CMPv2- Creating a GPU Virtual Machine

What You’ll Accomplish

This guide walks through every field of the CloudPe GPU VM creation wizard — GPU type, compute flavor, OS image, billing period, storage policy, network, security, and Cloud-Init — before clicking Create GPU VM.

Prerequisites

Ensure these are in place before you begin

🔑

Active Account

Verified account at app.cloudpe.com

💰

Billing / Credits

Sufficient balance or active billing method

📁

Project Created

At least one project in your account

🎖

GPU Quota

GPU quota approved — contact support if needed

🔐

SSH Key (Recommended)

Public key uploaded under Billing → SSH Keys

🌐

Network Available

Public or private network in your region

Navigate to GPU VMs

Access the GPU VM creation wizard from the sidebar

Open GPU VMs from the Sidebar

In the left sidebar, expand Compute and click GPU VMs. You’ll see the GPU VMs list page. Click the + Create GPU VM button at the top right, or go directly to app.cloudpe.com/dashboard/gpu-vms/create.

Select GPU Type

Choose the NVIDIA GPU accelerator for your workload

Review Available GPU Types

CloudPe offers 6 NVIDIA GPU types. Each card shows the GPU name, VRAM, and number of available flavors. Click a card to select it. The region is automatically assigned based on where that GPU hardware is available.

GPU Type Reference

Select Flavor

Choose the compute configuration — vCPU and RAM — for your GPU VM

Select a Flavor

Based on the GPU type you selected, available flavors are listed. Each flavor shows the vCPU count, RAM, attached region, and pricing. The GPU VRAM is the same across all flavors of the same GPU type — only the host vCPU and system RAM differ.

H200 Flavor Reference

Select Operating System

Choose a GPU-optimised image or your custom image

Choose from GPU Images or Custom Images

The Image step has two tabs. GPU Images contains CloudPe-managed images with NVIDIA drivers and CUDA pre-installed. Custom Images lists your own uploads and snapshots. Select Ubuntu-24.04-GPU-v2 from the GPU Images tab for new deployments.

Image Options

Select Billing Period

Choose how you want to be billed — longer periods offer better value

Select a Billing Period

Choose from Hourly, Monthly, Quarterly, Semi-Annual, or Annual billing. Committing to longer periods unlocks significant discounts (up to 20% off). Check the Auto-renew subscription box to renew automatically at period end.

Billing Period & Pricing (H200 · gpu.h200.384g)

Pricing Estimate — Monthly Bundle Breakdown

Configure Your GPU VM

Set project, name, storage, network, security, and advanced options

Select Project

Choose the project where this GPU VM will be deployed. Projects control billing grouping, quota, and team access. The region (e.g. Mumbai DC2 Zone B) is shown automatically based on the flavor you selected and cannot be changed here.

VM Name

Enter a unique name for your GPU VM. Use only letters, numbers, hyphens (-), and underscores (_). The name is permanent after creation and is used in API calls and the dashboard.

Boot Volume Size

Set the size of the root disk using the slider or by typing a value. The minimum is 30 GB for GPU images. For GPU workloads that store large model checkpoints or datasets on the boot disk, increase to 100–500 GB.

Storage Policy (NVMe Tier)

Additional Data Volumes (Optional)

Click + Add Volume to attach extra data volumes alongside the boot disk. Useful for storing large ML datasets, model weights, or checkpoints separately from the OS disk. Each additional volume can have its own size and NVMe tier.

Network Configuration

Security — Root Password & SSH Key

Advanced Options — User Data (Cloud-Init)

CloudPe pre-fills a GPU-ready Cloud-Init template that installs CUDA and Python 3 on first boot. You can extend it to auto-install your ML stack, download model weights, or start services automatically.

#cloud-config
packages:
  - nvidia-cuda-toolkit
  - python3-pip

#cloud-config
packages:
  - nvidia-cuda-toolkit
  - python3-pip

runcmd:
  - pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu124
  - pip3 install jupyter transformers datasets
  - jupyter notebook --ip=0.0.0.0 --port=8888 --no-browser &

Review & Create

Verify the Configuration Summary and launch your GPU VM

After Your GPU VM is Running

Quick-start actions after creation

👖

SSH into VM

ssh ubuntu@<public-ip> — default user is ubuntu. Your SSH key authenticates automatically.

💻

Verify GPU

Run nvidia-smi to confirm GPU detection, check VRAM, driver version, and CUDA version.

🐍

Install ML Stack

Python 3 and pip are pre-installed. Run pip3 install torch transformers jupyter or use Cloud-Init to auto-install on next boot.

📖

Run Jupyter

Start with jupyter notebook --ip=0.0.0.0 --port=8888 and open http://<ip>:8888 in your browser.

📷

Snapshot First

Before experiments, create a snapshot from VM Details → Snapshots. This saves your configured environment.

🛑

Stop vs Delete

Subscription billing may continue even when stopped. For cost savings, delete the VM and restore from snapshot when needed.

Frequently Asked Questions

Common questions about GPU VM creation on CloudPe