[Official] DX9 vs DX11 performance difference

Discussion in 'Technical & Support' started by 4thworld, May 4, 2017.

Thread Status:
Not open for further replies.
  1. Ari Antero

    Ari Antero Registered

    Joined:
    Jul 27, 2012
    Messages:
    1,882
    Likes Received:
    829
    I don`t think it is your CPU. If you read this post from @stonec you find the answer why: https://forum.studio-397.com/index....eezing-jumpy-display.57025/page-3#post-907561
    :)
     
  2. hitm4k3r

    hitm4k3r Registered

    Joined:
    Jun 15, 2016
    Messages:
    1,320
    Likes Received:
    3,121
    Thanks for th headsup Ari, but I am afraid it really is a problem with DX11and it's CPU utilization in general on my end with the recent build. I tested the USF and Radical now and I get those CPU spikes aswell. It is really wierd as I have been playing the DX11 beta pretty much since the beginning and never had those issues, especialy not in multiplayer, wich is a big slideshow with the 650s right now. This is how the CPU graphs look for the 650s in DX9 and DX11. They aren't 100% matching the same situation, but I guess you get the picture. That's the first lap of an AI race at Silverstone: CPU_Graph_650s_DX9.PNG CPU_Graph_650s_DX11.PNG
     
  3. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    @hitm4k3r Those spikes look like the CPU is throttling down, maybe due to overheating or excessiver overclock (it's best to disable dynamic clock speed adjustment unless you're very concerned about the electricity bill). Have a look at the clock frequency while you are gaming and check if it stays constant.
     
  4. hitm4k3r

    hitm4k3r Registered

    Joined:
    Jun 15, 2016
    Messages:
    1,320
    Likes Received:
    3,121
    I have noticed that the temps get a bit higher, but that might be even caused simply by the fact that we have quite hot weather atm or by how DX11 stresses the CPU itself. And in general I am not using any kind of overclocking or stuff like that. What I find strange is that I only have those spikes in DX11. Went back to DX9 yesterday to have a bit of fun with the Flat6 mod at Oulton Park and it is running smooth as silk with 20 cars on track and it doesn't skip a beat. The charts I posted above, are with 10 cars. What I might test though is, how it behaves without dynamic clocking, so thanks for that tip. At the end I just hope that they can find the issue (wich seems to have been introduced with one of the latest builds), so that I don't have to use such drastic methods and can use my PC as allways. I could run the ealrier DX11 builds without any issues with 20 cars and online worked perfect once they fixed the weather lags. So fingers crossed :)
     
  5. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    DX9:
    Frames, Time (ms), Min, Max, Avg
    7398, 60000, 95, 179, 123.300

    DX11:
    Frames, Time (ms), Min, Max, Avg
    6825, 60000, 90, 154, 113.750

    Had a spare 5mins before Work and decided to give this a go, slightly lower performance in DX11 but nothing substantial or worrying and it was perfectly acceptable to "race" with (used the AI control both times).
     
    Last edited: Aug 29, 2017
    SPASKIS likes this.
  6. Juergen-BY

    Juergen-BY Registered

    Joined:
    Jun 16, 2012
    Messages:
    3,089
    Likes Received:
    440
    DX11: Frames: 11028 - Time: 120000ms - Avg: 91.900 - Min: 71 - Max: 115
    DX9: Frames: 11639 - Time: 120000ms - Avg: 96.992 - Min: 76 - Max: 120
    To compare the both versions a wee bit more serious (for me) PP effects were complete disabled for the DX11 version (EPostProcessingSettings=0).
     
  7. DaVeX

    DaVeX Registered

    Joined:
    Feb 7, 2015
    Messages:
    637
    Likes Received:
    701
    Have you tried to move ffb to a separate core?
    Ifnot, go to your controller.json and locate an entry called "Use Thread". If that is set to "false" then change it to "true". That way SIM will use a separate thread (that can run on a different CPU core) to send the FFB update commands to the driver of your wheel.
     
    SimonJ and Ari Antero like this.
  8. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    Had some more spare time and did two different tests, All settings Off/Low then All settings Full/Max

    DX9 - Off/Low:
    Frames 15792
    Time (ms) 60000
    Min - 238
    Max - 296
    Avg - 263.200

    DX11 - Off/Low:
    Frames - 14112
    Time (ms) - 60000
    Min - 184
    Max - 285
    Avg - 235.200

    Obviously on my system there is no visual difference between the two results, DX11 slightly lower performance but in no way noticeable in game.

    DX9 - Full/Max:
    Frames - 4478
    Time (ms) - 60000
    Min - 59
    Max - 95
    Avg - 74.633

    DX11 - Full/Max:
    Frames - 4393
    Time (ms) - 60000
    Min - 54
    Max - 100
    Avg - 73.217

    The numbers are much closer than I was expecting with barely anything between the two versions, I was expecting a bigger drop in DX11 so that's a nice surprise to see it remain almost identical to DX9.

    -=-=-

    N.B. I will add doing a simple FRAPS Benchmark is not a guarantee that one is better than the other as with each Run of the Benchmark different things can happen on track resulting in different FPS, whilst the FPS change might not be huge it could result in varied Benchmarks. (If I get time I'll do the Replay Test as per the thread on the main forum)

    N.B.2 Post Effects was set to NONE and EPostProcessingSettings=0 in the Config_DX11.ini file for both DX11 tests.
     
  9. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    Purely because I was curious, DX11 with Ultra PE:

    Frames - 3246
    Time (ms) - 60000
    Min - 47
    Max - 61
    Avg - 54.100
     
  10. Emery

    Emery Registered

    Joined:
    Oct 24, 2010
    Messages:
    3,035
    Likes Received:
    1,654
    @MarcG & @Juergen-BY -

    What resolution/card and which track/car/weather combo? Otherwise the numbers are kind of useless without their context.
     
  11. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    1920x1080 Specs in profile, Default weather, Mclaren at Silverstone International
     
    Emery likes this.
  12. Juergen-BY

    Juergen-BY Registered

    Joined:
    Jun 16, 2012
    Messages:
    3,089
    Likes Received:
    440
    It`s not about the absolut numbers, but the difference. Cant see anything useless....tbh
     
    Davvid likes this.
  13. TravisRoe666

    TravisRoe666 Registered

    Joined:
    Jan 31, 2012
    Messages:
    40
    Likes Received:
    3
    havent been playing for months since the DX11 issues when released
    I'll give it another go now, is the general consensus now that the average performance is better in DX11? I have a nice i7 build with GTX1070 that was running awful in early DX11 builds but immediately perfect returning to DX9 build.
    I'll report back in 2 days, any thoughts shoot through to me in case i end up back on here troubleshooting! :) :)
     
  14. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    There hasn't been an update since July, so nothing has changed recently in that regard. I reckon there was one update in June that improved performance somewhat. Performance in DX11 on newer hardware should be within 10% of DX9 in my experience (usually a tad bit slower). On older Nvidia cards DX9 is much faster. Performance problems that are remaining I think are more due to ancient gMotor engine stuff (like excessive usage of RAM, CPU and PCI-E bandwidth), which sometimes prevents GPU from being 100% utilized.
     
  15. patchedupdemon

    patchedupdemon Registered

    Joined:
    Jan 3, 2017
    Messages:
    1,602
    Likes Received:
    1,151
    What is your typical ram and vram usage when racing.
    My vram is roughly around 1.5gb,so I have 4.5gb left doing nothing.
    Is there a way to make rf2 use up more vram,or is that on the devs
     
  16. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    You don't just use vram, it doesn't contribute to processing performance or anything like that.
     
  17. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    I repeated the benchmark with latest dx9 and dx11 builds with similar results to previously posted benchmarks. 20 % fps loss.

    ALT TAB does nothing wrt to FPS. I also tried disabling optimization for fullscreen under rFactor2.exe properties-compatibility with no effect in FPS.

    In ingame video settings, DX11 version shows more resolutions available compared to DX9 version. The available refresh rates also differ. Is this normal?

    On the good part, fullscreen scaling now properly works without changing Windows resolution.

    DX9 B2008937?
    2017-09-10 20:53:33 - rFactor2
    Frames: 8313 - Time: 114093ms - Avg: 72.862 - Min: 54 - Max: 111

    DX11 29jul
    2017-10-08 19:46:44 - rFactor2
    Frames: 6408 - Time: 113891ms - Avg: 56.264 - Min: 41 - Max: 89

    DX11 29jul
    2017-10-08 19:52:26 - rFactor2 ALT TAB
    Frames: 6414 - Time: 113953ms - Avg: 56.286 - Min: 42 - Max: 89

    DX11 29jul + FullScreen Optimization Disabled
    2017-10-08 19:57:34 - rFactor2
    Frames: 6393 - Time: 114125ms - Avg: 56.018 - Min: 41 - Max: 89

    DX9 B2008937
    2017-10-08 20:14:28 - rFactor2
    Frames: 7995 - Time: 113562ms - Avg: 70.402 - Min: 53 - Max: 107

    Per Second average FPS
    upload_2017-10-9_0-11-26.png

    Instant FPS (Stuttering?)
    upload_2017-10-9_0-28-57.png

    Video Res Differences
    upload_2017-10-9_0-37-34.png
     
    Last edited: Oct 8, 2017
  18. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    @SPASKIS Hm, is there a new build? My game still says DX11 July 29th in the right hand corner of the screen. Regarding performance, the trend I noticed is that older Nvidia cards perform worse. GTX 1000 and possibly 900 series have more similar performance in DX11 as in DX9.

    [​IMG]
     
    SPASKIS likes this.
  19. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    No. It is the same build you quote. I simply noticed that my latest report was previous to latest reported DX11 update from July 7th. So I gave it a try!

    BTW above results are using fullscreen and cockpit view. Changing to borderless/windowed reduces performance by 30% in DX9 and has no effect on DX11. For that reason I believe fullscreen is not properly working in DX11.
     
  20. patchedupdemon

    patchedupdemon Registered

    Joined:
    Jan 3, 2017
    Messages:
    1,602
    Likes Received:
    1,151
    Are you really suprised you get less fps in dx11 with a cpu that has a stock clock of 2.2ghz and that gpu,your hardware is obviously not up to the task of the extra horsepower dx11 requires.

    I'd have thought with all your knowledge this performance drop wouldn't be a surprise to you,yes you have a i7 but those clocks are too low
     
Thread Status:
Not open for further replies.

Share This Page