One-Command Docker Deployment for GPU Monitoring Dashboard GPU Hot

A clean, beautiful, and lightweight real-time GPU monitoring solution that you can deploy with just a single command.

Preface#

When using a machine day to day, GPU usage is often the metric we care about most. I used to rely on Grafana + Prometheus + Nvitop for GPU monitoring. That stack is undeniably powerful, but its downsides are obvious: complicated configuration, relatively heavy resource usage, slow startup, and when you only want to quickly check temperatures and VRAM usage on a few cards, it hardly feels worth spinning up a full monitoring system.

Until I randomly came across this project on GitHub —

psalias2006

gpu-hot

Waiting for api.github.com...

00K

Waiting...

Deployment is extremely simple, the UI is clean and modern, and the feature set hits a very practical sweet spot. You can even think of it as a “GPU version of Glances” or a “lightweight GPU dashboard”.

Preview UI#

Below are screenshots from an actual deployment. You can see that each GPU has its own dynamic charts, covering not just temperature, power, and utilization, but also fan speed and VRAM usage trends.

gpu-hot-preview-1 gpu-hot-preview-2 gpu-hot-preview-3 gpu-hot-preview-4

The overall style has a strong “Grafana vibe”, but is far more lightweight, and the data refresh is very smooth.

Features#

Seven charts per GPU:#

GPU utilization
Temperature (core / memory)
VRAM usage
Power draw
Fan speed
Clock frequency (graphics core / SM / memory)
Power efficiency

Monitoring capabilities:#

✅ Automatic multi-GPU detection
✅ GPU process tracking (PID, VRAM usage)
✅ System CPU / memory (RAM) usage
✅ Threshold markers:
- Temperature: 75°C / 85°C
- Utilization: 80%
- VRAM usage: 90%

Collected metrics#

Core Metrics:

GPU and VRAM utilization (%)
GPU core / memory temperature (°C)
VRAM usage (used / free / total, MB)
Power: current draw and limit (W)
Fan speed (%)
Clock frequencies: graphics core / SM / memory / video decode (MHz)

Advanced Metrics:

PCIe generation and lane width (current / maximum)
Performance state (P-State)
Compute mode
Encode / decode sessions and statistics
Driver and VBIOS versions
Throttle status

Deployment (One-Command Docker Startup)#

Requirements#

✅ Docker
✅ NVIDIA Container Toolkit
👉 Official installation guide

Run command#

You only need a single line:

1
docker run -d --name gpu-hot --gpus all -p 1312:1312 ghcr.io/psalias2006/gpu-hot:latest

Then open your browser and visit:

http://localhost:1312

You’ll be greeted with a beautiful GPU monitoring dashboard. No extra configuration, no Prometheus, no database setup required.

Docker Compose (optional)#

If you prefer managing containers with docker-compose, you can use:

1
version: '3'
2
services:
3
  gpu-hot:
4
    image: ghcr.io/psalias2006/gpu-hot:latest
5
    container_name: gpu-hot
6
    ports:
7
      - "1312:1312"
8
    deploy:
9
      resources:
10
        reservations:
11
          devices:
12
            - capabilities: [gpu]
13
    restart: unless-stopped

Then run:

1
docker compose up -d

Comparison#

Item	nvidia-smi	Prometheus + Grafana	GPU Hot
Deployment	★☆☆☆☆ (CLI)	★★★★★ (complex setup)	★☆☆☆☆ (one line)
Historical data	❌ None	✅ Yes	✅ Built-in short-term charts
Real-time refresh	❌ Manual	✅ Auto	✅ Auto
Visualization	❌	✅ Highly customizable	✅ Lightweight & clean
GPU process details	✅	✅	✅
Resource usage	Very low	Medium–high	Low
Use cases	Quick ad-hoc checks	Enterprise / cluster monitoring	Personal / small workstations

From a positioning standpoint, GPU Hot is a very appropriate “middle-ground solution”:

Not as heavyweight as Grafana, but far nicer to look at than nvidia-smi. Ideal for single machines, multi-GPU workstations, or personal servers for day-to-day monitoring.

Real-world experience: pros & cons#

Pros:

One-command deployment, no need to configure exporters or databases
Automatic multi-GPU detection
Extremely clean UI with an excellent dark theme
Smooth real-time charts with low latency
Runs happily inside official NVIDIA containers

Cons:

No data persistence or historical querying yet
NVIDIA GPUs only
No API interface (in the current version)

Conclusion#

If you’re tired of the cumbersome Grafana + Prometheus monitoring stack, or you just want to quickly check GPU load and temperatures, GPU Hot is a small but very worthwhile tool to try out.