Is this page helpful?

NVIDIA vGPU for Compute Installation#

Install NVIDIA vGPU for Compute to enable GPU virtualization. The procedure has three phases—host setup, guest VM driver installation, and AI workload deployment—each documented on its own page below.

Installation Overview

  1. Verify Prerequisites: Confirm hardware, BIOS settings, and licensing requirements

  2. Install NGC CLI: Download software from NVIDIA NGC Catalog

  3. Install Virtual GPU Manager: Deploy on hypervisor host (VMware, KVM, Nutanix)

  4. Verify Fabric Manager: Included in the NVIDIA AI Enterprise drivers for HGX multi-GPU configurations

  5. Install vGPU Guest Driver: Deploy in each virtual machine

  6. Configure Licensing: Connect VMs to NVIDIA License System

Refer to the NVIDIA AI Enterprise Product Support Matrix for supported platforms and versions.

Installation Tasks#

Table 44 Installation Subpages#

Subpage

Audience

Use when you need

Host Setup

Hypervisor administrator

To verify prerequisites and BIOS, install the NGC CLI, deploy the Virtual GPU Manager on the hypervisor, and confirm Fabric Manager on HGX servers.

Installing NVIDIA vGPU Guest Driver

Guest VM owner

To install the vGPU Guest Driver in each VM (Ubuntu, Red Hat, Windows, or other Linux), then license the VM and configure profiles.

AI Enterprise Workload Setup

AI / DS / workload deployer

To install the NVIDIA GPU Operator (Bash script) and pull NVIDIA AI Enterprise application containers via Docker, Podman, or Cloud Native Stack on licensed Guest VMs.

Next Steps#

After installing the vGPU Manager and guest drivers: