretty much it. For me the most important is avoid stutters and ghosting which is killing my immersion, the main reason why I sim in VR. I wish my rig could sustain 90hz in all scenario but I guess I would need a 4090 (or even more) for that, maybe later next year. I posted that before but I feel that I am not exploiting the full potential of the 8kX with the present hardware capabilities on the market. For now, I don't think I will get another HMD soon, a new GPU will come before that. performance track Caucasus F-18.trk

If you are testing with the Mariana map, it is very difficult to achieve good FPS. My test track is using the FA-18 on the Caucasus map, which is about the "average demanding" map. The results are quite different when on South-America for example, I have to drop my refresh rate to 75hz. I do not use MSAA anymore as the perfect cure to aliasing is high resolution. Aliasing (for me) is usually bad on the ground but very tolerable once airborne and this is with the 25PPD of the pimax 8kX. The Crystal has 35PPD which is much better and should have very little aliasing. Of course, personal preference and tolerance level can vary a lot between individuals.

Dismissing Pimax from 2018 hardware! What about the Crystal or the 8kX? considering that he is making this statement from a DCS sim perspective!! No doubt that the Quest pro is a very good HMD but not everyone want to fly with the FOV of a pair of binos... When you state that this HMD is the best, you need to consider the whole crowd.

I did some testing a few weeks ago to see the loss between VR and 2D. Running the same track with the same settings give me a 125-156 FPS in 2D using a 4k monitor, compared to 56-67 FPS in 3D using my Pimax 8kX at 4k per eye. This will vary a bit depending on the settings used but as a general rule, you can expect your 3D performance to degrade to 42 to 45% of what you get in 2D for the same settings in the same scenario. Since the Crystal and the 8kX share the same number of pixels, you should get similar results unless you have some settings that only apply to VR (or 2D) that could explain the difference.

My understanding is that the correction for barrel distortion is adapted depending on the lenses, actual panel size and geometry used. What you state means that the simulation rendering pipeline would need to know the particulars of each HMD in order to apply this transformation. And also, barrel distortion is more akin to an upscaling operation that rendering an actual image. If that was the case, you would have about 1/3 of the FPS in VR compared to 2D at thew same resolution, what we have is more between 43% and 47% of the FPS.

That is an 8 months old video. Wonder if he can say that now. He states that he tried a pimax 5k+ and 8k (released in 2018!) but did not talk about the 8kX? There are always holes in those youtubers videos.

Just going for FFR on my 8kX shreds 1 to 1.5 ms on my GPU frame time, DFR will give even more.

Then it seems that your GPU can't keep up with the refreshrate of the HMD. You have three possible approaches: 1-lower your settings, 2-lower your HMD refresh rate (I think that the Crystal does not allow lower than 90hz), 3-Upgrade your GPU.

Unfortunately, there is no ''standardized" benchmark for DCS. And the infinite possible hardware combination out there makes it very difficult to compare, not to mention all possible combination of airframe/terrain/mission type and settings you can use to bechmark. I suggest you set yourself a couple of tracks and create list of results and go from there. The best and most meaningful comparison is againt your own baseline. This will be usefull to see performance and quality effects cause by settings/software/hardware changes. You can also always search the forum for "similar systems' performance, most of us do post them time to time.

On my system, still no significant difference between HAGS on or off. In my test track (F-18/Caucasus), I get a 0.02ms difference in GPU average frametime (8.62 vs 8.64) and a difference of 0.05ms CPU average frametime (2.78 vs 2.83) which is within the margin of error.

I think motion blur is overrated. Better to use motion reprojection, same effect but at least you have a smoother experience...

I don't think it is related to smart smoothing. Never seen this with my 8kX, but it could be different with the crystal. It looks more like a power glitch. what is your setup? Could it be that your PSU is not keeping up with the power draw?

Well in that case, I would suggest to start with low settings and move your way up. I am not familiar with the reverb g1 but what I suggest is to try to avoid using motion reprojection and aim for native resolution plus some upscaling, the 4090 should be able to handle this.

Totally disagree, I can play in my 8kX for hours without any pain or discomfort other that I can be a little hot. Maybe they goofed with the crystal but that is not the case with all Pimax HMD!

Are you using motion reprojection? That could explain the "jitter", if you see it only on the sides. You should expand on what software setting/version you use, there might be a problem there also. You do have a high end gear and your experience should be silky smooth!

Some additional findings: Combination of PimaxXR 0.3.4 with Pimax client 1.11 offers the smoothest ride. Client 1.12 and 1.13 are not performing as fast as 1.11 and cause in some cases a 0.5ms increase in GPU frametime. If your sim with your HMD refresh rate at 90hz, then PimaxXR 0.4.0 causes some sporadic stutters. Reverting back to PimaxXR 0.3.4 solves the issue combined with Pimax client 1.11 Pimax client is hard to find, here is the link: https://drive.google.com/file/d/1uj9v26nPSkSe2Og4JPMIzxNIgOnlcFCl/view?pli=1

Same here, I did tests from a 4k desktop resolution down to 800x600 with a 0.1 ms GPU frametime difference, which is within the margin of error. No change in performance.