UnReal crash out of game.

olRocketBoyolRocketBoy Posts: 1Member
Third leg of Sector 2 for about the 10th time.. engaged by about 3 enemy ships and was turning to respond when it died.

Custom Graphics set to EPIC at 3440x1440. Looked great until everything went black..


LoginId:953f3c5a4ba1f8bb6f3337a1abb88c07
EpicAccountId:

LowLevelFatalError [File:Unknown] [Line: 200] Unreal Engine is exiting due to D3D device being lost. (Error: 0x887A0007 - 'RESET')

RSG_Win64_Shipping!FDebug::AssertFailed() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\core\private\misc\assertionmacros.cpp:425]
RSG_Win64_Shipping!TerminateOnDeviceRemoved() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\windows\d3d11rhi\private\d3d11util.cpp:213]
RSG_Win64_Shipping!VerifyD3D11Result() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\windows\d3d11rhi\private\d3d11util.cpp:249]
RSG_Win64_Shipping!FD3D11DynamicRHI::GetQueryData() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\windows\d3d11rhi\private\d3d11query.cpp:147]
RSG_Win64_Shipping!FD3D11DynamicRHI::RHIGetRenderQueryResult() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\windows\d3d11rhi\private\d3d11query.cpp:61]
RSG_Win64_Shipping!FetchVisibilityForPrimitives_Range<1>() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\renderer\private\scenevisibility.cpp:830]
RSG_Win64_Shipping!FetchVisibilityForPrimitives() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\renderer\private\scenevisibility.cpp:1378]
RSG_Win64_Shipping!OcclusionCull() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\renderer\private\scenevisibility.cpp:1527]
RSG_Win64_Shipping!FSceneRenderer::ComputeViewVisibility() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\renderer\private\scenevisibility.cpp:3002]
RSG_Win64_Shipping!FDeferredShadingSceneRenderer::InitViews() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\renderer\private\scenevisibility.cpp:3324]
RSG_Win64_Shipping!FDeferredShadingSceneRenderer::Render() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\renderer\private\deferredshadingrenderer.cpp:571]
RSG_Win64_Shipping!RenderViewFamily_RenderThread() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\renderer\private\scenerendering.cpp:2854]
RSG_Win64_Shipping!TGraphTask<`FRendererModule::BeginRenderingViewFamily'::`20'::EURCMacro_FDrawSceneCommand>::ExecuteTask() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\core\public\async\taskgraphinterfaces.h:829]
RSG_Win64_Shipping!FNamedTaskThread::ProcessTasksUntilQuit() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\core\private\async\taskgraph.cpp:573]
RSG_Win64_Shipping!RenderingThreadMain() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\rendercore\private\renderingthread.cpp:333]
RSG_Win64_Shipping!FRenderingThread::Run() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\rendercore\private\renderingthread.cpp:467]
RSG_Win64_Shipping!FRunnableThreadWin::Run() [i:\intermediate\rsg_configurable_win\engine\engine\source\runtime\core\private\windows\windowsrunnablethread.cpp:76]

Answers

  • PedalToTheMetalPedalToTheMetal Posts: 100Member
    https://docs.microsoft.com/en-us/windows-hardware/drivers/display/tdr-registry-keys Set TDR to a higher value, 8 seconds feels sane enough.
  • ROCKFISH_AndiROCKFISH_Andi Posts: 1,017Moderator
    Hi, you're getting the infamous and well-known Unreal issue "Unreal Engine is exiting due to D3D device being lost.". It seems to happen when it's too much for your graphics card to handle. Other players have managed to get rid of it by toning the overclocking of their graphics card down (yours might be factory overclocked) and/or playing at lower graphics settings.
  • PedalToTheMetalPedalToTheMetal Posts: 100Member
    Andi, I'd argue with this. Usually people report the same error with the exact error code reading 'HUNG'. In this particular case we have a 'RESET' code instead, which is something different (though seems both can be avoided by setting up a proper TDR delay). Honestly, nowadays overclocking your GPU is the least likely cause of any problems as, unlike before, the GPUs have very smart things running behind the scenes and tweaking your clocks, like TDP estimation, VRM load limiting and such, all either in their BIOS or even hard-wired into the chips, so unless you modify the BIOS there is virtually no way to make it run outside the factory limits.
    I guess we should try testing the TDR effect on some weaker GPU (sorry, I got none at the moment - either an OC 1080 or some integrated netbook AMD, one with only 64 shading units running 256 MHz, which barely handles UE4 at all - though probably the latter is worth a shot).
  • ROCKFISH_AndiROCKFISH_Andi Posts: 1,017Moderator

    Andi, I'd argue with this. Usually people report the same error with the exact error code reading 'HUNG'. In this particular case we have a 'RESET' code instead

    Oh, your're right - I missed that.

    Honestly, nowadays overclocking your GPU is the least likely cause of any problems as, unlike before, the GPUs have very smart things running behind the scenes and tweaking your clocks, like TDP estimation

    I don't now, I'm no expert on the topic, but we did get feedback that toning the overclocking down indeed helped in a lot of these cases (if you google the problem it seems to be the case for many unreal engine games)

  • PedalToTheMetalPedalToTheMetal Posts: 100Member
    Probably, but there is one thing in common between all those cases: this mostly happens on Windows 10 (probably 8 as well) - and I never read of anything like this on Windows 7. The clear difference between those systems is a reduced TDR time (from 8 to only 2 seconds), which will definitely decrease stability of any sort of one-time heavy GPU loads (loading-stage compute shaders, single-run long draw calls and such). The overclocking can only make things worse while not being a direct cause, as under the above conditions your GPU is no longer bottle-necked by any other things (refresh rate if under VSync, CPU, texture streaming and such), thus making an already hot chip even hotter. The BIOS will figure the system needs to cool down and will heavily limit the TDP by dropping the clock rates even way below the stock ones, thus increasing the shader execution times. Bad nobody actually bothered making a, say, GPU-Z clock/TDP graph to inspect the issue.
    The only other OC-related cause is a misrouted PSU rail (say, a GPU with two 12v connectors is fed from the same rail, or any GPU shares a rail with some heavy load (like MB's VRM), making the GPU BIOS (which tracks this value) "see" a high difference in - or overall high values of - the 12v input currents of the VRM (the latter happens when 12v is dropped too much, which requires the VRM to take more current in order to sustain the required power level). The BIOS will do many bad things in this case, though this is a somewhat rare thing anyway (the PSUs usually have the cables designed in such a way this will never happen unless using something shady like 4-to-8 pins jumper cables).
Sign In or Register to comment.