> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.ibee.co.in/docs/infrastructure/gpu-vms/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.ibee.co.in/_mcp/server.

# GPU VMs

> GPU-backed virtual machines on IBEE Solutions for AI inference, training, rendering, and other accelerated workloads — deployed and managed like Cloud VMs, with per-GPU metrics.

GPU VMs are KVM-based virtual machines with NVIDIA GPUs attached. Use them for model training and fine-tuning, inference APIs, rendering, video processing, and any workload that needs GPU acceleration. You deploy and manage them from the portal much like [Cloud VMs](/docs/infrastructure/cloud-vms), with a GPU plan catalog and an extra **GPU Metrics** tab.

> **Info**
>
> **Ready to deploy?** Follow [Create a GPU VM](/docs/infrastructure/gpu-vms/create-a-gpu-vm).

## What you can do

* Deploy GPU plans that combine a GPU model and GPU memory with vCPU, RAM, and disk — up to 5 identical VMs at once.
* Boot from GPU-compatible OS templates, or from an ISO, snapshot, or backup.
* Sign in with SSH keys (Linux) or a password, shown under **Overview → Connection Details**.
* Connect with a public IP, or place the VM in a VPC with **Private + NAT** or **Private only** networking. Attach Reserved IPs and protect it with a firewall group.
* Watch CPU, memory, disk, and network on the **Monitoring** tab, and GPU utilization, VRAM, temperature, and power on the **GPU Metrics** tab.
* Protect data with scheduled [backups](/docs/tools/backups) and manual [snapshots](/docs/tools/snapshots).
* Change plans and manage IPs, VPCs, SSH keys, and passwords from the **Settings** tab.

## Find GPU VMs in the portal

In the portal sidebar, open **Infrastructure → GPU VMs**. The **GPU Virtual Machines** page lists your GPU VMs with their **Instance**, **OS**, **Location**, **IP Address**, and **Status**. Click **Deploy GPU VM** to create one, or click a row to open its detail page.

## How GPU VMs differ from Cloud VMs

|                 | GPU VMs                                                                        | Cloud VMs                                                                                                  |
| --------------- | ------------------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------------------- |
| **Plans**       | Each plan lists the GPU model and GPU memory, plus vCPU, RAM, disk, and price. | vCPU, memory, storage, and bandwidth. See [Instance types](/docs/infrastructure/cloud-vms/instance-types). |
| **Billing**     | **Hourly**.                                                                    | **Hourly**, or **Monthly** / **Yearly** commitments where offered.                                         |
| **Templates**   | Only GPU-compatible images are listed.                                         | Only Cloud VM images are listed.                                                                           |
| **Detail tabs** | Overview, Monitoring, **GPU Metrics**, Backups, Snapshots, Settings.           | Overview, Monitoring, Backups, Snapshots, Settings.                                                        |

Everything else — location, SSH keys, network configuration, instance count, hostnames, power actions, and the delete dialog — works the same way as for Cloud VMs.

## Lifecycle of a GPU VM

| Stage       | What you do                                                                            | Where                                                                   |
| ----------- | -------------------------------------------------------------------------------------- | ----------------------------------------------------------------------- |
| **Deploy**  | Pick location, GPU plan, OS, SSH keys, network, and hostname.                          | [Create a GPU VM](/docs/infrastructure/gpu-vms/create-a-gpu-vm)         |
| **Set up**  | Confirm the GPU with `nvidia-smi`; install a compatible NVIDIA driver if it's missing. | [Create a GPU VM](/docs/infrastructure/gpu-vms/create-a-gpu-vm#verify)  |
| **Operate** | Start, stop, reboot; watch **Monitoring** and **GPU Metrics**.                         | [Power actions](/docs/infrastructure/cloud-vms/power-actions)           |
| **Connect** | Manage public IP, Reserved IPs, VPCs, and access.                                      | [Networking for VMs](/docs/infrastructure/cloud-vms/networking-for-vms) |
| **Protect** | Enable scheduled backups or take manual snapshots.                                     | [Backups](/docs/tools/backups) · [Snapshots](/docs/tools/snapshots)     |
| **Scale**   | Change plan from **Settings → Change Plan**.                                           | [Resize a VM](/docs/infrastructure/cloud-vms/resize-a-vm)               |
| **Retire**  | Delete the VM, optionally keeping its public IP as a Reserved IP.                      | [Power actions](/docs/infrastructure/cloud-vms/power-actions#delete)    |

## GPU Metrics

The **GPU Metrics** tab charts **GPU Utilization**, **VRAM Utilization**, **Temperature**, and **Power Draw** over **30m**, **1h**, **6h**, **24h**, or **7d**. Collection is off until you turn it on:

1. Start the VM, and make sure `qemu-guest-agent` is running inside it.
2. On **GPU Metrics**, click **Enable monitoring**. A lightweight IBEE agent is installed through the guest agent — it opens no network port on the VM.
3. If the tab shows **NVIDIA guest driver required**, install a compatible NVIDIA driver inside the VM. Monitoring starts automatically once `nvidia-smi` works; CUDA is optional.

If the agent goes offline or setup fails, the tab explains why and offers **Repair monitoring**. Historical values stay visible while the VM is stopped.

> **Info**
>
> Browser console access isn't available in the portal yet — the **Console** action is shown as **Soon**. Use SSH or RDP with the details under **Overview → Connection Details**.

## Related pages

* [Create a GPU VM](/docs/infrastructure/gpu-vms/create-a-gpu-vm)
* [Cloud VMs](/docs/infrastructure/cloud-vms)
* [Networking for VMs](/docs/infrastructure/cloud-vms/networking-for-vms)
* [SSH Keys](/docs/tools/ssh-keys)

> GPU-backed virtual machines on IBEE Solutions for AI inference, training, rendering, and other accelerated workloads — deployed and managed like Cloud VMs, with per-GPU metrics.

## Docs

- [Create a GPU VM](https://docs.ibee.co.in/docs/infrastructure/gpu-vms/create-a-gpu-vm.md): Deploy a GPU-enabled virtual machine on IBEE Solutions for AI, ML, rendering, and compute-intensive workloads.