[slurm-users] NVML autodetect "Failed to get supported memory frequencies" error

[Joshua Baker-LePain]

On Fri, 5 Mar 2021 at 7:51am, Kilian Cavalotti wrote

> Hi Joshua,
>
> On Thu, Mar 4, 2021 at 8:38 PM Joshua Baker-LePain <jlb at salilab.org> wrote:
>> slurmd: error:  _nvml_get_mem_freqs: Failed to get supported memory frequencies
>> slurmd: error:  for the GPU : Not Supported
>
>> slurmd:  4 GPU system device(s) detected
>> slurmd:  WARNING: The following autodetected GPUs are being ignored:
>> slurmd:      GRES[gpu] Type:geforce_gtx_1080 Count:1 Cores(1024):28-55
>
> It's just a warning, informing that getting (and setting) memory
> frequencies is not supported on GeForce GPUs.
> You won't be able to use the srun/salloc/sbatch --gpu-freq option for
> those GPUs, but that's about it. Everything else should work normally.

Ah, that makes sense.  The "GPUs are being ignored" bit was throwing me 
off, but that was simply b/c I hadn't added the GRES to the node 
configuration yet.  Thanks!

-- 
Joshua Baker-LePain
Wynton Cluster Sysadmin
UCSF