some1

What shadows settings do you use? Because these are "default" aka highest quality terrain object shadows. These work as expected, but most people in vr use flat or off.

You have the same shadows bug that everyone else. Note that the shadows in your 2.8 screenshots are smoother, while the 2.7 shadows are jagged. You've also lost the shadows under distant vehicles. The weird shadow glitches mostly appear at certain angles with terrain object shadows set to OFF, and you seem to have them set to FLAT? Posting 80 fps screenshot when your fps is locked to 80 in both cases is completely meaningless.

25% drop in cpu performance is larger than what we gained in the last three cpu generations from Intel and AMD. In other words, we would need several years of hardware development to close the gap that ED manged to introduce with a single patch. I hope it's just the bugged shadows that cause current performance issues.

Even on the thumbnail you can see 10% increase in gpu frame times, cpu frame times went up from 10.5 to 13 (more than 25٪) and his fps dropped by 8 or 15, depending on if you look at momentary or average fps. How's that an insignificant drop?

Added MB-339 from India Foxtrot.

It's only on the openbeta branch.

Here are my numbers on 2.8 https://forum.dcs.world/topic/311178-large-performance-drop-in-vr-after-updating-to-28/#comment-5074255 That's even more than 20% drop in the spot I compared.

This is Hornet Ready on the ramp instant action mission in 2.8, with RTX4090, Ryzen 5800X3D and Reverb G2. This is the same spot in 2.7 I made a week ago an posted on the forums, using the same hardware and settings: Compared to 2.7, the GPU frametime went up by roughly 10%. CPU frametime went up by roughly 30%. Yay. Notice the different look of shadows in 2.8. In 2.7 I used terrain object shadows set to "flat". Terrain object shadows options "flat" and "off" do not work correctly in 2.8, which may be the reason for the performance impact. See: https://forum.dcs.world/topic/311147-strange-behaviour-of-shadows-on-28/#comment-5074052

No difference.

The broken terrain shadows may be the reason for noticeably worse performance in VR. It's like the default "smooth" terrain object shadows can never be turned off unless you turn off the shadows completely using the option in the left column in the settings menu.

Unfortunately that's the way it is. For example, the guy you've linked previously uploaded an RTX4090 "test" on 11th Oct. That's the day when press embargo has been lifted, but the cards were not yet available to buy, and more importantly, the drivers which worked with 4090 were not yet available for download. Only the selected reviewers received them earlier. What are the chances nvidia would send a kit to this random guy? Not very high. Yes, they give cards to smaller reviewers and various influencers as part of marketing campaign, but those people are usually expected to show the card on a video, talk about it, etc. You know, do some marketing work or at least make an actual review and not just drop a bunch of numbers on the screen in total silence.

Comparison videos from small channels that look like this one you've posted, are usually fake.

Small YT channels with zero commentary and gazillion hardware comparisons never showing the actual hardware tested, are not necessarily the best source of information. Most of them are simple scams showing "emulated" results.

I wouldn't call a difference above 30% "a slight lead". The gap in 1% lows is even higher, although that measurement is typically not super accurate. Raptor Lake is also loosing to X3D in racing simulators like F1 2021 or Assetto Corsa. https://www.tomshardware.com/news/intel-core-i9-13900k-core-i5-13600k-cpu-review

Similar to Ismael's, except I got 5800X3D and Reverb G2, he has 5950x and Varjo.

Your CPU frametimes look really high even on low settings. I wonder if this is Varjo overhead, recording software inssue, or just 5800x3d doing its magic on my system. Do you have a full track for this benchmark? Because for me, the mission looks like this: Compared to:

It doesn't help much if the reprojection itself is of lower quality, with more artifacts showing. And the reprojection preemptively drops to 30 fps in places where it still maintains 45 on SVR. Anyway, I don't want this thread to turn into another SVR into OXR debate. I simply checked both and currently I can't tell a clear winner in terms of performance.

I may not be the right person to ask, I never saw any real gains with OXR on my system.

I've been thinking the same when I got the 3080. Yeah yeah, allocated doesn't mean it's actually used, gpu drivers can adjust to less VRAM, 10GB will be fine, etc. Then DCS got to a point where I couldn't even bring any VR menus or overlays in VR on top of DCS, because everything would slow down to a crawl and stutter massively from VRAM shortage. It may not always be 17 gig and up, and it may run fine on 16 GB card, but it certainly requires above 10GB. To my understanding, OpenXR only shows VRAM usage by the game, and does not take into account VRAM required for your system and other apps working in the background. I did some comparisons yesterday, but just as I didn't see any performance benefit on my system before, I don't see it now. FFR improves things a bit, but I can also enable FFR with SteamVR for similar results.

It think 16GB is a bare minimum. It may even be quite ok for now, but not necessarily for the next 2 years.

I had a 3080 before 4090 and unfortunately, with only 10GB of VRAM, RTX3080 really sucks with high resolution headsets. It's been sort of "okay" two years ago, but ED increased VRAM usage since then, especially with the new modules, terrains, and AI units. I have a simple mission with AH-64, overcast and IHADSS on, sitting on a tarmac in Caucasus. Even at SS 70%, the mission was nearly unplayable on a 3080, with GPU frametimes around 23 ms. Now with RTX4090 the frametimes dropped to 9.5 ms at the same location and settings. Also CPU frametimes dropped from 11ms to 8.5 ms, since it doesn't have to juggle the textures in and out of memory. Turns out that mission requires more than 17 GB of VRAM at the very beginning, and it only gets worse as you keep flying. So in this particular scenario at 70% SS, I went from low 40-ish, not enough for reprojection to kick in, to a pretty stable 90 fps. Even at 150% SS it still hovers around 90 fps, the increased resolution simply does not seem to take much toll on a new card.

If your GPU does not show close to 100% utilisation when gaming on a 2D monitor, and you are not at the vsync limit, then yes, you are most likely limited by CPU. With CPU it's different, basically no game will use it to 100% and looking at cpu utilisation is mostly pointless. In MSFS switching from rtx3080 I saw 40% increase on ultra settings in one spot i tested, and 100% increase in another scenario that is not cpu limited. Except I have 5800x3d, while 10900k performance in this game is comparable with 5600x/5900x

I see no point in going higher, at least for SP. It only makes missions load longer. Maybe it was a useful setting back in Lock On days, when your HDD topped at about 60 MB/s

I don't know if it's Varjo Aero or your CPU, but with 5800X3D and Reverb G2@150% (3868x3784) I see much better frametimes in the same spot despite higher resolution. This is High preset with mirrors enabled, like on your video: null compared to And this is with the settings I use and mirrors off, much more practical for VR than the "HIGH" preset. At these settings 4090 holds 90 FPS quite well in many scenarios, not everywhere of course.

Your setup is fine, the PSU has non-standard sockets on the PSU side. Your previous post was a bit confusing and suggested you run the whole card from two PCIe 8 pin cables.