Closed dasantonym closed 2 months ago
After further review, I don't see how this would be related to Firestarter. More likely something with our setup, sorry!
Actually, it was not the setup. gpu_burn worked fine and everything looks normal.
I currently can't reproduce it on our A100 system with CUDA 11.1 installed. We will have a more detailed look at the issue. Sorry for the inconvenience.
I now also tried it with CUDA 11.6 on our A100 partition and still cannot reproduce your problem. Could you list the output of
ldd ./FIRESTARTER_CUDA_11.0
? Thanks.
@dasantonym Could you check whether this persists, if we do not hear back, we will close the issue.
Closed due to no feedback. Maybe fixed with update.
I tried to run the prebuilt CUDA 11 binary on an Ubuntu 20.04 system with Cuda 11.6 and eight A100 GPUs.
When it gets to the GPUs, the following error occurs:
There seems to be no error for device 1.
nvidia-smi
reports all eight GPUs normally.