Jump to content

Flextremes

Members
 View Profile  See their activity

  • Posts

    77

  • Joined

  • Last visited

 Content Type 

Everything posted by Flextremes

  1. Flextremes
    Guys: you really need to report more detailed information. Like what hardware are you using and how are you running VR?
  2. Flextremes
    Let me start off by saying that this is a great conversation. It is very clear that many, like you, do not experience the significant improvement in fluidity so many are reporting. We need to figure out what is going on. I tested your track again until the end: I experienced zero stutters, perfectly smooth. The only reason I went 46 fps is that I have enabled all eye candy (that would destroy performance in ST, btw) and I am running at a very high resolution. Seems like there is a trend: folks on streamVR are generally not happy. It seems like this first version of MT DCS is optimized for native openXR. ED is aware of these stutter problems: they mention it specifically at release. I am hopeful they can fix this: because it seems for open XR native this does not seem a problem. I also suspect that SteamVR and Oculus have a considerably higher CPU overhead than native Open XR. (conversely: reprojection is no way near as good in openXR) Also, it is clear that DCS MT seems to prefer newer CPU's with higher core counts. MT can even render worse results on old CPUs with only a few cores. This makes sense because managing MT processes instead of ST does incur a sizable overhead for tracking and syncing all those threads. If you do not have enough cores, ST will the better option. Lastly, it is also clear that when somebody is GPU limited you should not expect higher FPS. But this is only half of the story: DCS always has been severely CPU bottlenecked: just check multiplayer, Supercarrier, high complexity missions, and even maps like the Marianas: CPU latency is all over the place. With MT these are fundamentally more fluid now. That is why I put so much emphasis on the Carrier example: Total CPU latency combined (app + rdr) in ST: 41.1ms, and in MT CPU combined is only 16.2 ms: this is x2.5 faster. Especially the 'app CPU' 22.5ms (ST) vs 4.0ms (MT) difference is stark: this is more than 5x faster. In VR this means that the chances of the CPU bottlenecking your VR experience is so much lower. I am a casual flyer since Flanker 2.0. (Yep I am that old). Many moons ago: only the DX11 update improved performance in DCS. But never in the history of DCS updates have we got this kind of performance improvement. But I do think we need to manage expectations: Always 90fps in VR without any drops in a simulation like DCS is close to impossible. DCS is such a complex program, simulating the constraints of reality like AI, war logic, physics, technology, weather, multiplayer, and graphics. DCS IMHO dwarfs MSFS in complexity: There is absolutely nothing (for consumers) that comes close. That is why I am so appreciative of MT in DCS.
  3. Flextremes
    My system: Highly optimized Win 10, AMD 5800x3D, 4090, Reverb G2, Native OpenXR, almost maxed out DCS settings with 80% FSR upscaled resolution to 3700x3600 with 100% sharpening. With your track I got 90 fps, until we got close to the island. Then 'app GPU' frametimes go above 9ms: reprojection kicks in and fps is 45. It is all perfectly smooth, reprojection is fine and acceptable for me. I created screenshots at full resolution for you during the track: note the frame times and fps reported by openXR in the right corner below. Click on the image to get the high-res picture, and download (right-click, download save as) it to appreciate the insane fidelity DCS-MT is rendering on my Reverb. So freaking beautiful. Look at both "rdr CPU" and "app CPU" frametimes: they remain ultra low because of MT. This would be way higher in ST. For context: These are my DCS settings... I
  4. Flextremes
    I understand your perspective. No offense taken But just check the improvements in frame times with the carrier on-ramp screenshots here as an example. These are not just some numbers: actually flying in VR with all eye candy enabled is butter smooth, a big difference with Single Threaded performance. I can only double down on that 'sensational' statement. If you focus on fps only, you probably will not fully understand the impact of MT. For instance: the Marianas map, Syriah and the SuperCarrier maps were clearly CPU bottlenecked. Now with MT it makes a big difference that is immediately noticed while flying.
  5. Flextremes
    Honestly, that sucks! It annoys me because I do not understand. Did you compare your CPU and GPU frame times instead of general FPS? I just cannot understand why I am experiencing such a dramatic improvement across the board and you do not. Look at my CPU frame times improvements: those are not dependent on GPU. On a CPU level, you should at least also experience a difference. Is your 4070ti the main bottleneck? I do not believe so. With that said.. I think we are all missing something here: why are the results so mixed. I just cannot fathom what we are overlooking. Grrr... You did start DCS.exe from the Bin-MT folder right? (apologies for such a stupid question, better safe than sorry)
  6. Flextremes
    Compare below the OpenXR CPU and GPU frame times in Single threaded and Multithreaded: F-18 on carrier ramp in Persian Gulf: 1st: Single Threaded... Secondly: Multi-Threaded TOTAL CPU combined in ST: 41.1ms ... in MT CPU combined is only 16.2 ms. Especially the 'app CPU' 22.5ms (ST) vs 4.0ms (MT) difference is stark.
  7. Flextremes
    Ahhhh... Yes the numbers should be the same Disabling Proces Lasso should not make a big difference: affinity is persistent until you reboot or restart a process. So compare before and after you enable Process Lasso. Not the other way round. Easy way to validate frame times yourself: Choose a stable, non-moving, scene (while being parked) in the external view (with loads of objects, I really like to use external view the F18 carrier mission on the Persian Gulf map) that has high but relatively stable 'rdr CPU' frame times. Enable advanced FPS in openXR. This is about 'rdr CPU' and 'app GPU'. 'app CPU' should be unaffected. Wait for frame times to stabilize before comparing/screenshotting 'rdr CPU' and 'app GPU' values. Screenshot before you enable Process Lasso. Enabling Process Lasso, and make a second making sure you have the exact same view angle in VR. Make it easy by enabling screenshots in the OpenXR toolkit companion app, enable the Advanced OpenXR FPS counter, and select CTRL + F12 to create a screenshot create before and after with the exact same view in VR.
  8. Flextremes
    Thanks for the experiment, it really looks like this is not recommended for Intel. Other intel users may thank you.
  9. Flextremes
    Then I am dumbfounded We have very, very similar systems, with similar tuning (I have been keenly watching your tuning thread). Are you running Windows 10? If you want, I can share screenshots of everything I have done to tune/tweak Windows for VR and how I step up openxr and dcs.exe.
  10. Flextremes
    That is correct: Did you configure all processes while DCS was running in VR? You can sort on affinity in the overview to see how processes are configured.
  11. Flextremes
    Did you perform Step 1 and 2? Did you check openXR frame times?
  12. Flextremes
    All info is in the first post Super Curious about your findings! I am quite sure we tuned our Windows in similar fashion: and I have been eyeing your tuning thread extensively.
  13. Flextremes
    I finding really great results with my 5800x3D, but I am on a 4090. Did you try this?
  14. Flextremes
    Cool! Ensure DCS in VR mode is started first before you select all in Process Lasso and assign to other cores. Otherwise you will miss processes that are responsible for VR and they will try to use the same cores as DCS.exe.
  15. Flextremes
    That explains a lot. STEP 1 is crucial.
  16. Flextremes
    OK, then it seems safe to assume using Process Lasso to set the affinity for DCS is not recommended for Intel CPU's. But something is different: it looks like you are using Steam VR, is this correct? Also, did you force all other processes to other cores than DCS.exe? On my AMD 5800x3D, using native openXR, it makes a tangible and measurable difference, consistently, no matter what maps, planes, mission or view I use. It also works in MSFS. Luckily it is easy to test, hopefully, more will chime in with their experience.
  17. Flextremes
    Awesome, I will update my recommendation. I have only experience with 5800x3d, thanks for trying this out. Windows scheduler already dedicates P-core for DCS.exe and all other processes to e-cores? (is this Game Bar working or disabled?) What happens if your force DCS.exe to run only on p-cores and all other processes on e-cores? Scratch that: after reading your comment twice: that is super weird: DCS.exe only on p-cores should be faster than e-cores. I can't explain this at all
  18. Flextremes
    I envy you: 60hz is an instant headache for me. Luckily I can stand reprojection at 45fps.
  19. Flextremes
    None: this is CRTL-F2 -> enable advanced FPS in openXR.
  20. Flextremes
    100% 90fps is currently impossible now for sure. For 90fps 'app CPU' & 'rdr CPU' frame times combined and 'app GPU' need to be both beneath 10ms. However, did you already try these tweaks: and...
  21. Flextremes
    Cool, Screenshot my openXR settings:
  22. Flextremes
    Yep: I think that is a matter of fact. That is why I am embracing reprojection in openXR. My goal is to stay above 45fps. I have to say that reprojection looks way better now?!!! Clearly not perfect, but less wiggly because so much less dropping to 30fps is happening. Currently I am running at almost maxed-out settings in DCS and a very high FSR upscaled render resolution (3880x3800). It is so damn smooth and beautiful now.
  23. Flextremes
    Good question: I would only assign 1 thread for all and 3 for DCS.exe. But really validate your frame times when trying. I can easily imagine your CPU to a bottleneck and fixing Affinity does not work for you. Easy way to validate frame times yourself: Choose a stable, non-moving, scene (while being parked) in the external view (with loads of objects) that has high but relatively stable 'rdr CPU' frame times. Screenshot before and after enabling Process Lasso, making sure you have the exact same view angle in VR. Make it easy by enabling screenshots in the OpenXR toolkit companion app, enable the Advanced OpenXR FPS counter, and select CTRL + F12 to create a screenshot create before and after with the exact same view in VR. Wait for frame times to stabilize before comparing/screenshotting 'rdr CPU' and 'app GPU' values. Please try and report back.
  24. Flextremes
    This is so interesting: when I moved from 5900x to 5800x3D I saw massive improvements. Maybe try to disable Game Bar and start using Process Lasso? Or disable a CCD in bios?
  25. Flextremes
    G2/openXR on the 4090 runs extremely smooth for me... Is it possible to use openXR on the VP2? Also: check-out
×
×
  • Create New...