Olares

English

简体中文

English

简体中文

Appearance

Install Olares on PVE via ISO with GPU Passthrough ​

GPU passthrough in Proxmox Virtual Environment (PVE) allows virtual machines (VMs) to directly access the physical GPU, enabling hardware-accelerated computing for workloads like AI model inference and graphics processing.

This tutorial provides a comprehensive process for:

  • Configuring a PVE host for GPU passthrough.
  • Installing Olares from its official ISO image into a new VM that fully leverage the dedicated GPU.

Not recommended for production use

Currently, Olares on PVE has certain limitations. We recommend using it only for development or testing purposes.

Prerequisites ​

Before proceeding, ensure that your setup meets the following requirements:

  • CPU: At least 4 cores, with IOMMU enabled in BIOS
    • Intel: VT-d
    • AMD: AMD-Vi/IOMMU
  • GPU: NVIDIA GPU that supports GPU passthrough
  • RAM: Recommended 16 GB or more
  • Storage: Minimum 200 GB SSD (installation may fail on HDD)
  • PVE Version: 8.3.2
  • Olares ISO Image: Download the official Olares ISO image before you start.

Configure GPU passthrough in PVE ​

To use GPU-accelerated workloads in Olares, you must first enable GPU passthrough for the PVE host.

Enable IOMMU ​

The Input-Output Memory Management Unit (IOMMU) is a hardware feature that allows the operating system to control how devices access memory, which is essential for passthrough.

  1. In the PVE command line (Shell), run the following command to open the GRUB configuration file:

    bash
    nano /etc/default/grub

  2. Find the line: GRUB_CMDLINE_LINUX_DEFAULT="quiet"

    Replace it with the line corresponding to your CPU vendor:

    bash
    GRUB_CMDLINE_LINUX_DEFAULT="quiet intel_iommu=on iommu=pt"
    bash
    GRUB_CMDLINE_LINUX_DEFAULT="quiet amd_iommu=on iommu=pt"

  3. Save and close the GRUB configuration file, then apply GRUB changes:

    bash
    update-grub
reboot

  4. After restarting, check whether IOMMU is enabled on the PVE host:

    bash
    dmesg | grep -e DMAR -e IOMMU

    If successful, you should see output similar to the following:

    bash
    [0.061644] DMAR: IOMMU enabled
...
[0.408103] DMAR: Intel(R) Virtualization Technology for Directed I/O
    bash
    [1.219719] AMD-Vi: Found IOMMU at 0000:00:00.2 cap 0x40

Add VFIO modules ​

Virtual Function I/O (VFIO) enables a VM to directly access PCI devices such as GPUs.

  1. On the PVE host, run the following command to open the modules file:

    bash
    nano /etc/modules

  2. Add these lines to the end of the file:

    vfio
vfio_iommu_type1
vfio_pci
vfio_virqfd

  3. Save and close the file.

Blacklist host GPU drivers ​

To prevent the Proxmox host from using the GPU you plan to pass through, it's best to blacklist its default drivers. This ensures the GPU is available for vfio-pci.

  1. Run the following command on the PVE host to create the blacklist configuration:

    bash
    nano /etc/modprobe.d/blacklist.conf

  2. Add the following lines to block NVIDIA drivers:

    blacklist nouveau 
blacklist nvidia 
blacklist nvidiafb
blacklist nvidia_drm
blacklist nvidia_modeset

  3. Save and close the file.

Bind GPU to VFIO ​

  1. Run the following command on the PVE host to find your GPU's PCI address:

    bash
    lspci | grep NVIDIA

    Example output:

    01:00.0 VGA compatible controller: NVIDIA Corporation AD106 [GeForce RTX 4060 Ti] (rev a1)
01:00.1 Audio device: NVIDIA Corporation AD106M High Definition Audio Controller (rev a1)

    Here, the GPU's PCI address is 01:00, with two functions listed.

  2. Get the PCI idenfiers of your GPU:

    bash
    lspci -n -s 01:00

    Example output:

    01:00.0 0300: 10de:2803 (rev a1)
01:00.1 0403: 10de:22bd (rev a1)

    In this case, the IDs of the GPU are 10de:2803 and 10de:22bd.

  3. Bind the IDs to VFIO (replace the IDs with your own):

    bash
    echo "options vfio-pci ids=10de:2803,10de:22bd" > /etc/modprobe.d/vfio.conf

  4. Apply all module and driver changes by updating the initramfs (the root file system), then reboot:

    bash
    update-initramfs -u
reboot

  5. After rebooting, check if the GPU is now using the vfio-pci driver:

    bash
    lspci -v

    You should see the output similar to:

    01:00.0 VGA compatible controller: NVIDIA Corporation AD106 [GeForce RTX 4060 Ti] (rev a1) (prog-if 00 [VGA controller])    
Subsystem: Gigabyte Technology Co., Ltd AD106 [GeForce RTX 4060 Ti]
Flags: fast devsel, IRQ 255, IOMMU group 11
...
Kernel driver in use: vfio-pci

Set up VM and install Olares ​

With the GPU passthrough enabled, you can now install Olares in PVE.

Create and configure the VM ​

This section creates and configures a VM using the Olares ISO image:

  1. Upload the official Olares ISO you downloaded to your PVE storage (e.g., local).

    1. In the PVE web interface, select your target storage (e.g., local).

    2. Click ISO Images > Upload.

    3. Click Select File, choose the Olares ISO file you downloaded, and then click Upload.

  2. Click Create VM.

  3. Configure the VM settings as follows:

    • OS:
      • ISO image: Select the official Olares ISO image you just downloaded.
    • System:
      • BIOS: Select OVMF (UEFI).
      • EFI Storage: Choose a storage location (for example, a local LVM or directory) to store UEFI firmware variables.
      • Pre-Enroll keys: Uncheck the option to disable Secure Boot.
    • Disks:
      • Disk size (GiB): At least 200GB.
    • CPU:
      • Cores: At least 4 cores
    • Memory:
      • Memory (MiB): At least 8GB

  4. Click Finish. Do not start the VM yet.

    Below is a sample configuration for the VM hardware settings in PVE. PVE Hardware

Bind GPU to the VM ​

Follow these steps to bind the GPU to your VM:

  1. In the PVE interface, select your VM and go to Hardware > Add > PCI Device. Add PCI

  2. Select Raw Device, then pick your GPU by the PCI address (for example, 01:00).

  3. In the bottom-right corner, select the Advanced options and check PCI-Express.

  4. Click Add to save. Add GPU

Now your VM is ready to use GPU passthrough.

Install Olares ​

Once the VM is set up, follow these steps to install Olares using the installer ISO image.

  1. Select and start the VM you just created.

  2. From the boot menu, select Install Olares to Hard Disk and press Enter.

  3. In the Olares System Installer interface, select the installation disk.

    1. Review the list of available disks (for example, sda 200G QEMU HARDDISK).

    2. Select the first disk by typing /dev/ plus its name (for example, /dev/sda).

    3. When the on-screen warning appears, just type yes to continue.

    Note

    During installation, warnings related to the NVIDIA graphics driver will appear. Press Enter to ignore them.

  4. Once the installation completes, you'll see the message:

    Installation completed successfully!

  5. Reboot the VM in the Proxmox web interface.

Verify the installation and GPU passthrough ​

After the VM restarts, it will boot into the Ubuntu system.

  1. Log in to Ubuntu using the default credentials:
  • Username: olares
  • Password: olares

  1. Confirm that Olares has been installed successfully using the following command:

    bash
    sudo olares-check

    The installation is successful if you see results like:

    ...
check Olaresd:  success
check Containerd:  success

  2. Finally, verify that the GPU is successfully passed through and recognized by Olares using the NVIDIA System Management Interface tool:

    bash
    nvidia-smi

    If successful, this command will display a table with your NVIDIA GPU's details, including its name, driver version, and memory usage.

Next steps ​

Olares is now installed and running with full GPU acceleration. To start using Olares, activate the device and log in with your account. For detailed steps, see our official guides: