[Official] DX9 vs DX11 performance difference

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

Thread Status:
Not open for further replies.
  1. TJones

    TJones Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,074
    Likes Received:
    257
    My Specs:
    CPU: Intel Xenon E3-1231 @ 3.40 GHz
    GPU: nVIDIA GeForce GTX 1070 (driver: 381.65)

    DX9:
    Build ID: 1796904
    Serttings: FXAA: on; Sync: off; AA: off; Screen res.: 2560x1080; Refreshrate: 120Hz; fullscreen.
    Ingame settings: all maxed out
    Car: Nissan GTR GT-500
    Track: NOLA-A with 10 AI-cars
    FPS count: Clear sky: 150-210 fps ; Mostly cloudy: 130-180 fps

    DX11:
    Build ID: 1803126
    Settings: Sync: off; AA: Level 3; Screen res.: 2560x1080; Refreshrate: 120Hz; fullscreen; Post effects: None.
    Ingame settings: all maxed out
    Car: Nissan GTR GT-500
    Track: NOLA-A with 10 AI-cars
    FPS count: Clear sky: 90-150 fps ; Mostly cloudy: 50-80 fps

    08.05.2017 21:55
    Build ID: 1807720
    All settings same as above, also car and track.
    FPS count: Clear sky: 115-170 fps ; Mostly cloudy: 95-150 fps
    Note: All fps readings are notet by hand, using GeForce Experience tool.
    Added 2 PerfLog's.

    09.05.2017 21:15
    Build ID: 1807962
    All setting as above ...
    FPS count: Clear sky: 125-180 fps ; Mostly cloudy: 95-145 fps
    Added 1 PerfLog

    10.05.2017 21:33
    Build ID: 1813052
    All settings as ...
    FPS count: Clear sky: 115 - 225 fps ; Mostly cloudy: 113 - 190 fps
    Excellent !
    Added 1 PerfLog

    14.05.2017 11:20
    Build ID: 1818133
    All settings as before, except FXAA now on (not much difference to off)
    FPS count: Clear sky: 110 - 220 fps ; Mostly cloudy: 100 - 190 fps

    Something strange happend with my video settins in Launcher, at first I was able to activate FXAA and AA level 3. But after exiting and re-enter video settins, AA level was off again. After a few tries FXAA was greyed out and AA level was activateable again. So it's very likely FPS counts from 14.05. and maybe 10.05. where done with AA off. Excuse my inconvenience.

    16.05.2017 20:55
    Buil ID: 1825438
    All settings as before (FXAA off, AA level 3)
    FPS count: Clear sky: 115 - 170 fps
    Added 1 PerfLog

    21.05.2017 10:43
    Build ID: 1832786
    All settings as ...
    FPS count: Clear sky: 120 - 170 fps

    23.05.2017 19:44
    Build ID: 1837619
    All settings ...
    FPS count: Clear sky: 129 - 182 fps

    Thank's for your great effort !
     

    Attached Files:

    Last edited: May 23, 2017
  2. patchedupdemon

    patchedupdemon Registered

    Joined:
    Jan 3, 2017
    Messages:
    1,602
    Likes Received:
    1,151
    Nah he's more important than the rest of us
     
  3. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    Quote me please saying so. I really don't care when they fix it as far as it is addressed in a reasonable amount of time.

    The only thing I said is that it would have nice to read some statement regarding the not recognizition of the GPU. Apparently the bug didn't exist.

    And the smart guy of before tells me to be happy because they read my posts. Wow, what a privilege, mine!

    It is pissing the way people jump to the neck everytime someone shows disconformity.
     
  4. Rui Santos

    Rui Santos Registered

    Joined:
    Jan 8, 2012
    Messages:
    1,081
    Likes Received:
    1,206
    Guys, please, we don't need this and the dev's don't deserve too, let's stay focused in what matters... don't start fights when we're trying to solve problems for the good of us all!

    Cheers!
     
  5. Navigator

    Navigator Registered

    Joined:
    Jan 15, 2012
    Messages:
    2,275
    Likes Received:
    389
    System:
    i5 3570 - 8 Gb ram 1600Mhz - GTX 680 2 Gb

    Settings:
    in both cases the same, no PP and no AA (as I don't know if they compare 5/8? so off)
    DX11 car and track used. For testing DX9, both versions, no 2 fps between them.

    DX9:
    Build: 1796904
    min: 142 max: 255 av: 204,3

    DX11:
    Build: 1803126
    min: 92 max: 148 av: 122,5

    Note:
    Compared to the "closed beta" I had a drop (cockpit, just in car, garage) from 170 in DX9 to 123 in DX11. Now, in the latest open beta, I get 85 fps.
    Sad to say its getting worse every time there is something new.
    Is this due to all the messing around for AMD users? Can you keep Nvidia users in mind?

    Edit:
    Forgot performance log: https://www.dropbox.com/s/5qrswyu03oi4t2i/PerfLog_2017-5-7-23-4-48-.txt?dl=0
     
    Last edited: May 7, 2017
  6. DaVeX

    DaVeX Registered

    Joined:
    Feb 7, 2015
    Messages:
    637
    Likes Received:
    701
    I don't think is related to AMD optimization (nvidia on my side), more the software evolve more bugs will surface.
     
  7. Louis

    Louis Registered

    Joined:
    Dec 27, 2016
    Messages:
    1,379
    Likes Received:
    840
    This happen to me since the beggining of rf2 with my nvidia 680 i had before. But Tim told me not to worry about this
     
  8. davehenrie

    davehenrie Registered

    Joined:
    Jul 6, 2016
    Messages:
    7,454
    Likes Received:
    4,369
    I hate to butt in, but where do you get the build #?
     
  9. Emery

    Emery Registered

    Joined:
    Oct 24, 2010
    Messages:
    3,035
    Likes Received:
    1,654
  10. The Iron Wolf

    The Iron Wolf Registered

    Joined:
    Feb 20, 2016
    Messages:
    984
    Likes Received:
    984
    You can also right click on rF2->Properties->local files in Steam, you'll see build #.
     
  11. 4thworld

    4thworld

    Joined:
    Dec 17, 2016
    Messages:
    286
    Likes Received:
    950
    Hi Mate,
    thanks for your help, quick answer to your points before continuing:
    -The beta is changing fast because we want to tackle down issues faster releasing targeted updates.
    -The patches are incremental updates, don't worry about missing updates.
    -The graphics card detection things that you mentioned are just metadata, this does not cause the issues that you wrote.

    We read every single post, despite not being able to respond to everyone of them, we study them and we prioritise them.

    Thanks for all your help guys,
    we got very interesting feedback and nice clues where the problems with different configs may lie.
    This week we'll be able to make good progress with the systems that still have issues.
    This thread is supposed to follow a rigid structure as specified in the first thread and in different other posts, this helps me having a good signal to noise ratio.
    For any other issue, please open a new thread.
    Thanks a lot!
     
    Louis, roby13, Mulero and 11 others like this.
  12. hitm4k3r

    hitm4k3r Registered

    Joined:
    Jun 15, 2016
    Messages:
    1,320
    Likes Received:
    3,121
    Ok, seems that I found the worst offender for my system: it's the cloud rendering. Did an extensive benchmarking regarding sunny vs. overcast for both DX9 and DX11 (Latest beta build and last stable beta build) at Indy, with no AI, no PP and "Agressive Threading" turnt off as it makes no difference for my system. Tested also at Atlanta MP with the same results, but not included in my zip file. Dxdiag file included in the zip file aswell as detailed display settings for each benchmark. There we go.
     

    Attached Files:

  13. Drathuu

    Drathuu Registered

    Joined:
    Oct 14, 2010
    Messages:
    349
    Likes Received:
    97
    My Specs:
    CPU: Intel i5 2500
    GPU: nVIDIA GeForce GTX 1070

    Interestingly in the closed beta i found the opposite than the below with Frames Higher in DX11 (nearly twice the frames vs dx9) .. albeit we were using Joesville as the benchmark track.. Possibly it has something to do with the polycount being rendered ???

    SINGLE SCREEN

    DX9: Fullscreen
    Build ID: 1796904
    Serttings: FXAA: off; Sync: off; AA: x2; Screen res.: 3440x1440 100hz (4.9MIO PIXEL)
    Ingame settings: Max - Excepting - Shadows on Fast, Reflections on low (both of them)
    Car: Renault 3.5 2015
    Track: NOLA-Indy with 20
    Mostly Cloudy
    Frames: 19819 - Time: 140812ms - Avg: 140.748 - Min: 100 - Max: 190

    DX11: Fullscreen
    Build ID: 1803126
    Serttings: FXAA: off; Sync: off; AA: x2; Screen res.: 3440x1440 100hz (4.9MIO PIXEL)
    Ingame settings: Max - Excepting - Shadows on Fast, Reflections on low (both of them)
    Car: Renault 3.5 2015
    Track: NOLA-Indy with 20
    Mostly Cloudy
    Frames: 9426 - Time: 145641ms - Avg: 64.721 - Min: 55 - Max: 75

    DX11: PP ULTRA Fullscreen
    Build ID: 1803126
    Serttings: FXAA: off; Sync: off; AA: x2; Screen res.: 3440x1440 100hz (4.9MIO PIXEL)
    Ingame settings: Max - Excepting - Shadows on Fast, Reflections on low (both of them)
    Car: Renault 3.5 2015
    Track: NOLA-Indy with 20
    Mostly Cloudy
    Frames: 9453 - Time: 149062ms - Avg: 63.417 - Min: 53 - Max: 73

    TRIPLE SCREENS

    DX9: Fullscreen
    Build ID: 1796904
    Serttings: FXAA: off; Sync: off; AA: x2; Screen res.: 5760x1080 (6.2MIO PIXEL)
    Ingame settings: Max - Excepting - Shadows on Fast, Reflections on low (both of them)
    Car: Renault 3.5 2015
    Track: NOLA-Indy with 20
    Mostly Cloudy
    Frames: 17086 - Time: 141172ms - Avg: 121.030 - Min: 90 - Max: 173

    DX9: Fullscreen - MULTIVIEW
    Build ID: 1796904
    Serttings: FXAA: off; Sync: off; AA: x2; Screen res.: 5760x1080 (6.2MIO PIXEL)
    Ingame settings: Max - Excepting - Shadows on Fast, Reflections on low (both of them)
    Car: Renault 3.5 2015
    Track: NOLA-Indy with 20
    Mostly Cloudy
    Frames: 12072 - Time: 144469ms - Avg: 83.561 - Min: 55 - Max: 135

    DX11: Fullscreen
    Build ID: 1803126
    Serttings: FXAA: off; Sync: off; AA: x2; Screen res.: 5760x1080 (6.2MIO PIXEL)
    Ingame settings: Max - Excepting - Shadows on Fast, Reflections on low (both of them)
    Car: Renault 3.5 2015
    Track: NOLA-Indy with 20
    Mostly Cloudy
    Frames: 8096 - Time: 135156ms - Avg: 59.901 - Min: 43 - Max: 68

    DX11: Fullscreen - MULTIVIEW
    Build ID: 1803126
    Serttings: FXAA: off; Sync: off; AA: x2; Screen res.: 5760x1080 (6.2MIO PIXEL)
    Ingame settings: Max - Excepting - Shadows on Fast, Reflections on low (both of them)
    Car: Renault 3.5 2015
    Track: NOLA-Indy with 20
    Mostly Cloudy
    Frames: 6883 - Time: 186516ms - Avg: 36.903 - Min: 32 - Max: 43

    DX11: PPULTRA - Fullscreen - MULTIVIEW
    Build ID: 1803126
    Serttings: FXAA: off; Sync: off; AA: x2; Screen res.: 5760x1080 (6.2MIO PIXEL)
    Ingame settings: Max - Excepting - Shadows on Fast, Reflections on low (both of them)
    Car: Renault 3.5 2015
    Track: NOLA-Indy with 20
    Mostly Cloudy
    Frames: 5206 - Time: 141453ms - Avg: 36.804 - Min: 32 - Max: 42

    Summary -
    Single Screen - DX9 to DX11 on current build less than half the framerate 140->64
    Triple Screen - DX9 to DX11 half frame rate on full screen 121->59
    Triple Screen Mutliview - DX9 to DX11 - more than half the framerate 83->36

    Post processing has neglible effect (less than 1 frame difference) - (With it off Vs Full)

    PERFLOG_2017-5-8-23-38-41-.txt is the Single Screen 3440x1440 Fullscreen dx11 version
    PERFLOG_2017-5-8-23-46-49-.txt is triple screens (fullscreen mode) (distorted side screens of course)
    PERFLOG_2017-5-8-23-53-53-.txt is triple screens with Multiview

    The GPU load only ever maxed at 60, however sat around 45 on average - If this is a percentage ?? then it only ever used on average 45% of the GPU.. (Fans didnt even start up - never got hot).
     

    Attached Files:

    Last edited: May 8, 2017
  14. Rony1984

    Rony1984 Registered

    Joined:
    May 10, 2012
    Messages:
    500
    Likes Received:
    28
    one pic says more than a thousand words..

    Operating System: Windows 10 Home 64-bit
    CPU: Intel Core i7 3770 @ 3.40GHz 36 °C Ivy Bridge 22nm Technology
    RAM: 12,0GB Dual-Channel DDR3 @ 798MHz (11-11-11-28)
    Motherboard: MEDION H77H2-EM (SOCKET 0) 28 °C
    Graphics: NV Surround (5760x1080@60Hz), NVIDIA GeForce GTX 970 (MSI) 36 °C

    2 pics taken under the exact same conditions and same time, 1 on dx9, and 1 on dx11(latest beta 05-05), when exiting the pitlane

    In addition to the overall higher GPU-load on dx11, there is also a lot of input lag, even when the experience is smooth. The on-screen wheel simply lags a lot behind my actual wheel. An effect which is fully absent with dx9. Edit: I've multiview enabled and frames capped at 60.

    Edit 2: aggressive threading is devastating for the smoothness on my system. It causes profound stuttering.

    Dx11
    20170508184902_1_beta.jpg

    Dx9
    20170508193131_1_non_beta.jpg
     
    Last edited: May 8, 2017
  15. Navigator

    Navigator Registered

    Joined:
    Jan 15, 2012
    Messages:
    2,275
    Likes Received:
    389
    System:
    i5 3570 - 8 Gb ram 1600Mhz - GTX 680 2 Gb

    Settings:
    in all cases the same, no PP and no AA (as I don't know if they compare 5/8? so off)
    DX11 car and track used. For testing DX9, both versions, no 2 fps between them.

    DX9:
    Build: 1796904
    min: 142 max: 255 av: 204,3
    (100%)

    DX11:
    Build: 1803126
    min: 92 max: 148 av: 122,5
    (-40%)


    DX11:
    Build: 1807720
    min: 94 max: 192 av: 143,3
    (-29,8%)

    Note:
    Same test with newer build. A bit better, but still far from what the "closed Beta" was.......
    The test you wanted in the closed one, the "jump in car, write down fps directly" is with this one still 83; no change there. (For comparison; DX9 was 170, closed build was 123, this and former build is 83 fps)

    Note 2:
    Tested this several times, but "aggressive threading" seem to utilise 1 core of the cpu less.
    4 cores, usage for 1,2,3 and 4 about: 85%, 55%, 15%, 15%
    Same with AT on: 100%, 25%, 15%, 15%
    No gain whatsoever with AT on.

    Note 3:
    Trackside cameras seem to have improved the most on the 1807720 build. Before, with every version of DX11 it looked like the fps was 10 while it was not, but this "bad" it ran. Now; it seems fluent.
     
    Last edited: May 8, 2017
    Rui Santos likes this.
  16. Mulero

    Mulero Registered

    Joined:
    Jan 16, 2012
    Messages:
    351
    Likes Received:
    153
    @4thworld Tested the latest version and I think there is a small improvement. I've noticed that Control + F is having a big impact here. I do not know if in previous versions or on other computers.

    FPS.jpg

    Track: Nola
    Car: Indycar
    Resolution: 1920 * 1080, All to Max, 35 FOV
    AA: Level 8 in DX9, Level 5 in DX11 (no PP)
    Conditions: 1 full lap from the finish line to the finish line.
    Build ID: 1807720
    Aggressive Threading OFF
     

    Attached Files:

  17. demerzel

    demerzel Registered

    Joined:
    Aug 30, 2015
    Messages:
    197
    Likes Received:
    127

    Attached Files:

  18. Proscribo

    Proscribo Registered

    Joined:
    Nov 16, 2013
    Messages:
    6
    Likes Received:
    0
    Performance is improved greatly, minimum is the same as in DX9 and average is just 10 lower anymore (100 vs 90). However, I don't know how you've done it, because my GPU usage is still really ugly. :eek:
    I've added results of DX9, DX11 with aggressive threading on and off (fraps, perflog).
     

    Attached Files:

  19. MaD_King

    MaD_King Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,827
    Likes Received:
    611
    @4thworld Same here with latest build, I passed from 60 FPS to 90 FPS, great improvement !!!
     
  20. Rony1984

    Rony1984 Registered

    Joined:
    May 10, 2012
    Messages:
    500
    Likes Received:
    28
    also a minor improvent on my system with build 1807720, but Atlanta with the gt500 and some opponents is still a stuttery and laggy mess. Note that the lag is evident even when the GPU is hardly loaded and no stutters are noticable.
     

    Attached Files:

Thread Status:
Not open for further replies.

Share This Page