Managed Kubernetes GPU Nodes
UpCloud Managed Kubernetes supports GPU node groups to accelerate AI, machine learning, and compute-intensive workloads. GPU nodes come with NVIDIA drivers pre-installed and GPU resources exposed to the container runtime:
- NVIDIA Driver: Version 595 (nvidia-open)
- CUDA: Version 13.2
- NVIDIA Container Toolkit: Pre-configured for container runtime integration
Getting started
GPU node groups can be added to your cluster through the API, Terraform, OpenTofu or upctl. GPUs are automatically available to all containers running on GPU nodes. To use Kubernetes resource limits (nvidia.com/gpu) for GPU scheduling, the NVIDIA device plugin must be installed separately.
See the Getting Started with GPU Workloads in Managed Kubernetes guide for setup instructions.