Skip to content

Dedicated GPU

Dedicated GPU Service provides GPU acceleration for cloud virtual machines through a request-based provisioning workflow.

Unlike shared GPU resources, each approved GPU is exclusively allocated to the VM.

Introduction

Use Dedicated GPU Service when a VM requires GPU acceleration for AI, machine learning, visualization, or high-performance computing workloads.

GPU requests are reviewed and provisioned by the SITE Cloud team.

Key Features

Feature Description
Request-based provisioning GPU requests are reviewed and provisioned by the SITE Cloud team.
Dedicated allocation After approval, GPUs are dedicated to the VM.
Multi-GPU support Supports up to 8 GPUs per VM.

Common Use Cases

AI model training

  • Large language model training.
  • Distributed training across multiple GPUs.
  • Computer vision workloads.
  • Deep learning research.

Professional visualization

  • 3D rendering.
  • CAD and CAE applications.
  • Media production.
  • Scientific visualization.

High-performance computing

  • Cryptographic operations.
  • Scientific computing.
  • Simulation workloads.
  • Financial modeling.

Getting Started

Request GPUs for a VM

  1. Navigate to the Virtual Machines page.
  2. Select the VM that needs GPUs.
  3. Open the GPU tab.
  4. Click Request.
  5. Select the GPU type and number of GPUs.
  6. Click Submit.

The request is sent through Cloud Support portal for GPU attachment.

Use the configured GPUs

After the GPU is allocated and the status is Active in Cloud Portal:

  1. Install the required GPU drivers from the provider's download center.
  2. Verify that the GPUs are detected (In case of using Nvidia GPU): 
    nvidia-smi

Remove a GPU from a VM

  1. Navigate to the Virtual Machines page.
  2. Select the VM that has the GPU.
  3. Open the GPU tab.
  4. Open the actions menu for the GPU.
  5. Select Remove GPU.

The request is sent through Cloud Support portal for GPU removal.

Operating Guidelines

Plan Dedicated GPU requests with the following service rules:

  • GPUs can be requested for VMs that do not have Backup or Replication regimes enabled.
  • A VM can have up to 8 GPUs.
  • A VM can use one GPU type at a time.
  • Backups, Snapshots, and Replication are unavailable after GPUs are attached.
  • GPU pricing starts after Cloud Operations allocates and activates the GPU.
  • GPU pricing stops after Cloud Operations confirms GPU removal.
  • Remove GPUs before deleting a VM with attached GPUs.