Optimizing VM Configuration for Performant AI Inference on Blackwell Ultra HGX

NVIDIA B300 NVL8 full-node passthrough with KVM and libvirt.

• Introduction
  • About this paper
  • Scope
  • Problem Statement
  • Target Audience
  • Environment
• System Setup and Installation
  • Blackwell Ultra System Configuration
  • B300 NVL8 Physical Topology
  • Blackwell Ultra System Software Configuration
  • Installing the Hypervisor
  • Understanding system topology
  • Sample ‘lstopo’ output
  • Additional System Setup Considerations
  • Optimal device selection
• Configuring Virtual Machines
  • VM Configuration Pre-requisites
  • Editing a VM’s XML definition
  • Scenario: Full passthrough- All GPUs, NICs, and NVSwitches
  • Virtual CPU configuration
  • Virtual PCI Express configuration
  • Passthrough device assignment
• Known Limitations
• Appendix
  • ATS and ACS configuration
  • Example physical topology map for B300 NVL8 8-GPU platform
  • Example libvirt domain XML for B300 NVL8 8-GPU platform
• Notices
  • Notice
  • Trademarks
  • Copyright