#HPC #supercomputing #Linux Note, there seems to be a bug within the latest F31 BIOS from Gigabyte and NVIDIA's 590.48.01 CUDA drivers. Previously, we could reset GPU's after each job, but with this version combination, the GPU goes "missing" and requires a reboot and driver reinstall. This behavior persists, as well. So far, rolling back to F30 and F19 BIOS' seems to negate the problem.
#HPC #supercomputing #Linux Downgrading to a previous BIOS allows one to skip the driver re-installation.
#HPC #supercomputing #Linux The downgrade to Gigabyte BIOS F30 still produced the issues. Downgrading to F19 restored previous behavior.
Annnnnnd the BIOS with a known issue is still available for download. #wtf #hpc #supercomputing #linux