Hello. I’m using Unreal 4, 4.25. When I try to render out a movie or take a high resolution screenshot, Unreal 4 and my GPU crash. I’m using Raytracing and Direct X 12. It will crash and a window will appear saying,
GPU Crashed or D3D Device Removed.
Use -d3ddebug to enable the d3d debug device.
Check log for GPU state information.
I’m not sure if it’s my GPU and need to send it in for repair, or my GPU isn’t getting enough power. Or Unreal 4 issue. I have a 2080Ti, Gaming X Trio. 650 Watt power supply. 64GB Ram, 3900X Ryzen. Any information will help. Thanks.
I had this issue for 2 days now and it seems to be fixed if you switch to DirectX 11 in the project settings. For some reason DX 12 is unstable.
Hope this helps.
Thanks for the reply. When I switch to Direct X 11, it seems fixed but the quality of my scene is drastically reduced and I have no control over the ray trace feature for the post process volume. Which sucks. I guess it is an engine problem and not my Graphics card problem.
I am having issues with this same error when rendering out ray traced scenes in the movie render que. Rig; RTX 3080 ,3700x & 32gbRAM.
At first I thought it was because I had a 650w PSU, so I upgraded to 850w but it made no difference. Swapping from game to studio drivers made no difference.
Previously I had a RTX 2070s and it would sometimes crash as well, but not in the same way. The 2070 would crash by timeout or running out of memory, but that isn’t the case with the 3080. It crashes the moment it finishes the warmup frames and starts to render frame 1.
Video of crash happening- https://youtu.be/A76ekBBG_a8
For a workaround I set the samples per pixel to 1 for all (Gi, Ref, Trans) and set aa to 20 temporal and 3 spatial samples, and changing nothing else, or alternatively using tiling but I need bloom so this is usually a no go. With all spp at 1 it does 3-4 sub samples per second. With spp any higher it crashes.
It is strange because with the 2070 I could really push it on the samples per pixel front, rendering 3200X1800 with 4 or 8sppGi going out to 8-10 seconds per frame/subsample and it would only crash if it really just couldn’t make it, and it would go like that overnight most of the time. but with the 3080, it either renders the frame or the subsample in less than a second or it crashes.
When the 2070 would crash I would have to Hard re-start the PC cos everything would just freeze, or get an out of memory error, but as you can see in this case the GPU even keeps recording the screen vid flawlessly though the crash.
I’m not sure where the issue is, in the driver, unstable clock speeds, momentary loss of power from everything going all out at once, not enough Vram or maybe something with the engine, but at the moment, its really not much better for rendering than the 2070, even though it has double the horsepower along with the expected gains in viewport FPS.
Things I still have left to try:
-Full clean driver wipe and re-install
-put 2070 back in to compare on this scene with same settings
-Drop texture resolutions and increase rays per pixel.
I might also try underclocking the GPU and lowering the voltage a little, see if that makes it more stable. I only have a 60hz screen so it’ll still be huge overkill for gaming lol.
I used DDU to re-install the drivers and tried a few other things, but in the end it was this TDR tool that fixed the problem.
Now I can render at 4k with many Spp on Gi, trans and Ref on this very heavy foliage scene with 95% Vram usage and no crashes at warmup stage like before.
Hi,
The TDR tool can change how long windows will wait for a response for the GPU before deciding it has crashed and reset the driver.
Windows restarting the driver because it thinks it has crashed is what causes the crash in this case.
it is only adding some time between the crashes
, but this really helped me!