Configure GPU workloads
Configure PCI passthrough on the control plane and data plane nodes to provision instances that support GPU-accelerated AI and machine learning workloads.
GPU workloads require direct access to your hardware for optimal performance. If you set up your RHOSO environment to support GPUs, you can run demanding applications such as the Red Hat AI inference server with the processing power needed for business operations.
To complete this configuration, you set up PCI passthrough on both your control plane and data plane nodes. You can then launch a GPU-ready instance to test and verify the setup.
- Enable GPU support for instances
Configure PCI passthrough on control plane and data plane nodes to enable GPU access for cloud instances. - Verify GPU configuration with an AI inference server
Deploy a Red Hat AI inference server instance to verify that your GPU passthrough configuration works correctly.