DirectX 12 causing computer to freeze (black screen) even after adjusting TDRdelay while rendering in Movie Render Queue

Development Rendering
, , ,
1

I’ve already done some research on this problem from previous posts on the forum and made some changes on my settings, but the situation did not improve. And most of all, the symptoms that my computer showed is somewhat distinct compared to those mentioned in those posts.

I am using Lumen and Nanite to render in Unreal 5.3. I am rendering a sequence from sequencer in Movie Render Queue. The current RHI is Directx12 (essential for nanite and lumen). For anti-aliasing, Spatial Sample is set to around 50 and temporal sample 1 (I have physics simulation in the sequence and thus cannot bump up the temporal sample). My computer is quite beefy, with a 13th Gen i9-13900KF and a rtx4080 with 16GB of vram.

When I start rendering, the computer have quite a high chance to stuck in a black screen, and the fan will starts spinning really fast, but the temperature of both GPU and CPU is normal, around 60-70 C degrees. The only way to exit this status is to restart the computer by a long press on the power button, since it will not take any other keyboard or mouse input.

When the log for that Unreal project is checked, there are always these 3 lines of warning at the end of the log when the crash happens:

[2024.03.26-09.29.34:899][823]LogD3D12RHI: Warning: GPU timeout: A payload (0x000006A087092100) on the [0x0000069F5AF26B00, 3D] queue has not completed after 5.000000 seconds.

[2024.03.26-09.29.34:900][823]LogD3D12RHI: Warning: GPU timeout: A payload (0x000006A08709A600) on the [0x0000069F5AF27100, Compute] queue has not completed after 5.000000 seconds.

[2024.03.26-09.29.34:900][823]LogD3D12RHI: Warning: GPU timeout: A payload (0x000006A08709BE00) on the [0x0000069F5AF26E00, Copy] queue has not completed after 5.000000 seconds.

I am not familiar with RHI, but judging from it being a “timeout”, I followed the official guide from Epic: How to Fix a GPU Driver Crash and added the TDRdelay items in my registry, but the situation did not improved. What I did found, is that by quiting background softwares such as creative cloud from Adobe and Autodesk desktop client, the chance of the crash will drop down. However, even with these softwares running, by no means my Vram or Ram or any other hardware is being drived over its limits (judging from my monitoring of the task manager), which is why the crash confused me a lot.

Any help or hint or just a general direction for me to work to will be greatly appreciated!
Thanks.

2

Hi,

I don’t know the direct cause or immediate resolution based on this but I can give a few points and tips that may resolve the issue or similar issues.

  1. Upgrade your video card drivers (you already did this but listing for others)
  2. Try Disabling MPO - After updating to NVIDIA Game Ready Driver 461.09 or newer, some desktop apps may flicker or stutter when resizing the window on some PC configurations | NVIDIA
    Note: This will often fix weird context menu disappearing.
  3. Turn Off Instant Replay in Geforce Experience and Record in Background for Game Bar.
  4. Download GPU-Z and see if you notice anything weird like being throttled by PWR or THML. TechPowerUp - GPU-Z . You might need to replace your power supply or thermal paste.
  5. Try running a GPU benchmark program and see if your PC crashes during it. This may hint at a memory problem. UNIGINE Benchmarks
  6. Try running OBS while doing this task (you don’t need to stream or record). It does a pretty good job at controlling the RHI active window which could be the root of the problem.
  7. Try closing Discord while doing the render. Discord is constantly pushing out patches that mess with directx.
1 Like