Skip to main content

GPU as a Service

Hikube provides access to NVIDIA accelerators in passthrough mode, for workloads that require hardware acceleration (AI/ML, rendering, HPC).

The console has no dedicated GPU page: you choose the GPU when you create the resource that uses it, in the Hikube console.

Usage modes​

ModeWhere to choose the GPUUse cases
GPU on VMVM Instances > Create an Instance, Configuration step, Hardware Acceleration (GPU) sectionNative CUDA, interactive environments, rendering
GPU on KubernetesKubernetes > Create cluster, Nodes step, GPU section of a node groupContainerized training and inference, batch

Guides​

GPU on VM

Create a VM with one or more NVIDIA GPUs from the console wizard.

  • VM
  • Passthrough

GPU on Kubernetes

Add a GPU node group to a managed Kubernetes cluster, with the GPU Operator.

  • Kubernetes
  • GPU Operator

Overview

Available GPU models, availability, and choosing between VM and Kubernetes.