I am getting the error message below after usually 2 frames rendering fine. this basically stalls the render. would anybody have an idea how to debug this? the scene used to render fine, I have since tried a variety of GPU drivers (game- and studio) and I also downloaded the bubblewrap scene from sideFX which renders fine across multiple frames. enabling /disable denoise also did not change anything

KarmaXPU: device Type:Optix ID:0 has registered a critical error , so will now stop functioning. Future error messages will be suppressed

so any hint on how to narrow this down and get my scene back up and rendering would be much much appreciated!
many thanks


this happens in both 20.0.625 and 20.0.547 under win11, 3080ti

The old standby answer might help. Monitor your GPU temperature. Also consider lowering your overclock by a half-step.

Enivob
The old standby answer might help. Monitor your GPU temperature. Also consider lowering your overclock by a half-step.

thanks for your input, I am not overclocking at all and the GPU chugs along at a balmy 70-80 degrees (C), memory usage around 7GB (of 12GB)


interesting finding:
If I reduce the amount of samples, the error message comes later after rendering more frames successfully.

300 (and any value above) samples = crash after 2 successfully rendered frames
30 samples = crash after 7 frames
10 samples = crash after 104 frames

I am monitoring the vram usage and i am well below the hardware limit at all times

sort of narrowing it down. I have emissive spheres inside a piece of geometry with a shader that has transmission as well as SSS active. if I disable the SSS it seems i am able to render, obvs it changes the look though, but maybe there is something fundamental that I don't understand how to achieve what I'm after here

'be curious to know why this is happening and not just resulting in too long render times (again, VRM usage looks fine and below the HW limit)

anyone at sideFX care to chime in?

It might help if you could post the scene. Then others could test on their hardware. It could be a SSS bug, though. Use the Support menu up above to report the bug and supply your file.