Create a GPU node

Creating a GPU node uses the same create-node flow as a regular node, with one difference in order: you pick the GPU card first, and the operating system, version, and everything else follows on the plan screen.

Pick a GPU card for the node

Open Compute > Nodes, click Add New Node, and switch from the CPU tab to the GPU tab. The GPU tab appears only in the Delhi region, so switch your region to Delhi first if it is set to Chennai. To ask for a card that is not listed, contact support.

Select a card (A100, L40S, A40, A30, L4, V100, or T4), and you land straight on Select Plan. See Choose a GPU card for what each one is built for.

Configure the operating system and plan

The GPU tab asks for the operating system and version on the plan screen, next to the plan itself, rather than before it the way the CPU tab does. Everything else works the same as a regular node:

• Plan shows vCPU count, RAM, GPU memory, CUDA version, disk space, and IOPS together for each option. The number of GPU cards is fixed per plan; a multi-card plan (2x, 4x, 8x) is its own entry in the list.
• Billing offers the same on-demand and committed options as any node; GPU plans belong to the GDC or GDC3 family.
• Volume is hidden for most GPU plans. Create the node first, then attach a volume from its details page afterward; see Attach a volume.
• Encryption, VPC, reserved IP, security group, and SSH keys all work exactly as they do for a regular node; see Create a node for the full reference on each.

Launch the node and check the card

Review the Summary panel, then click Launch. The node shows as Creating and then Running on your node list; a GPU node can take a little longer than a plain CPU node while the driver initializes on first boot.

Once it is Running, connect and confirm the card: see Connect to a Linux GPU node or Connect to a Windows GPU node. If the card does not show up once you are in, see Troubleshooting GPU Node.