Any way to only add 1 out of 2 GPUs to gigahorse #889
whattheflux
started this conversation in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi team, I have dell r820 with 2x NVIDIA Tesla P4s. Unraid 6.11.5
I want to run one GPU attached to machinaris and leave one free for plotting direct from unraid CLI
Nvidia Driver Version:530.41.03
Open Source Kernel Module:No
Installed GPU(s):
0:
Tesla P4
04:00.0
GPU-d3fa90ce-8ecb-b9da-3ccb-0f38a4890172
1:
Tesla P4
44:00.0
GPU-99847a33-d393-b653-8eee-302e6399794d
In my docker configs I have extra parameters: --runtime=nvidia
Then in variables I have:
NVIDIA_DRIVER_CAPABILITIES: all
NVIDIA_VISIBLE_DEVICES:
Under NVIDIA VISIBLE DEVICES I've tried: 0 and 1 and GPU-d3fa90ce-8ecb-b9da-3ccb-0f38a4890172 but no matter what I do it always adds both to the container. I can see because when I run nvidia-smi from the CLI within the contianer both GPUs are visible and both are doing chia_harvester.
I disabled the CHIAPOS checking setting so at least it can farm now. Is there a bug that adds both or am I doing something wrong?
Thanks a bunch!
Beta Was this translation helpful? Give feedback.
All reactions