Documentation › Virtual machines

Hyper-V

Creating the VM with the script, installing, checkpoints, the boot order and using the GPU for AI

Hyper-V is in Windows 10 and 11 Pro, Enterprise and Education. Hyper-V never gives a Linux VM your graphics card for its screen: the desktop is drawn by Hyper-V's virtual display, in software. Your GPU can still run Aurora AI models, from Windows.

#Turn on Hyper-V (once)

Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Hyper-V -All

In PowerShell as administrator, then restart Windows.

#Create the VM with the script

tools/hyperv/New-AuroraVM.ps1 from Aurora's repository creates everything from the ISO's path. In PowerShell as administrator:

powershell -ExecutionPolicy Bypass -File .\New-AuroraVM.ps1 -IsoPath "$env:USERPROFILE\Desktop\aurora-os-VERSION-amd64.iso"

It creates “Aurora” (Generation 2, 8 GB, 4 processors, 60 GB disk in C:\VMs), turns Secure Boot off, sets the boot order (DVD, disk, network), keeps Windows' standard and automatic checkpoints, sets the resolution and enhanced session, starts the VM and opens its window. Run it again with a newer ISO: it keeps the VM and its disk and only swaps the ISO. Options: -Name, -MemoryGB, -Processors, -DiskGB, -Folder, -Switch, -Width, -Height, -NoStart, -Recreate.

#Install

Install as on a PC (Installing Aurora OS). After the restart the VM starts from its disk; remove the ISO from the DVD drive in the VM's settings when you no longer need it.

#If the installed system doesn't start

It shows “Start PXE over IPv4”

The network is before the disk in the boot order:

Set-VMFirmware -VMName Aurora -BootOrder (Get-VMDvdDrive -VMName Aurora), (Get-VMHardDiskDrive -VMName Aurora), (Get-VMNetworkAdapter -VMName Aurora)

The firmware can't read the disk

On some hosts the VM's firmware reads the virtual disk only through a checkpoint. Keep checkpoints on:

Set-VM -Name Aurora -CheckpointType Standard -AutomaticCheckpointsEnabled $true

A report for help

tools/hyperv/Get-AuroraBootReport.ps1 -Name Aurora collects the VM's settings, Hyper-V's log and the boot files (read-only) into a folder on the Desktop.

#Use your GPU for Aurora AI

Run Ollama or LM Studio on Windows (they use your GPU), allow connections from the network, and in Aurora choose Settings → AI → Provider → OpenAI-compatible API with Windows' address (for Ollama, http://WINDOWS-IP:11434/v1). The full guide, including experimental GPU partitioning, is HYPERV.md in Aurora's repository.