I've been kicking around with this all day and hoping someone can help. I've successfully added the --runtime=nvidia parameter and the NVIDIA_DRIVER_CAPABILITIES variable and plex updates and starts. However, adding NVIDIA_VISIBLE_DEVICES variable with the UUID value (or 0 or all) throws up the following:
root@localhost:# /usr/local/emhttp/plugins/dynamix.docker.manager/scripts/docker run -d --name='plex' --net='host' -e TZ="Europe/London" -e HOST_OS="Unraid" -e 'PUID'='99' -e 'PGID'='100' -e 'VERSION'='latest' -e 'NVIDIA_DRIVER_CAPABILITIES'='all' -e 'NVIDIA_VISIBLE_DEVICES'='GPU-0e64b88f-fa47-56a5-d740-5f989b63fc65' -v '/mnt/user/appdata/plex':'/config':'rw' --runtime=nvidia 'linuxserver/plex'
458cbf493f1f14c2f87df492eb789360284d4ed61b3f24bcf8e23fd61fbb405e
/usr/bin/docker: Error response from daemon: OCI runtime create failed: container_linux.go:344: starting container process caused "process_linux.go:424: container init caused "process_linux.go:407: running prestart hook 0 caused \"error running hook: exit status 1, stdout: , stderr: exec command: [/usr/local/bin/nvidia-container-cli --load-kmods --debug=/var/log/nvidia-container-runtime-hook.log --ldcache=/etc/ld.so.cache configure --ldconfig=@/sbin/ldconfig --device=GPU-0e64b88f-fa47-56a5-d740-5f989b63fc65 --compute --compat32 --graphics --utility --video --display --pid=11393 /var/lib/docker/btrfs/subvolumes/bbb1ee999ab52a47f5db1e25ffe306072910df3fafff7d19c1c72d092e7e8f8d]\\nnvidia-container-cli: initialization error: cuda error: initialization error\\n\""": unknown.
The command failed.
The log:
May 19 15:57:08 Tower login[10845]: ROOT LOGIN on '/dev/pts/0'
May 19 15:57:15 Tower kernel: resource sanity check: requesting [mem 0x000c0000-0x000fffff], which spans more than PCI Bus 0000:00 [mem 0x000d4000-0x000d7fff window]
May 19 15:57:15 Tower kernel: caller _nv000934rm+0x1bf/0x1f0 [nvidia] mapping multiple BARs
May 19 15:59:33 Tower kernel: resource sanity check: requesting [mem 0x000c0000-0x000fffff], which spans more than PCI Bus 0000:00 [mem 0x000d4000-0x000d7fff window]
May 19 15:59:33 Tower kernel: caller _nv000934rm+0x1bf/0x1f0 [nvidia] mapping multiple BARs
May 19 15:59:33 Tower kernel: nvidia 0000:01:00.0: swiotlb buffer is full (sz: 4194304 bytes)
May 19 15:59:33 Tower kernel: nvidia 0000:01:00.0: swiotlb buffer is full