Set Up a Cloud GPU VM
You can set up a Cloud GPU VM in the DCD using Canvas mode or using the Xpress mode. This topic covers adding the GPU element to your data center, configuring its storage and network connectivity, and provisioning your changes to deploy it in your infrastructure.
Important:
Cloud GPU VMs can only be added to data centers in the Germany / Frankfurt am Main (de/fra/2) location.
The procedure listed here applies only to the Canvas mode. For configuring the Cloud GPU VM using the Xpress mode, see Use the Xpress Mode.
If you do not have a data center, create a data center using either of these two modes: Canvas or Xpress.
After creating a data center using the Canvas mode, continue with the following steps to add and provision a Cloud GPU VM within it.
For instructions on creating and provisioning a Cloud GPU VM using the Xpress mode, see Use the Xpress Mode.
Create a Cloud GPU VM
Get started
1. In the DCD, go to Menu > Virtual Data Centers.
2. Select the data center in the Germany / Frankfurt am Main (de/fra/2) location to configure a Cloud GPU VM within it.
3. Drag the Cloud GPU element from the Palette into the Workspace.
4. Select the Cloud GPU element.

5. Use the Inspector pane on the right to configure the properties for this individual instance. In the Inspector pane, configure your Cloud GPU VM Settings:
Name: Specify a name for your Cloud GPU VM.
CPU Architecture: By default, Cloud GPU VM supports AMD EPYC Turin.
Template: Choose the appropriate configuration template. The following are automatically set when you choose a template: GPUs, CPU Cores, RAM in GB, and Storage in GB.
6. The GPU element includes an Unnamed SSD Storage by default. Select it to configure the following in the Inspector pane:

Name: Specify a storage name unique to the chosen Virtual Data Center (VDC).
Availability Zone: The zone is set to Auto, by default.
Size in GB: The storage size defaults to the value defined in the template.
Performance: By default, Cloud GPU VMs use the Premium performance class to ensure maximum performance.
Image: Select the drop-down list to choose from and associate an IONOS CLOUD from Linux Images.
Password: Enter a password between 8 and 50 characters using letters and numbers only.
SSH Keys: Select premade SSH Key. You must first have a key stored in the SSH Key Manager. For more information, see Connect to Virtual Machines via SSH.
Ad-hoc SSH Key: Copy and paste the public portion of your SSH key into this field.
Cloud-Init user data: Provide custom initialization scripts or configuration data to be executed when the instance first starts. Use this to automate setup tasks such as installing packages, configuring services, or setting environment variables. For more information, see Boot with Cloud-init.
Backup Unit: Select a backup unit from the drop-down list or create a new one to back up all data with version history to local storage or your private cloud storage.
Connect a GPU to the Internet
1. Each compute instance has a NIC, which is activated through the Autoport symbol. Connect the GPU to the Internet by dragging a line from the GPU's Autoport to the internet's NIC.

2. In the Inspector pane, select the Network tab to configure your LAN device.
Name: Enter a name for your NIC in the Virtual Data Center (VDC).
MAC: The MAC address is assigned automatically on provisioning.
LAN: Select a LAN. For more information, see LAN.
Firewall: Configure a Firewall.
Configure IPv4 addresses
Primary IPv4: The assignment of the primary IP address depends on the network type:
Public LANs: The primary IP address is automatically assigned by the IONOS CLOUD DHCP server. To use a specific address, select a reserved IP address from the drop-down list. The NIC must be connected to the Internet for public routing.
Private LANs: The primary IP address is automatically assigned by the IONOS CLOUD DHCP server. If you choose to specify a private IP manually, ensure it falls within the same CIDR range as the existing LAN. Note that the IONOS CLOUD DHCP server does not track manually assigned addresses, so you must manage these static IP addresses carefully to avoid address conflicts.
DHCP: It is often necessary to run a DHCP server in your virtual data center (for example, PXE boot for fast rollout of VMs). If you use your own DHCP server, clear this checkbox so that your IPs are not reassigned by the IONOS CLOUD DHCP server.
Result: The GPU is connected to the internet and assigned an IP address.
Provision changes
1. Start the provisioning process by selecting PROVISION CHANGES in the Inspector pane.

2. The Provision Data Center dialog opens. Review your changes in the Validation tab.
3. Confirm changes with your password. Resolve outstanding errors without a password.
4. Once ready, select Provision Now to start provisioning resources.

Result: The new Cloud GPU VM now appears in the data center. The DCD displays a Provisioning Complete notification when the infrastructure is ready.
Last updated
Was this helpful?