5Stack

Appearance

NVIDIA GPU Setup ​

Game streaming on NVIDIA GPUs needs three things on every node that will run a streamer pod:

  1. The NVIDIA driver and CUDA userland.
  2. The NVIDIA Container Toolkit (so containerd can expose GPUs into pods).
  3. The Kubernetes nvidia overlay applied to your cluster (registers the device plugin so pods can request nvidia.com/gpu).

1. Install drivers and the container toolkit ​

On Ubuntu / Debian-based hosts:

bash
sudo apt install -y \
    nvidia-cuda-toolkit \
    nvidia-container-toolkit \
    nvidia-container-runtime \
    cuda-drivers-fabricmanager-590 \
    nvidia-headless-590-server-open \
    nvidia-utils-590-server \
    nvidia-driver-590-server-open

Driver version

The 590 package suffix tracks the driver branch. Pick the branch that matches your GPU and CUDA version — nvidia-driver-590-server-open is a recent open-kernel branch that works for the streamer pipeline. If apt doesn't have it, you may need NVIDIA's cuda-keyring repo: see NVIDIA's installation guide.

Reboot, then verify:

bash
nvidia-smi

You should see your GPU(s) listed with the driver version, and you can continue to the game streaming setup.