Skip to content

Latest commit

 

History

History
62 lines (42 loc) · 1.89 KB

README.md

File metadata and controls

62 lines (42 loc) · 1.89 KB

intel-npu-top

A minimalist, zero-dependency monitoring tool for Intel Meteor Lake NPU usage in real-time via sysfs on Linux. Features clean ASCII-based visualization in just 200 lines of Python.

NPU Usage Monitor Screenshot

Features

  • Dead Simple: Single Python file, zero dependencies - just copy and run
  • Real-time Monitoring: Live NPU usage with ASCII bar chart visualization
  • Usage History: Visual graph of the last 40 samples
  • Clean Interface: Pure Python + ASCII = prettier than intel_gpu_top
  • Auto-refresh: Continuous monitoring out of the box

Installation

Standalone Script (Recommended!)

Just copy STANDALONE.py to your system and run it - that's it! No dependencies, no package management, no fuss.

Also available via PyPI

pip3 install intel-npu-top

From Source

git clone https://github.com/DMontgomery40/intel-npu-top cd intel-npu-top python3 -m pip install .

Docker (if you really want to...)

docker-compose up Note: Docker is overkill for local monitoring - but great if you're pulling NPU data from another VM!

Requirements

  • Linux system with Intel Meteor Lake NPU
  • Python 3.6 or higher
  • Root privileges for NPU sysfs access (required for all installation methods)
    • Verify with: ls -lh /sys/devices/pci0000:00/0000:00:0b.0/accel/accel0

Usage

If using standalone script:

sudo python3 STANDALONE.py

If installed via pip:

sudo intel-npu-top

The interface shows:

  • Current NPU usage percentage with visual bar
  • Device information (vendor ID, device ID, etc.)
  • Historical usage graph
  • Runtime status and NUMA node details

Acknowledgements

This project builds upon ZoLArk173/nputop. Thanks for the inspiration and groundwork!

License

MIT License - See LICENSE file for details.

Just 200 lines of pure Python. No dependencies. No complications. It just works.