gpustat

What it is and what it does

gpustat is a command-line utility that queries NVIDIA GPU status and displays it in a human-readable format. It shows GPU temperature, utilization percentage, memory usage (used/total), and a list of processes currently running on each GPU with their memory footprint. The tool is designed as a simpler, more readable alternative to nvidia-smi for quick GPU monitoring.

The package requires NVIDIA driver R450.00 or higher and uses NVIDIA's official python bindings (pynvml) to access GPU information. It supports Python 3.6 and later, offers watch mode for continuous monitoring, and can output results in JSON format. Options allow filtering by GPU index, displaying full command lines, fan speeds, power usage, and codec utilization. It is NVIDIA-only; AMD GPUs are not supported.

Use it for:

• Monitor GPU utilization and memory usage in real-time while running deep learning training jobs.
• Identify which processes are consuming GPU memory and by how much on a shared GPU cluster.
• Watch GPU temperature and thermal throttling during intensive compute workloads.
• Quickly check GPU availability and status before launching CUDA applications.
• Integrate GPU monitoring into shell scripts or monitoring dashboards via JSON output.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Displays real-time NVIDIA GPU status, memory usage, temperature, and running processes in a command-line interface, similar to nvidia-smi but with a simpler output format.

Yes, if you have NVIDIA GPUs and need quick, readable GPU status monitoring. The high install friction (external driver dependency) and NVIDIA-only support are real constraints, but the tool is actively maintained, permissively licensed, and widely used (top 5000 PyPI packages). Install it only if you already have NVIDIA drivers R450.00+ installed.

Install

gpustat on PyPI

pip install gpustat

uv add gpustat

poetry add gpustat

Installing gpustat

Before you install

High install friction: requires NVIDIA driver R450.00 or higher and the nvidia-ml-py library. The package itself has no runtime dependencies listed, but depends on external NVIDIA driver infrastructure. Maintenance is active with recent commits and a large community (4392 stars).

License in practice

MIT license is permissive and places no restrictions on use, modification, or redistribution in commercial or private projects.

Quickstart

pip install gpustat

gpustat

# With options:
gpustat --watch --show-cmd --show-user

Requires NVIDIA driver version R450.00 or higher and NVIDIA GPU hardware; will not work with AMD GPUs.

Verify before relying

• Whether the package works correctly with the latest NVIDIA driver versions beyond R450.00.
• Performance overhead when running with watch mode on systems with many GPUs or processes.

Package facts

Evidence: gpustat-1.1.1.tar.gz

Keywords: nvidia-smi, gpu, cuda, monitoring, gpustat

Tags

More Monitoring packages