Bsod on Nvidia 50 series when running "SynthBenchmark"

Development Pipeline & Plugins
, ,
1

The 5080 can run through all 7 benchmarks once, if your have 70 iterations it will bsod. The 5070 crashes if any 3 are run at the same time.

​

5.5 did introduce new gpu profiler, I haven’t tested disabling this, but it’s next on my plate.

2

Steps to Reproduce
Run “SynthBenchmark” on any 50 series Nvidia card. It’ll bsod. It happens in our title, and in vanilla ue5.5 editor. I confirmed it happens on a founders 5070, and a retail 5080.

​

It does not happen on vanilla editor 5.4, both are the prebuilt editors from EGS.

3

Hi, thanks for reporting this issue. We’ve had several reports of this on the 50 series with driver versions 576.xx and higher and posted a technote for TwinMotion about it and we’ll likely recommend something similar for vanilla UE:

https://dev.epicgames.com/community/learning/knowledge-base/pBjb/twinmotion-known-issues-with-nvidia-572-06-drivers-on-rtx-50-series-gpus

The technote has several recommendations depending on the specific GPU you have. We haven’t identified what is happening in the benchmark to expose this crash, but NVidia is working on a driver fix.

4

For us. We can run the full gamut of benchmarks on the 5080, up to about 30 iterations and it doesn’t crash. Any more and it will. For the 5070, we can run any two benchmarks with one iteration without crashing, but once you try any 3, BSOD.

We wrote some code to mask benchmark passes, and modify the iteration count for testing.

We also have a chat with Nvidia, but the back and forth is slow.

5

Have you found that lowering the r.SynthBenchmark.GPUWarmupRounds andr.SynthBenchmark.GPUWarmupThreshold cvars helps?

6

i haven’t tested that but i will and get back to you.

7

when i set the warmup frames to 0 it lets us raise the iteration count to about 40 before crashing, full benchmark suite.

8

Thanks for added information - NVidia just released a new driver (576.80) that appears to fix this issue, can you let us know if it resolves the issue for you?

9

Yup, this driver fixes it.

10

Hey, i already got crash because of „SynthBenchmark” When launching fortnite and waiting for applying settings for my hardware. GPU Memory: 7807/0/16315 MB
[2025.11.24-21.52.08:687][844]LogSynthBenchmark: Display: GPU warmup (round 1) test: 0.16s
[2025.11.24-21.52.08:687][844]LogSynthBenchmark: Display: … 34.587 s/GigaPix, Confidence=81% ‘ALUHeavyNoise’ (likely to be inaccurate)
[2025.11.24-21.52.08:687][844]LogSynthBenchmark: Display: … 38.578 s/GigaPix, Confidence=89% ‘TexHeavy’ (likely to be inaccurate)
[2025.11.24-21.52.08:687][844]LogSynthBenchmark: Display: … 37.892 s/GigaPix, Confidence=81% ‘DepTexHeavy’ (likely to be inaccurate)
[2025.11.24-21.52.08:687][844]LogSynthBenchmark: Display: … 117.858 s/GigaPix, Confidence=89% ‘FillOnly’ (likely to be inaccurate)
[2025.11.24-21.52.08:687][844]LogSynthBenchmark: Display: … 1.428 s/GigaPix, Confidence=100% ‘Bandwidth’ (likely to be inaccurate)
[2025.11.24-21.52.08:687][844]LogSynthBenchmark: Display: … 6.198 s/GigaVert, Confidence=89% ‘VertThroughPut1’ (likely to be inaccurate)
[2025.11.24-21.52.08:687][844]LogSynthBenchmark: Display: … 22.015 s/GigaVert, Confidence=100% ‘VertThroughPut2’ (likely to be inaccurate)
[2025.11.24-21.52.13:918][844]LogD3D12RHI: Error: GPU crash detected:

  • Device 0 Removed: DXGI_ERROR_DEVICE_HUNG

[2025.11.24-21.52.13:919][844]LogD3D12RHI: Error: Shader diagnostic messages and asserts:

Device: 0, Queue 3D:
    No shader diagnostics found for this queue.

Device: 0, Queue Copy:
    No shader diagnostics found for this queue.

Device: 0, Queue Compute:
    No shader diagnostics found for this queue.

[2025.11.24-21.52.13:919][844]LogRHI: Error: Active GPU breadcrumbs:

Device 0, Pipeline Graphics: (In: 0x8000c19f, Out: 0x8000bf6f)
    (ID: 0x8000bf6b) [     Active]    Frame 40844
    (ID: 0x8000bf6c) [     Active]        WorldTick
    (ID: 0x8000bf6f) [     Active]            Scale:1.500000
    (ID: 0x8000c198) [     Active]                Iteration:69
    (ID: 0x8000c19f) [     Active]                    Benchmark Method:6

Device 0, Pipeline AsyncCompute: (In: 0x8000bf6e, Out: 0x8000bf6e)
    No breadcrumb nodes found for this queue.

[2025.11.24-21.52.16:923][844]LogNvidiaAftermath: Warning: Timed out while waiting for Aftermath to start the GPU crash dump.
[2025.11.24-21.52.16:923][844]LogD3D12RHI: Error: DRED: No breadcrumb head found.
[2025.11.24-21.52.16:923][844]LogD3D12RHI: Error: DRED: No PageFault data.
[2025.11.24-21.52.16:923][844]LogD3D12RHI: Error: Video Memory Stats from frame ID 795:
[2025.11.24-21.52.16:923][844]LogD3D12RHI: Error: Local Budget: 7039.00 MB
[2025.11.24-21.52.16:923][844]LogD3D12RHI: Error: Local Used: 459.57 MB
[2025.11.24-21.52.16:923][844]LogD3D12RHI: Error: System Budget: 15546.54 MB
[2025.11.24-21.52.16:923][844]LogD3D12RHI: Error: System Used: 155.52 MB
[2025.11.24-21.52.16:956][844]LogD3D12RHI: Error: GPU Crashed or D3D Device Removed.

Check log for GPU state information.