[Official] DX9 vs DX11 performance difference

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

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

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    Although I cannot do a VR Benchmark (nVidia has one but it's a faff to work with, one day!) I have seen a visual performance increase in Build 1859720. I have been running the same Mod & Track these past few days without Graphical/PPD Changes and today I see less stuttering at Race Starts (it's still there but not as apparent) and a smoother experience during the Race.

    This could of course be a Placebo as I have no definitive way of benchmarking (yet) but I'm convinced performance is better with this Build.
     
    bravotangosix likes this.
  2. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    upload_2017-6-2_2-22-15.png
    upload_2017-6-2_2-22-29.png
     

    Attached Files:

  3. Ari Antero

    Ari Antero Registered

    Joined:
    Jul 27, 2012
    Messages:
    1,882
    Likes Received:
    829
    Video sync seems to be random on and off. Anybody else has same issue?
     
  4. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    I don't use V or other sync. However I have the same lack of consistency feeling in general. Considering the results I posted last night, it would seem that fullscreen is not being properly applied showing a similar fps drop with respect to dx9 as when using borderless or windowed mode. I still haven't compared performance between full screen, windowed and borderless using the same settings to see if what I am saying makes sense or not. Ill give it a try later....
     
    Matt Beavis likes this.
  5. Ari Antero

    Ari Antero Registered

    Joined:
    Jul 27, 2012
    Messages:
    1,882
    Likes Received:
    829
    Feels like full screen randomly switches to borderless or windowed mode
     
    Matt Beavis, demerzel and SPASKIS like this.
  6. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    In my case the biggest sympton is that ALT tab doesn't properly work. After pressing it, rF2 window turns to unscaled 1280x720 (the full screen resolution I use) covering greatest area (94% x 94%) of my native 1366x768 laptop screen. If using external 1920x1080 TV the covered area is 2/3 x 2/3.

    Considering that to switch resolution or video mode you need to go through video settings I think that the behaviour is because full screen is not effectively working. Getting similar FPS as with dx9 using windowed or borderless is an additional clue for me.

    Last one is that all the available resolutions for fullscreen are not available for borderless or windowed modes.
     
    Rony1984 and Ari Antero like this.
  7. Ari Antero

    Ari Antero Registered

    Joined:
    Jul 27, 2012
    Messages:
    1,882
    Likes Received:
    829
    Randomly is perhaps wrong word to use. When I hit the race button I have 120 fps, esc and the race button again I have over 200 fps.
     
    SPASKIS likes this.
  8. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    Last edited: Jun 2, 2017
  9. suryaprihadi

    suryaprihadi Registered

    Joined:
    Apr 16, 2017
    Messages:
    87
    Likes Received:
    32
    AI driving all , so that time is consistent ( me can't drive as fast as AI:p )

    For Silverstone and Nola-C , its pit to 1 round of track ( rolling start ?? ) and 1 lap, where the game counter indicated lap time.

    For Nords-Autumn Tourist, its pit to 1 round of track only passing the pit, its too long a track:D
    .
    Silverstone GT DX9 & 11  2-june-17.JPG

    .
    Nola-C DX9 & 11  2-june-17.JPG

    .
    Nords - Tourist Autumn - DX9 & 11  2-june-17.JPG

    Is Silverstone and Nola-C more DX11 optimized ?
    I am seeing a bit better FPS in DX11 by average.
    Nola-C beggining DX9 faster, but then DX11 perform a bit better.
    Nords, DX9 wins.

    Question to graphic experts :
    My previous GTX980 is only a 4GB GPU. This new GTX1080Ti is 11GB.
    When I use full all best setting DX9 which is more memory intensive so it seems, GPU memory used will be 4GB.

    What happen when our GPU memory is at 2 or 3 GB but we use all best setting , where I am seeing
    from my experience for Nola-C and Silverstone easily 3.9GB used if DX9.
    Will the impact be very bad for FPS ? Or what worse will happen ?

    Thanks
     
    Ari Antero likes this.
  10. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    First of all, today's build 1861431 didn't have any effect in terms of performance.

    However, I just finished testing and the results further suggest that full screen mode is not being "effectively applied". By "effectively applied" I refer to having the same behaviour as with DX9. According to my theory some kind of borderless mode is being applied. The biggest proof of this is that with DX9 I suffered a 30% fps loss when using Borderless or Windowed Mode with respect to Full Screen Mode. Curioulsy, in DX11 all 3 modes show the exact same performance.... You can see the figures in the attached graph. Coincidence? I bet not.

    • The optimistic view: Given that problem seems to be isolated we could expect to get it hopefully fixed soon. If that happens, it would be the greatest news. Considering that the FPS loss in DX11 wrt DX9 is about 15% but the expected gain for effectively applied FullScreen mode is 30% I would be expecting a net 15% fps gain when it gets sorted. Just as informative I included a test with full screen at a higher resolution to account for its comparison in performance. Going from 1280p to 1366p only implies a 7% fps loss. Half the amount of pixel increase)
    • The pessimistic view: If on the other hand we consider the recent history around similar problems regarding fullscreen, borderless fps difference and, of course, Alt Tab Issue (ATI).... it could happen that the issue doesn't get ever sorted despite being a known fact. If Alt Tab Issue would have been sorted in its origin by those games affected by it (I remind other games are ATI free), borderless or windowed mode might not have even existed.
    From what I have talked to some IT friends, It seems that Direct X is very poorly documented by Microsoft in several aspects. This could explain the general inability in gaming industry to solve Alt Tab Issue. Let me remind that in the thread where I posted my findings regarding ATI, the more general solution for fixing car components that were systematically presenting the problem was (believe it or not) to rename one texture file, mainly for animated light materials. Even more difficult to understand was that when Yoss updated AM's GT3 mod including a fix for ATI, I could still reproduce the systemtic ATI for some vehicle loading order while using other loading orders would be impossible to force it. Absolutely puzzling to understand what could be going behind the scenes....​
    upload_2017-6-2_23-4-52.png
     
    Last edited: Jun 2, 2017
  11. Mulero

    Mulero Registered

    Joined:
    Jan 16, 2012
    Messages:
    351
    Likes Received:
    153
    No significant changes during the last two weeks. Control + F is no longer affecting performance, even the FPS are a bit better, 2 or 3 FPS more than average.

    Performance is still far below the DX9 version and making the game unplayable for me at the moment.

    FPS.jpg
     

    Attached Files:

    demerzel likes this.
  12. Mulero

    Mulero Registered

    Joined:
    Jan 16, 2012
    Messages:
    351
    Likes Received:
    153
    I have been doing some tests with other cars and if the performance in DX11 is already bad in itself (in my case), if we add other cars to the combination, the results are terrible.

    First lap, Silverstone, 40 Camaro GT3 (16 cars visible). Measurements made in replay.

    FPS race.jpg
    I need to put everything to the minimum, deactivate reflections, shadows and effects to avoid dropping below 60 FPS. Even so, FPS remain low, 75FPS on average and below 60 FPS during the first few corners.

    In DX9, with an Athlon II x2 @ 2400mhz and an AMD 5850, I was getting many more FPS apart from a higher graphics quality.
     
  13. Roover

    Roover Registered

    Joined:
    Feb 18, 2017
    Messages:
    18
    Likes Received:
    11
    Can somebody explain to me how to switch between DX9 and DX11? And how i can check my dx version in game? Sorry if i repeating this question...
     
  14. Trevor Greenfield

    Trevor Greenfield Registered

    Joined:
    Apr 19, 2017
    Messages:
    37
    Likes Received:
    11
    Steam > Library > right click on rfactor2 > left click properties > left click betas > change "select the beta..." drop down menu to v1108-dx11-open-beta". To switch back, change to just v1108. In game, there is text at the top of the screen that says the beta. For a specific build number to reference look in the properties box at the local files tab, at the bottom you will see buildid.
     
  15. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    I made an edition of the obtained results in order to show the expected fps if fullscreen mode worked as in DX9.
    sketch-1496600392674.png
     
    Last edited: Jun 4, 2017
    stonec and Ari Antero like this.
  16. Stan

    Stan Registered

    Joined:
    Jan 18, 2016
    Messages:
    166
    Likes Received:
    88
    @SPASKIS Hi, my fps is as steady as possible at 60, not moving more than 0.1 (with a very low cpu/gpu usage, less than 50%), but none the less, it is still not smooth at all and I also suspect that it's regarding of the FullScreen problem!?!
    With my Nvidia (MSI-780GTX-OC), I am not able either to access Windows7 desktop.
    I really hope they'll be able to fix it...
     
    Last edited: Jun 4, 2017
  17. Roover

    Roover Registered

    Joined:
    Feb 18, 2017
    Messages:
    18
    Likes Received:
    11
    Thank you, Trevor!
     
  18. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    I can't see the kind of performance boost as in @SPASKIS graph when going fullscreen/windowed in DX9, it's 1-2 FPS difference at most for me. This is on Windows 10 and a 2 GB Nvidia GPU. Earlier versions of Windows I reckon had a heavier Aero implementation for desktop, which took more performance away from games that were running through the desktop (i.e., not fullscreen). If you run Windows 7, a good test would be to run it with the classic desktop (transparency/aero off), I have seen people report better performance with that. For newer Windows versions this option is no longer available.
     
    SPASKIS likes this.
  19. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    As it was reported back when Alt Tab problem was discussed, borderless or windowed mode didn't affect the same way every PC.
    This was the main reason why some of us keep asking to solve the ALT TAB issue in full screen mode. Windowed or borderless were not an option for lower end machines showing the fps drop.

    As you, others didn't report fps impact when using alternative display modes.
     
    Ari Antero likes this.
  20. MarcG

    MarcG Registered

    Joined:
    Oct 4, 2010
    Messages:
    6,854
    Likes Received:
    2,234
    I'll post this here as it's relevant to the current discussion, this is the FCAT Benchmark in VR with all the Display Settings, as you can see (for me) there is essentially no difference when running in VR in any of the Display options. (ignore the first 2-3seconds of Fullscreen)

    Build 1861431 - Display.png
     
Thread Status:
Not open for further replies.

Share This Page