Home > Storage > PowerFlex > White Papers > Dell Validated Design for Virtual GPU with VMware and NVIDIA on PowerFlex > VM configuration
The following VM configuration is required to use a vGPU in the VM:
1. Create a VM on an ESXi host with the Ubuntu 22.04 operating system.
2. Edit the following settings for the VM:
a. Set Boot Options under VM Options to EFI mode.
b. Add the following advanced configurations:
pciPassthru.use64bitMMIO: TRUE
pciPassthru.allowP2P: TRUE
pciPassthru.64bitMMIOSizeGB: 64
c. Add a PCI device to the VM and set the NVIDIA GRID vGPU Profile. For example. grid_a100-7-40c.
Figure 8. Setting the NVIDIA profile
3. Boot the VM, install the operating system, and set the IP address for this VM.
4. Install the NVIDIA vGPU Linux Driver package by downloading it from the NVIDIA Licensing Portal and placing a copy on the VM. For example, /tmp/NVIDIA-Linux-x86_64-510.85.02-grid.run
Note: After installing the NVIDIA vGPU Linux Driver package, the console window in vSphere is no longer able to display the GUI. To control the VM, you need to use the SSH to log in to the VM.
5. Before running the driver installer, exit the X Server and terminate all OpenGL applications.
a. Use CTRL+ALT+F1 to switch to a console prompt.
b. Shut down the display manger with the following command:
sudo service gdm stop
6. Install the driver by running the following commands:
chmod +x NVIDIA-Linux-x86_64-510.85.02-grid.run
sh ./chmod +x NVIDIA-Linux-x86_64-510.85.02-grid.run
7. After the driver is installed, reboot the VM and verify the driver installation with nvidia-smi and check for a similar output to the ESXi host with minor changes.
For more information about installing the NVIDIA vGPU software graphics driver, see Installing the NVIDIA vGPU Software Graphics Driver on Linux in the NVIDIA Virtual GPU Software User Guide.
8. Reboot the VM to complete the driver installation.
9. Copy the client token generated above to /etc/nvidia/ClientConfigToken/ of every VM created.
10. License the GPU driver by following this procedure: https://docs.nvidia.com/grid/ls/latest/grid-license-server-user-guide/index.html#installing-nvidia-grid-license-server-linux.
11. Restart the nvidia-gridd service after applying the license and then verify the licensing.
systemctl restart nvidia-gridd
cat /var/log/syslog | grep Lic
nvidia-smi -q | grep License
The installation of NVIDIA vGPU drivers is now complete and the VM is ready to run AI workloads.