gpustat
A command-line utility for querying and monitoring NVIDIA GPU status with a human-readable output.
What is gpustat?
gpustat is a Python-based command-line utility for monitoring NVIDIA GPU status. It provides a human-readable, colorized summary of GPU utilization, temperature, memory usage, and running processes, acting as a more concise and user-friendly alternative to `nvidia-smi`. It solves the need for quick, at-a-glance GPU monitoring in development and research environments.
Developers, researchers, and system administrators working with NVIDIA GPUs for machine learning, deep learning, scientific computing, or any GPU-accelerated workloads.
Developers choose gpustat for its faster query performance, cleaner output, and additional features like JSON export and watch mode, which are not as easily accessible in the standard `nvidia-smi` tool.
Overview
📊 A simple command-line utility for querying and monitoring GPU status
Use Cases
Best For
- Quickly checking GPU utilization and memory usage during model training
- Monitoring multiple GPUs in a server or workstation with a concise view
- Scripting GPU status checks with JSON output for automation
- Identifying which processes are using GPU resources and by whom
- Watching GPU metrics over time with periodic updates
- Ensuring consistent GPU indexing between CUDA and monitoring tools
Not Ideal For
- Environments with AMD or non-NVIDIA GPUs
- Systems using NVIDIA drivers older than version 450.00
- Projects needing raw, detailed GPU telemetry beyond summarized views
- Users requiring a graphical interface without separate alpha-stage tools
Pros & Cons
Pros
Displays essential GPU metrics like temperature, utilization, and memory in a single colorized line, making it faster to scan than nvidia-smi's verbose output.
Offers flags to show process owners, full commands, power usage, and encoder utilization, allowing tailored views for different monitoring needs.
Supports periodic updates with configurable intervals via --watch, enabling real-time monitoring without manual refreshes, ideal for tracking GPU trends.
Provides JSON export functionality, making it easy to integrate GPU data into automation scripts and tools for programmatic analysis.
Cons
Only supports NVIDIA GPUs with no AMD compatibility, as stated in the README, limiting its use in mixed or AMD-only hardware setups.
Requires specific NVIDIA driver versions (e.g., 450.00+ for gpustat 1.0+) and Python versions, which can complicate installation on older or restricted systems.
Lacks a built-in GUI; the separate web interface is in alpha and not part of the core tool, so users needing visual dashboards must look elsewhere.
Frequently Asked Questions
Related Projects
spdlogFast C++ logging library.
ctopTop-like interface for container metrics
psutilCross-platform lib for process and system monitoring in Python
nvtopGPU & Accelerator process monitoring for AMD, Apple, Huawei, Intel, NVIDIA and Qualcomm
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.