GPU Hot

Real-Time NVIDIA GPU Monitoring Dashboard

Single-container web dashboard for NVIDIA GPU monitoring with real-time charts.

Overview

Self-contained dashboard for monitoring NVIDIA GPUs on remote servers. Access utilization and health metrics from a browser without SSH.

Runs in a single container on one port. No configuration required - start the container and open a browser.

---

Quick Start

Using Pre-built Docker Image (Recommended)

docker run -d --name gpu-hot --gpus all -p 1312:1312 ghcr.io/psalias2006/gpu-hot:latest

Open http://localhost:1312

Building from Source

docker-compose up --build

Open http://localhost:1312

Requirements: Docker, NVIDIA Container Toolkit (install guide)

---

Why Not Just Use...

nvidia-smi CLI:

• Requires SSH access
• No historical data or charts
• Manual refresh only
• Hard to compare multiple GPUs

prometheus/grafana:

• Complex setup (exporters, databases, dashboard configs)
• Overkill for simple monitoring needs
• Higher resource usage

This is the middle ground: web interface with charts, zero configuration.

---

Features

7 Charts per GPU:

• Utilization, Temperature, Memory, Power Draw
• Fan Speed, Clock Speeds (graphics/SM/memory), Power Efficiency

Monitoring:

• Automatic multi-GPU detection
• GPU process tracking (PID, memory usage)
• System CPU/RAM monitoring
• Threshold indicators (temp: 75°C/85°C, util: 80%, memory: 90%)

Metrics Collected:

Core Metrics

• GPU & Memory Utilization (%)
• Temperature - GPU core & memory (°C)
• Memory - used/free/total (MB)
• Power - draw & limits (W)
• Fan Speed (%)
• Clock Speeds - graphics, SM, memory, video (MHz)

Advanced Metrics

• PCIe Generation & Lane Width (current/max)
• Performance State (P-State)
• Compute Mode
• Encoder/Decoder sessions & statistics
• Driver & VBIOS versions
• Throttle status

---

Installation

Pre-built Image (Easiest)

docker run -d --name gpu-hot --gpus all -p 1312:1312 ghcr.io/psalias2006/gpu-hot:latest

Build from Source

git clone https://github.com/psalias2006/gpu-hot
cd gpu-hot
docker-compose up --build

Local Development

pip install -r requirements.txt
python app.py

Verify GPU Access

docker run --rm --gpus all nvidia/cuda:12.1.0-base-ubuntu22.04 nvidia-smi

If this fails, install NVIDIA Container Toolkit first.

---

Configuration

None required. Optional customization:

Environment Variables:

NVIDIA_VISIBLE_DEVICES=0,1    # Specific GPUs (default: all)

Application (app.py):

eventlet.sleep(2)              # Update interval (seconds)
socketio.run(app, port=1312)   # Port

Charts (static/js/charts.js):

if (data.labels.length > 30)   // History length (data points)

---

API

HTTP

GET /                    # Dashboard UI
GET /api/gpu-data        # JSON metrics

WebSocket

socket.on('gpu_data', (data) => {
  // Real-time updates every 2s
  // data.gpus, data.processes, data.system
});

---

Extending

Add New Metric

1. Backend (app.py):

def parse_nvidia_smi(self):
    result = subprocess.run([
        'nvidia-smi',
        '--query-gpu=index,name,your.new.metric',
        '--format=csv,noheader,nounits'
    ], ...)

2. Frontend (static/js/gpu-cards.js):

// Add to createGPUCard() template
<div class="metric-value" id="new-metric-${gpuId}">
    ${gpuInfo.new_metric}
</div>

3. Chart (optional static/js/charts.js):

chartConfigs.newMetric = {
    type: 'line',
    data: { ... },
    options: { ... }
};

---

Project Structure

gpu-hot/
├── app.py                      # Flask + WebSocket server
├── static/js/
│   ├── charts.js               # Chart configuration
│   ├── gpu-cards.js            # UI rendering
│   ├── socket-handlers.js      # WebSocket events
│   ├── ui.js                   # View switching
│   └── app.js                  # Bootstrap
├── templates/index.html        # Dashboard
├── Dockerfile                  # nvidia/cuda:12.1-devel-ubuntu22.04
└── docker-compose.yml

---

Troubleshooting

GPU not detected:

# Verify drivers
nvidia-smi

# Test Docker GPU access
docker run --rm --gpus all nvidia/cuda:12.1.0-base-ubuntu22.04 nvidia-smi

# Restart Docker daemon
sudo systemctl restart docker

Debug logging:

# app.py
socketio.run(app, debug=True)

---

Contributing

Pull requests welcome. For major changes, open an issue first.

git checkout -b feature/NewFeature
git commit -m 'Add NewFeature'
git push origin feature/NewFeature

License

MIT - see LICENSE

---

Report Bug • Request Feature