Unreal 5.5 Path Tracer GPU Dumps Abnormally

I am running into a weird issue when trying to use the path tracer in my Unreal Engine project. What’s odd is that I’ve been able to use the path tracer in the past on this PC and I’m not sure what’s changed since the time it was working.

I’m able to use the path tracer on my other PC which is running the same project via Perforce source control. The only difference is that PC is a 4090 while the problem PC is a 5090.

Essentially what happens is when I try to switch the viewport to Path tracing mode, all of my static meshes turn black and It says that shaders are being prepared in the top left of my viewport. Then after 30 seconds or so it issues a GPU dump and crashes. As soon as I switch the viewport mode to path tracing my PC essentially freezes up and I can’t do anything until it crashes.

I really am at a loss. I have already tried these solutions:

  • Updated to latest GPU Drivers
  • Changed TDR Delay to 60sec
  • Downgraded to Dec. 2025 Drivers
  • Verified UE5.5 Install
  • Lowered Sample Count for Path Tracing
  • Deleted “Saved, Intermediate, & DerivedDataCache” folders then regenerated upon launch
  • Increased Virtual Texture Pool Size to 4000mb
  • Disabled Virtual Textures

Here is the crash report and log (looks like a typical GPU dump crash):

LakeAve.log (15.5 MB)

GPU Crash dump Triggered

UnrealEditor_D3D12RHI
UnrealEditor_D3D12RHI
UnrealEditor_D3D12RHI
UnrealEditor_D3D12RHI
UnrealEditor_D3D12RHI
UnrealEditor_Core
UnrealEditor_Core
kernel32
ntdll

Hello there @RB_Brendan!

Checking through the massive log, the main takeaway is that the path tracing shader compile process does complete, then crashes due a timeout. Yet, there are more than a few angles to consider:

[2026.01.23-22.06.12:413][809]LogD3D12RHI: Error: PageFault: PageFault at VA GPUAddress “0xFEFE000000” (GPU 0)

This specific error code points towards a handshake or memory access issue, which could be tied to either the drivers, or the GPU hardware itself.

[2026.01.23-22.05.09:006][807]LogStaticMesh: Warning: Invalid material [Air] used on Nanite static mesh [Sculpted_Site_GH_HotTub_Grass]. Only opaque or masked blend modes are currently supported, [BLEND_Translucent] blend mode was specified. (NOTE: “Disallow Nanite” on static mesh components can be used to suppress this warning and forcibly render the object as non-Nanite.)

[2026.01.23-22.05.09:048][807]LogStaticMesh: Warning: Invalid material [Flower_petal] used on Nanite static mesh [Salvia_nemorosa_004]. Only opaque or masked blend modes are currently supported, [BLEND_Translucent] blend mode was specified. (NOTE: “Disallow Nanite” on static mesh components can be used to suppress this warning and forcibly render the object as non-Nanite.)

There are multiple entries like this, close to a hundred, pointing at invalid material references on Nanite meshes using translucent blend modes.

[2026.01.23-22.03.26:623][ 0]LogScript: Warning: Attempted to access TRASH_InstancedStaticMeshComponent_48 via property proxySM, but TRASH_InstancedStaticMeshComponent_48 is not valid (pending kill or garbage)

This particular blueprint has several entries tied to invalid components and array conflicts.

[2026.01.23-22.05.09:494][809]LogD3D12RHI: Creating RTPSO with 135 shaders (126 cached, 9 new) took 12.51 ms. Compile time 9.82 ms, link time 2.65 ms.
[2026.01.23-22.05.24:466][809]LogD3D12RHI: Warning: GPU timeout: A payload (0x00000B66AA7A0000) on the [0x00000B61F1342800, Compute] queue has not completed after 5.000001 seconds.
[2026.01.23-22.05.24:474][809]LogD3D12RHI: Warning: GPU timeout: A payload (0x00000B678852E700) on the [0x00000B61F1342300, 3D] queue has not completed after 5.000001 seconds.
[2026.01.23-22.06.12:412][809]LogD3D12RHI: Error: GPU crash detected: Device 0 Removed: DXGI_ERROR_DEVICE_HUNG

And here’s the point where the compilation does complete, then the GPU hangs during execution, and the whole thing crashes. Now that we know all the points of conflict in the log, let’s try to fix them one by one:

  • Use DDU to fully clear your drivers, then install the latest Studio drivers from nVidia, as these are the most stable for engine work

  • Check all your assets using translucent materials with Nanite enabled, and either disable Naninte from them, or convert the materials to masked/opaque. A more nuclear approach would be to use no Nanite at all, setting up r.PathTracing.UseNanite 0

  • Open any BPs tied to BP_ia_scatter_C and look for red error nodes, null references, and any other conflicts in the arrays. You could also disable it temporary and check the changes in the log

  • Disable HAGS in Windows, as other cases in the community point it as a source of conflict. To do so, go to Windows Settings > System > Display > Graphics > Advanced graphics settings, and there, turn off “Hardware accelerated GPU scheduling”

  • You mentioned setting up TdrDelay to 60, great approach, but please remember to set TdrDdiDelay to 60 as well, as it was not mentioned in your notes

  • Optionally, you can test path tracing on a new, empty level, in order to rule out specific project issues

  • Finally, after all these changes, please remember to repeat the cache clear (Deleted “Saved, Intermediate, & DerivedDataCache” folders then regenerated upon launch). After that, test your project again.

Hope this helps, and good luck!