Getting Intel GPUs to work in your Kubernetes cluster is fairly straightforward.
For this tutorial I assume you have the following:
- A Working K3S cluster
- At least one node with some sort of Intel GPU
- Helm installed
This guide has been tested on:
- Debian 13
No additional changes should be needed for Talos Linux. You will need either the siderolabs/i915 extension or the siderolabs/xe extension
In order to effectively use your Intel GPU for quicksync we need to install a package, intel-media-va-driver-non-free. Since I use Debian's cloud images the step to enable to correct repositories is slightly different for some reason:
$sudo sed -i 's/^Components: main$/Components: main non-free non-free-firmware/' /etc/apt/sources.list.d/debian.sources
If you installed Debian normally you would want to edit /etc/apt/sources.list, the first line is likely the correct one to edit:
$sudo nano /etc/apt/sources.list
deb http://mirror.csclub.uwaterloo.ca/debian/ bookworm main
Stick non-free non-free-firmware onto the end so that it looks something like this:
deb http://mirror.csclub.uwaterloo.ca/debian/ bookworm main non-free-firmware contrib
Press Ctrl+X to save, then Y to confirm and then enter.
Nothing extra needs to be done.
$sudo apt-get update
$sudo apt-get install intel-media-va-driver-non-free
$sudo reboot
Optionally you can install vainfo to see what codecs your GPU supports:
$vainfo
error: can't connect to X server!
libva info: VA-API version 1.17.0
libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so
libva info: Found init function __vaDriverInit_1_17
libva info: va_openDriver() returns 0
vainfo: VA-API version: 1.17 (libva 2.12.0)
vainfo: Driver version: Intel iHD driver for Intel(R) Gen Graphics - 23.1.1 ()
vainfo: Supported profile and entrypoints
VAProfileNone : VAEntrypointVideoProc
VAProfileNone : VAEntrypointStats
VAProfileMPEG2Simple : VAEntrypointVLD
VAProfileMPEG2Simple : VAEntrypointEncSlice
VAProfileMPEG2Main : VAEntrypointVLD
VAProfileMPEG2Main : VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointVLD
VAProfileH264Main : VAEntrypointEncSlice
VAProfileH264Main : VAEntrypointFEI
VAProfileH264Main : VAEntrypointEncSliceLP
VAProfileH264High : VAEntrypointVLD
VAProfileH264High : VAEntrypointEncSlice
VAProfileH264High : VAEntrypointFEI
VAProfileH264High : VAEntrypointEncSliceLP
VAProfileVC1Simple : VAEntrypointVLD
VAProfileVC1Main : VAEntrypointVLD
VAProfileVC1Advanced : VAEntrypointVLD
VAProfileJPEGBaseline : VAEntrypointVLD
VAProfileJPEGBaseline : VAEntrypointEncPicture
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointEncSlice
VAProfileH264ConstrainedBaseline: VAEntrypointFEI
VAProfileH264ConstrainedBaseline: VAEntrypointEncSliceLP
VAProfileVP8Version0_3 : VAEntrypointVLD
VAProfileVP8Version0_3 : VAEntrypointEncSlice
VAProfileHEVCMain : VAEntrypointVLD
VAProfileHEVCMain : VAEntrypointEncSlice
VAProfileHEVCMain : VAEntrypointFEI
VAProfileHEVCMain10 : VAEntrypointVLD
VAProfileHEVCMain10 : VAEntrypointEncSlice
VAProfileVP9Profile0 : VAEntrypointVLD
VAProfileVP9Profile2 : VAEntrypointVLD
The codecs that have VAEntrypointEncSlice are the ones your gpu can encode, in my case I can encode up to and including H265 10bit. Your results may vary.
The Intel GPU Device Plugin chart we'll be installing will create a DaemonSet with a node selector of intel.feature.node.kubernetes.io/gpu: true so that any pods created by the DaemonSet will only run on nodes with Intel GPUs. So to label your nodes correctly we'll use Node Feature Discovery.
Node Feature Discovery uses rules to determine what to label and why. The Intel GPU Resource Driver that we'll be installing comes with pre-made rules to automatically label your nodes.
$helm repo add node-feature-discovery https://kubernetes-sigs.github.io/node-feature-discovery/charts
$helm repo update
$helm upgrade --install node-feature-discovery node-feature-discovery/node-feature-discovery
$helm repo add intel https://intel.github.io/helm-charts/
$helm repo update
$helm upgrade --install intel-gpu-resource-driver intel/intel-gpu-resource-driver --values values.yaml
Now that the chart's installed we need to create a ResourceClaimTemplate, the one provided in this repo can be used without modification by any deployment that you want to have use your Intel GPU (Jellyfin, Frigate, Tdarr, etc.)
$kubectl apply -f resourceClaimTemplate.yaml
And that's it! You are now ready to use your Intel GPU in a pod.
Note
ResourceClaimTemplates are namespaced objects. If your pods that you want to have use your Intel GPU are in different namespaces you'll want a ResourceClaimTemplate for each namespace.
To make sure it works run the test pod in vainfo-pod.yaml:
$kubectl apply -f vainfo-pod.yaml
Give the cluster a few seconds to start the pod then check its output:
$kubectl logs vainfo
libva info: VA-API version 1.22.0
libva info: User environment variable requested driver 'iHD'
libva info: Trying to open /usr/local/lib/dri/iHD_drv_video.so
libva info: Found init function __vaDriverInit_1_22
libva info: va_openDriver() returns 0
Trying display: drm
vainfo: VA-API version: 1.22 (libva 2.22.0)
vainfo: Driver version: Intel iHD driver for Intel(R) Gen Graphics - 24.3.3 (9876014)
vainfo: Supported profile and entrypoints
VAProfileNone : VAEntrypointVideoProc
*snip*
It works!
Remove the deployment:
$kubectl delete -f vainfo-pod.yaml
- October 16, 2026 - Overhauled guide, replacing obsolete charts with Intel GPU Resource Driver
- June 10, 2026 - Added information for Ubuntu 26.04, updated jellyfin tag
- February 9, 2026 - Spring cleaning, added test pod, adjusted valus.yaml slightly
- March 5, 2024 - Initial commit