GPU passthrough allows a virtual machine to directly access a physical GPU on the host node. A virtual machine can directly access the physical GPU resource on a Kubernetes node, thereby significantly improving the performance of graphics-intensive applications, e.g., deep‑learning training, scientific computing, or graphics rendering.

Before enabling GPU passthrough, the GPU can only be used by pods. After enabling, a virtual machine can access and use the physical GPU. The administrator can flexibly allocate GPU resources to different virtual machines as needed.

Procedure

1. Go to EonKube > Cluster > Nodes.
2. Find a node with a GPU installed.
3. Select the node and click Enable GPU Passthrough.
4. Restart the node.
5. Go to EonKube > Virtualization > GPUs. You can view the enabled GPU resource for the virtual machine.