GPU Resources

VKS offers several GPU models. Choose the one that fits your workload when provisioning a cluster.

GPU model table

Using GPUs

In a container you can request and limit GPUs by setting the resource key (the GPU label) and the GPU count.

Request example:

• GPU label: nvidia.com/gpu-h800
• GPU count: 1
• cpu: 18
• memory: 200G

Limit example:

• GPU label: nvidia.com/gpu-h800
• GPU count: 1
• cpu: 18
• memory: 200G

Tip: For each GPU, the CPU and memory must not exceed the per-GPU caps in the table above. With multiple GPUs, scale CPU and memory in proportion. Write the CPU as a plain number — no c suffix (e.g. cpu: 18).

Code example

spec:
  imagePullSecrets:
    - name: your-harbor-secret
  containers:
    - name: your-container-name
      image: your-image:tag   # replace with your own image
      command: your-command
      resources:
        requests:
          memory: "200G"
          cpu: "18"
          nvidia.com/gpu-h800: 1 # request 1 GPU
        limits:
          memory: "200G"
          cpu: "18"
          nvidia.com/gpu-h800: 1 # limit 1 GPU