LXC container GPU sharing is the most efficient method for homelab hardware video transcoding and local AI inference. By mounting kernel device nodes directly, you bypass QEMU emulation overhead, achieving bare-metal performance while consuming zero additional idle watts.
Step 1: Identify Host Device Major/Minor Numbers
In the Proxmox host terminal, inspect permissions for your GPU devices:
ls -l /dev/dri
# Output:
# crw-rw---- 1 root video 226, 0 Oct 31 10:00 card0
# crw-rw---- 1 root render 226, 128 Oct 31 10:00 renderD128
Take note of major number 226 and minor numbers 0 and 128, alongside the GID of group render (usually 104 or 106).
Step 2: Configure Container Configuration File
Edit your container file on the host at /etc/pve/lxc/{CTID}.conf and append:
# Pass through Intel QuickSync DRI nodes
lxc.cgroup2.devices.allow: c 226:0 rwm
lxc.cgroup2.devices.allow: c 226:128 rwm
lxc.mount.entry: /dev/dri/card0 dev/dri/card0 none bind,optional,create=file
lxc.mount.entry: /dev/dri/renderD128 dev/dri/renderD128 none bind,optional,create=file
Start your container and verify with vainfo inside the guest: Intel QuickSync will report full H.264, HEVC, and AV1 profile support.
Verified Hardware Tested in This Blueprint
Tested in our physical engineering lab for power consumption, thermal stability, and Linux driver compatibility.