Recurrent NVIDIA Driver Crashes and Erratic GPU Usage

Discussion in 'Technical & Support' started by ECAR_Tracks, May 9, 2016.

  1. The Iron Wolf

    The Iron Wolf Registered

    Joined:
    Feb 20, 2016
    Messages:
    984
    Likes Received:
    984
    I used to have, not anymore (since I had to remove all overclocks till this gets solved). My observation is - it does not make difference with crashes.
     
  2. Gijs van Elderen

    Gijs van Elderen Registered

    Joined:
    May 1, 2014
    Messages:
    1,393
    Likes Received:
    469
  3. The Iron Wolf

    The Iron Wolf Registered

    Joined:
    Feb 20, 2016
    Messages:
    984
    Likes Received:
    984
    Thanks Gijs, I'll probably give it a try after I rule out a hardware problem with Oculus by running other game with native support to see if I get crashes there.

    I have to admit, I am running with unusual resolution (1600x1440), 100 FOV and SGSSAA, so with all of this I am asking for trouble :) Also, as I said if I switch to 1080p and usual monitor output, I've not seen crashes, though I don't play much that way. It's even not clear if it's rF2 fault, maybe vorpX does something to video driver that makes it hang.

    I couldn't find any other AA setting to make shadows not flicker/jumpy other that SGSSAA enabled through inspector. What AA are you using, in game?
     
  4. Gijs van Elderen

    Gijs van Elderen Registered

    Joined:
    May 1, 2014
    Messages:
    1,393
    Likes Received:
    469
    Level 8, that's level 4 for gtx9xx and gtx10xx cards.

    A gtx1080 should be able to run rF2 just fine with all display setting at maximum with a 1920x1080 or 1600x1440 resolution.
    (Shadow blur reduces shadow flickering)

    I've a titan black (gtx780ti with more vram). I've everything at max. When i'm alone on track i get +120fps in 1920x1080 single screen and +60fps with tripple screens + multiview.

    I have to admit: I've reduced visible cars to 12 to keep fps high enough in first laps. I usually do 1-3 hour races offline. Field is spread out after a few laps.
    For oval races 12 visible cars isn't enough.

    100 cars at the nordshleife was not a problem for my rig. :)
     
    The Iron Wolf likes this.
  5. The Iron Wolf

    The Iron Wolf Registered

    Joined:
    Feb 20, 2016
    Messages:
    984
    Likes Received:
    984
    780Ti is awesome card. Gave it to my son, he plays any FPS no problem, with a 10 year old processor :)

    All settings in game, you do not modify CP settings or Inspector, right?

    I believe game uses MSAA, I wasn't happy with shadows at level 4 AA of in game. Everything else is maxed out.
     
  6. Voltaic

    Voltaic Registered

    Joined:
    Sep 14, 2011
    Messages:
    327
    Likes Received:
    65
    @Marcel Offermans I reformatted my startup drive (C) and installed a clean Windows 10 from scratch. I installed a fresh Steam on my "C" drive and fresh rF2 on my "H" drive, after which I was able to run a few laps without issues. Just now I tried to load the Enduracers Corvette in the garage and got a sim motor error, something to do with the wipers, which crashed rF2.

    I will keep trying, however the whole subscription system seems to be screwed up. I keep getting the "Downloading Workshop Items...4 Items Remaining" message, I had the same message before I reformatted my HD and installed a new Steam setup. Once again, I have Steam installed on my SSD startup drive and Steam Apps on a second larger drive where I re-downloaded and re-installed my Steam apps from scratch. I don't know what else I can try :(
     
  7. Marcel Offermans

    Marcel Offermans Registered

    Joined:
    Oct 4, 2010
    Messages:
    647
    Likes Received:
    2,934
    @Voltaic I am taking the discussion to a personal conversation to see if I can help you.
     
  8. Voltaic

    Voltaic Registered

    Joined:
    Sep 14, 2011
    Messages:
    327
    Likes Received:
    65
    I did as you suggested, I unsubscribed to all non-ISI/Studio 397 and kept a few official cars and tracks. I went into a track, exited and when I tried to load Silverstone again rF2 crashed.

    Other than the SimCommander for my AccuForce and the SimDash for my wheel's display, I have no other plugins. I can try without the SimDash if you suspect it, I can't drive w/o SimCommander though.

    If the plugins are not at fault; could this be caused by something else in my system? Bad RAM? Although I have stress tested my PC and everything checks out fine.

    PS; I just ran a full memory test, no errors.
     

    Attached Files:

    Last edited: Dec 26, 2016
  9. rfuctor

    rfuctor Registered

    Joined:
    Jan 12, 2012
    Messages:
    106
    Likes Received:
    4
    Has there been a fix/workaround for this issue?
     
  10. Marcel Offermans

    Marcel Offermans Registered

    Joined:
    Oct 4, 2010
    Messages:
    647
    Likes Received:
    2,934
    I have had an extensive private conversation with @Voltaic about this matter and he does seem to have resolved this. I would like to invite him to share his findings with the community (as in general I don't make private conversations public without permission, and he can describe in more detail what he did).
     
    Ernie and bravotangosix like this.
  11. WhiteShadow

    WhiteShadow Registered

    Joined:
    Feb 16, 2015
    Messages:
    681
    Likes Received:
    3
    It might be good idea to look Control Panel > Administrative Tools > Event Viewer > Windows Logs > Application and System logs to find out what is causing rF2 crashes.
     
  12. Voltaic

    Voltaic Registered

    Joined:
    Sep 14, 2011
    Messages:
    327
    Likes Received:
    65
    Sorry for the delayed reply Marce.

    After some research I was able to narrow down my issues to a hardware/software problem. As it turns out, there seems to be a USB .DLL problem with some chipsets, here is some of the discussion at SimXperience;

    "@Marcel Offermans

    Marcel, this is the meat of the discussion at SimXperience:

    "I found the Project CARS/OSW similar thread about pid.dll.

    http://forum.projectcars...-completely-stop-working

    Post #30 is potentially interesting...

    http://forum.projectcars...p;viewfull=1#post1167830

    Text copied below.

    "Got some (hopefully) great news - I wrote a wrapper .dll to sit between PCars and pid.dll to see what was going on with FF. I spat out logs from both PCars and Assetto Corsa to check the differences. PCars destroys the FFB effects every time you quit the track, the crashes always happen when subsequently recreating (downloading) effects. Assetto Corsa doesn't destroy effects until quitting, but I spotted that after destroying the final effect, it always calls SendForceFeedbackCommand(dwId, DISFFC_RESET) to reset the FFB (PCars doesn't do this). I inserted this into my wrapper when it detects all forces have been destroyed and hey presto, PCars no longer crashes when creating new effects again! Hopefully the devs will see this."


    Yes, I have always suspected that it was the way that rFactor resets the FFB

    when rFactor2 crashes the last line of the log is

    hwinput.cpp 7874: Entered HWInput::Restart()

    Originally Posted by: Richard Macnaughtan [​IMG]

    I have a attached the rFactor2 log when it crashes

    The last line is

    hwinput.cpp 7874: Entered HWInput::Restart()

    [​IMG] trace.txt (32kb) downloaded 2 time(s).


    I figured that when it restarts the FFB then it caused some kind of bug to crash the game. It looks like they destroy the FFB every time you go back to the pits or start the game which creates this bug

    Originally Posted by: Richard Macnaughtan [​IMG]

    I think it is at a hardware driver error. I suspect that rFactor 2 is trying to restart the FFB or send some kind of command that the Accuforce doesn't like.

    It will either crash as soon as you spawn in to the garage or when you get in the car or when you go back to the garage after a session.


    And when I debugged the crash I saw that it was PID effect operation 56 which caused the crash. Maybe this is the same command which causes Project cars to crash.

    Originally Posted by: Richard Macnaughtan [​IMG]
    A bit more debug infomation

    Exception Information

    PID!PID_EFFECTOPERATION+56In rFactor2__PID__183932__Date__11_03_2015__Time_10_35_43PM__528__Second_Chance_Exception_C0000005.dmp the assembly instruction at pid!PID_EffectOperation+56 in C:\Windows\System32\pid.dll from Microsoft Corporation has caused an access violation exception (0xC0000005) when trying to read from memory location 0x9c340010 on thread 14

    Here is the full crash report

    [​IMG] CrashHang_Report__Date_11_03_2015__Time_10_38_29PM__422.zip (19kb) downloaded 4 time(s).


    Originally Posted by: Richard Macnaughtan [​IMG]
    Even more debug infomation.

    It looks like it is some type of memory access violation. I don't know why it only happens with the Accuforce plugged in and with the game running in 64bit mode

    FAULTING_IP:
    pid!PID_EffectOperation+56
    00007ffd`922d2ada 8b540838 mov edx,dword ptr [rax+rcx+38h]

    EXCEPTION_RECORD: (.exr -1)
    ExceptionAddress: 00007ffd922d2ada (pid!PID_EffectOperation+0x0000000000000056)
    ExceptionCode: c0000005 (Access violation)
    ExceptionFlags: 00000000
    NumberParameters: 2
    Parameter[0]: 0000000000000000
    Parameter[1]: 0000009b9c340010
    Attempt to read from address 0000009b9c340010


    SYMBOL_NAME: pid!PID_EffectOperation+56

    MODULE_NAME: pid

    IMAGE_NAME: pid.dll

    STACK_COMMAND: .ecxr ; kb

    BUCKET_ID: INVALID_POINTER_READ_pid!PID_EffectOperation+56

    PRIMARY_PROBLEM_CLASS: INVALID_POINTER_READ_pid!PID_EffectOperation+56

    FAILURE_PROBLEM_CLASS: INVALID_POINTER_READ

    FAILURE_EXCEPTION_CODE: c0000005

    FAILURE_IMAGE_NAME: pid.dll

    FAILURE_FUNCTION_NAME: PID_EffectOperation

    FAILURE_SYMBOL_NAME: pid.dll!PID_EffectOperation

    FAILURE_BUCKET_ID: INVALID_POINTER_READ_c0000005_pid.dll!PID_EffectOperation




    I am not a programmer so I can't write a wrapper but I suspect this PID effect operation 56 is trying to destroy the FFB effects like Project CARS does and this causes a crash with some USB controllers while in 64 bit mode.

    All this info backs up my hunches about how the games are resetting the FFB but I don't know why this is only happening on some USB controllers AND in 64 bit mode AND with the Accuforce. It seems like just that combo of hardware causes these problems and that is why it is so hard to reproduce.

    Originally Posted by: Danyel Grenzer [​IMG]


    Many many thanks to you, Richard [​IMG] You really solved (at least) my problem, but USB 3.0 ins´t the issue in my case, it´s really only the chipset. My Gigabyte GA-H97M-D3H have to USB 2.0/1.1 physical ports with the same problems.

    ...the two USB-case-ports seems to solve this issue in rFactor 2 and pCars[​IMG]



    So maybe those two ports are using a different USB controller chipset that isn't affected by this bug.

    It really does seem like it is something to do with how certain USB controllers handle this FFB reset/PID effect operation 56 command.

    The chipset for my USB 2.0 and back USB 3.0 works fine but the front USB must be using a chipset that has this bug. Same problem with Danyel, he must have 2 different USB chipsets, one which works and one which doesn't.

    I guess the solution for this would be to just keep trying different USB ports until you find one that works and if all else fails then buy a PCIe USB card which doesn't have this bug like my Etron one.

    Originally Posted by: Berney Villers [​IMG]
    Richard/Mark - I should find a spot on payroll for you guys!

    Nice work!

    I probably would have sat here for another week trying to reproduce the issue and never would have since I don't have a USB 3.0 driver with the issue.


    No probs. Glad to help"

    "Originally Posted by: Berney Villers [​IMG]
    As an FYI. Many motherboards have the Intel chipset provided USB and another USB chip that provides some USB 3.1, 3.0 and 2.0 ports.

    Here is an example of the USB specs for the Gigabyte Gaming 7 motherboard, which has 3 variants:

    Chipset:
    1.5 x USB 3.0/2.0 ports on the back panel
    2.4 x USB 2.0/1.1 ports (available through the internal USB headers)

    Chipset+Renesas® USB 3.0 Hub:
    1.4 x USB 3.0/2.0 ports (available through the internal USB headers)

    Chipset+Intel® Thunderbolt™ 3 Controller:
    1.1 x USB Type-C™ port on the back panel, with USB 3.1 support
    2.1 x USB 3.1 Type-A port (red) on the back panel

    This might explain why some folks have success by simply changing USB ports. By doing so, they're essentially changing USB chip and drivers to one that meets the standards and/or behaves better."

    I am receiving a new PCI-e USB 3.0 board today that, according to other AccuForce users, resolves this issue. I will get back to you once I test it. BTW, I asked SimXperience to grant you access to their forums. Not that I carry any weight but I think SIM developers should be involved.
    "
     
  13. Voltaic

    Voltaic Registered

    Joined:
    Sep 14, 2011
    Messages:
    327
    Likes Received:
    65
    I then ordered one of the confirmed-compatible PCI-e card;

    "@Marcel Offermans

    Marcel

    After limited testing once I received and installed this PCI-e card, I am happy to report I have not had any issues. I will be stress testing rFactor 2 tomorrow and get back to you.
    "

    And it has been working perfectly since then:

    "Hello and happy new year Marcel

    I am happy to say I have not experienced anymore rFactor 2 crashes since I installed the new PCI-e USB card. I hope this helps you and your team in the future, I will keep you informed if anything changes.
    "

    I don't know if this problem affects all controllers, I can say it was affecting all simulation games, I focused on rFactor 2 because it is my go to SIM.

    If you own an AccuForce you will be able to login at SimXperience and view the ".DLL" thread with further information.

    A great thank you to Marcel for his engagement with this issue! I hope this .
     
    Flaux and Euskotracks like this.

Share This Page