Skip to main content

Getting Started with NovaPod

This guide walks you through launching, connecting to, and managing your first GPU instance on NovaPod. Within two minutes, you will have a live JupyterLab notebook, an SSH terminal, or an OpenAI-compatible REST endpoint running on enterprise NVIDIA or AMD accelerators.

1. Create an Account & Add Credits​

NovaPod operates on a non-custodial pay-as-you-go billing model:

  1. Visit Sign Up and create an account using your email or WebAuthn Passkey. No credit card is required.
  2. Go to Add Credits.
  3. Select your preferred cryptocurrency deposit rail:
    • USDT (TRON TRC-20, Ethereum ERC-20, BNB Smart Chain BEP-20)
    • USDC (Solana, Base, Ethereum)
    • Native Assets (SOL, ETH, BNB)
  4. Send funds to your dedicated, personal deposit address. Once confirmed on-chain (typically 15 to 45 seconds), your balance will reflect automatically at $1.00 USD credit per stablecoin unit.
Tiered Deposit Bonuses

Transfers of $200+ receive an automatic +3% bonus, $500+ receive +5%, $1,000+ receive +7%, and $2,000+ receive +15% credited directly to your balance.

2. Launching Your First GPU Pod​

Option A: Via Web Console​

  1. Navigate to the Marketplace or Templates.
  2. Filter the catalog by your required GPU model:
    • NVIDIA RTX 4090 / 5090: Best for fine-tuning, image generation, and quick prototyping ($0.20 – $0.55/hr).
    • NVIDIA H100 SXM5 / H200 SXM5: Best for 70B+ LLM inference, large-scale training, and NVLink inter-node scaling ($1.70 – $4.00/hr).
    • AMD Instinct MI300X: Best for 192GB VRAM single-card inference with ROCm 6.2 ($2.50/hr).
  3. Select a curated template:
    • PyTorch 2.4: Includes CUDA 12.4, cuDNN 9, FlashAttention-2, and JupyterLab.
    • vLLM Inference Server: OpenAI-compatible REST server with continuous batching.
    • ComfyUI Flux.1: Node-based generation studio with custom nodes pre-cached.
  4. Set your NVMe storage disk allocation (default 80 GB, expandable up to 2 TB).
  5. Click Launch Pod.

Option B: Via NovaPod CLI​

# Authenticate
np setup --api-key np_live_your_key_here

# Search the live ask orderbook
np search --gpu RTX_4090 --count 1

# Launch instance
np launch --gpu RTX_4090 --template pytorch --disk 80GB --name my-first-pod

3. Connecting to Your Instance​

Once the container transitions to Running status (under 15 seconds):

JupyterLab Web UI​

Click the JupyterLab button in your dashboard, or copy the direct tokenized URL:

https://pod-8842.node.novapod.io/?token=f9a8b1...

You can immediately start creating .ipynb notebooks, opening bash terminals, or uploading datasets.

Direct SSH​

Copy the SSH command displayed in your instance details card:

ssh -p 22042 root@node-42.novapod.io

If you uploaded an SSH public key in Console Settings, authentication is passwordless and instant.

4. Managing Compute & Halting Costs​

When you are done with your computational session:

  • Pause Instance:
    • Shuts down the GPU to stop compute charges immediately.
    • Keeps all installed packages, Python virtual environments, and downloaded checkpoints under /workspace intact.
    • You only pay a standby storage fee of $0.00014 per GB per hour (~$0.10/month for 100 GB).
  • Resume Instance:
    • Re-allocates an identical GPU and brings your exact environment back live in seconds.
  • Destroy Instance:
    • Completely wipes ephemeral disks and stops all billing.