GPU Virtual Machines

On this page, you can find an explanation of how to create GPU Virtual Machine, and instructions for other steps to manage Virtual Machines in Cloud Console.

Table of contents

• Table of contents
  • Introduction
  • Create GPU Virtual Machine
  • Post-Creation Management

Introduction

GPU virtual machines (VMs) offer significant advantages for tasks that require intensive graphics processing or parallel computations.

Currently, our GPU services are exclusively available in the Upper-Austria region. We offer a range of GPU cards, including the H100 and L40S series, each available in multiple configurations to suit various performance needs:

• GPU-H100-1: 1 H100, 12 vCPUs, 250 GiB memory;
• GPU-H100-2: 2 H100, 24 vCPUs, 500 GiB memory;
• GPU-H100-4: 4 H100, 48 vCPUs, 1000 GiB memory;
• GPU-L40S-1: 1 L40S, 12 vCPUs, 224 GiB memory;
• GPU-L40S-2: 2 L40S, 24 vCPUs, 448 GiB memory;
• GPU-L40S-4: 4 L40S, 48 vCPUs, 960 GiB memory.

ℹ️ By default, these flavors are provided, but if you need a custom flavor, please contact our support team for assistance.

Choosing the appropriate GPU model allows you to enhance your applications for improved performance and efficiency within our cloud environment.

Create GPU Virtual Machine

Creating a GPU VM involves a similar process to setting up regular Ubuntu or Windows VMs, but with a few additional steps specific to GPU Flavor configurations.

GPU VM Creation Process:

1. Go to the Virtual Machines page and click on the CREATE VM icon in the upper left corner;
2. Fill in the form on the next opened Create Virtual Machine window:

• Name - specify a name for the virtual machine.

• Source - choose the source for creating the VM: image, volume, or snapshot;
  by default, “image” is pre-selected.

• OS Platform - choose the operating system platform: Linux or Windows;
  in this example we will create the Linux VM.

• Linux Distribution - if Linux OS is selected, choose the desired distribution (e.g. Ubuntu, CentOS, etc.).

• Image Version/Volume name/Snapshot name - based on the selected source, specify the corresponding image version, volume name, or snapshot name.

• Choose Flavor Type - select GPU Flavor Type if you need a VM with GPU capabilities.

• Choose Flavor - pick the type of GPU card and size for your new VM;
  ℹ️ available GPU flavors:

  ℹ️ If the list of GPU flavors is empty, your project is not currently associated with any GPU flavors.
  🔔 Please contact support to request access:

• Key pair - required for Linux VMs; choose an existing SSH key (created on the SSH Keys page) or create a new one to connect to the VM via SSH.

• Networks / Ports — choose how the VM will connect to the network:

  • Networks — select one or more available networks. A new port will be created automatically for each selected network;
    by default, “public” network is pre-selected;
    optionally specify Fixed IPs within the subnet when using the network option.
  • Ports — select an existing port to attach to the VM; the assigned firewalls will be displayed below (read-only).

• Firewalls - define which network rules control VM traffic:

  • using Networks, you can select Firewalls manually;
    by default, “default” Firewall is pre-selected;
  • using Ports, Firewalls will be inherited from the selected port.

  💡 The default Firewall allows outgoing Internet access but restricts incoming connections, except from resources in the same Firewall group.

• Tags - optional; use this field to assign tags to the VM.

• Volume size (GB) - provide the preferred disk size for the VM;
  valid range: 10–1000 GiB;
  minimal available size for Linux VMs - 10 GB; for Windows VMs - 50 GB;
  by default, “50 GB” is pre-selected.

• Delete Volume after VM deletion – enable this option to automatically delete the volume when the VM is removed. 

3. Complete VM Creation: Click on the CREATE icon to finalize the creation of your GPU VM.

After these steps, the newly created GPU VM will be added to the Virtual Machine page with the status ACTIVE.

Post-Creation Management

The management actions for GPU VMs, such as Editing, Resizing, Associating/Disassociating a Floating IP, or Deleting, are identical to those for non-GPU VMs.

💡 For additional details on these processes, refer to the Virtual Machines article.