NVIDIA Jetson

Run PC2 on an NVIDIA Jetson device for GPU-accelerated private AI. Your Jetson becomes a sovereign AI node — running local language models with hardware acceleration, no cloud APIs needed.

Why Jetson? Jetson devices have dedicated NVIDIA GPUs that accelerate AI inference 3-10x faster than CPU-only. This means faster responses from your local AI agent, entirely private, entirely on your hardware.

Supported Devices

Model	RAM	GPU	AI Performance	Price
Jetson Nano (4GB)	4GB	128 CUDA cores	Entry-level	~$150
Jetson Orin Nano	8GB	1024 CUDA cores	Mid-range (recommended)	~$250
Jetson Orin NX	16GB	2048 CUDA cores	High-end	~$900
Jetson AGX Orin	64GB	2048 CUDA cores	Enterprise	~$2000

⚠️

Jetson Nano (4GB): Can run PC2 and small AI models (deepseek-r1:1.5b). For larger models, use an Orin device with 8GB+ RAM.

Step 1: Flash JetPack OS

Download JetPack

Go to developer.nvidia.com/jetpack and download the latest JetPack SD card image for your device.

Flash the Image

Use balenaEtcher or the NVIDIA SDK Manager to flash the image to your SD card or NVMe SSD.

First Boot

Connect keyboard, monitor, and ethernet. Power on and complete the Ubuntu setup wizard. Set your username and password.

Step 2: Remove Desktop (Recommended)

The GNOME desktop environment uses 1-2GB of RAM. Since PC2 provides its own web-based interface, you don’t need it. Removing it frees up RAM for AI models.

# Remove GNOME desktop to free RAM for AI
sudo systemctl set-default multi-user.target
sudo apt remove --purge -y ubuntu-desktop gnome-shell gdm3
sudo apt autoremove -y
sudo reboot

After reboot, your Jetson will boot to a terminal. SSH in from another computer:

ssh your-username@your-jetson-ip

Find your Jetson’s IP address by checking your router’s connected devices, or run hostname -I on the Jetson before removing the desktop.

Step 3: Install Node.js

# Update system
sudo apt update && sudo apt upgrade -y
 
# Install Node.js 20 LTS
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt install -y nodejs
 
# Install build tools (needed for native modules)
sudo apt install -y build-essential git libcairo2-dev libpango1.0-dev \
  libjpeg-dev libgif-dev librsvg2-dev
 
# Verify
node --version  # Should show v20.x.x
npm --version

Step 4: Install Ollama (Local AI)

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh
 
# Verify Ollama detects GPU
ollama --version

Ollama automatically detects CUDA on Jetson devices. No extra GPU configuration is needed if JetPack was installed correctly.

Download an AI Model

Choose a model based on your device’s RAM:

# For Jetson Nano (4GB) — lightweight model
ollama pull deepseek-r1:1.5b
 
# For Jetson Orin (8GB+) — more capable model
ollama pull llama3.2:3b
 
# For Jetson Orin NX/AGX (16GB+) — full-size model
ollama pull mistral:7b

Test it:

ollama run deepseek-r1:1.5b "Hello, what can you do?"

You should see the response generated using your Jetson’s GPU.

Step 5: Install PC2

Option A: One-Line Install (Recommended)

curl -fsSL https://raw.githubusercontent.com/Elacity/pc2.net/main/scripts/start-local.sh | bash

This installs PC2, builds it, and starts it with PM2.

Option B: Manual Install

# Install PM2 (process manager)
sudo npm install -g pm2
 
# Clone repository
git clone https://github.com/Elacity/pc2.net
cd pc2.net
 
# Install dependencies
npm install --legacy-peer-deps --ignore-scripts
 
# Build PC2
cd pc2-node
npm install --legacy-peer-deps
npm run build
 
# Start with PM2
pm2 start npm --name pc2 -- start
pm2 save
pm2 startup

Step 6: Access Your Cloud

Open a browser on any device on your network and go to:

http://your-jetson-ip:4200

You’ll see the PC2 setup wizard. Connect your wallet and pick a username.

Remote Access: Once connected, your Jetson automatically registers a username.ela.city domain through the Boson network. Anyone can reach your PC2 at https://username.ela.city — even though your Jetson is behind your home router.

Managing PC2

Common Commands

# Check status
pm2 status
 
# View logs
pm2 logs pc2
 
# Restart
pm2 restart pc2
 
# Stop
pm2 stop pc2
 
# Monitor (real-time dashboard)
pm2 monit

Update PC2

pm2 stop pc2
cd ~/pc2.net
git pull origin main
cd pc2-node
npm run build
pm2 restart pc2

Auto-Start on Boot

pm2 startup    # Follow the printed command
pm2 save       # Save current process list

Now PC2 starts automatically when your Jetson powers on.

Performance Optimization

Use NVMe SSD

SD cards are slow. For significantly better performance, use an NVMe SSD:

Install an M.2 NVMe SSD in your Jetson’s M.2 slot
Flash JetPack directly to the SSD, or move the root filesystem to SSD

Increase Swap

For running larger AI models, add swap space:

# Create 4GB swap file
sudo fallocate -l 4G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
 
# Make permanent
echo '/swapfile none swap sw 0 0' | sudo tee -a /etc/fstab

Set Max Performance Mode

Jetson devices have power modes. For maximum AI performance:

# Set to maximum performance (Orin)
sudo nvpmodel -m 0
 
# Set max clock speeds
sudo jetson_clocks

⚠️

Maximum performance mode uses more power and generates more heat. Ensure adequate cooling.

Monitor GPU Usage

# Check GPU utilization
sudo tegrastats
 
# Or use jtop (Jetson's htop equivalent)
sudo pip3 install jetson-stats
sudo jtop

Troubleshooting

Ollama Not Using GPU

# Check if CUDA is available
nvcc --version
 
# If not found, install JetPack CUDA toolkit
sudo apt install -y nvidia-jetpack
 
# Restart Ollama
sudo systemctl restart ollama

Out of Memory

If PC2 or Ollama crashes with OOM errors:

Remove the desktop environment (Step 2)
Use a smaller AI model (deepseek-r1:1.5b instead of mistral:7b)
Increase swap space (see above)
Close unnecessary services: sudo systemctl disable bluetooth

Build Errors (node-pty)

If you get native module errors during npm install:

sudo apt install -y build-essential libcairo2-dev libpango1.0-dev \
  libjpeg-dev libgif-dev librsvg2-dev
npm rebuild

Can’t Access from Browser

Check PC2 is running: pm2 status
Check the port: curl http://localhost:4200/health
Check firewall: sudo ufw allow 4200
Use the Jetson’s IP address, not localhost, from another device

WASM Warning

You may see wasm streaming compile failed in the logs. This is harmless — it falls back to ArrayBuffer instantiation and works correctly.

Your hardware. Your AI. Your sovereignty.

Raspberry Pi First Login