+442038075483
support@nicktailor.com
My Git Repos

Day: January 21, 2026

| January 21, 2026

nvitop – The Ultimate Interactive NVIDIA GPU Monitoring Tool

If you’re working with NVIDIA GPUs  whether for deep learning, HPC, or systems administration  you’ve likely used the default nvidia-smi tool to check GPU status. But what if you want something more dynamic, interactive, and user-friendly? Enter nvitop, an incredible interactive NVIDIA GPU process viewer that makes monitoring your GPUs intuitive and informative.

nvitop in action - animated demonstration
nvitop in action — real-time GPU monitoring with an interactive interface

nvitop is a Python-based GPU monitoring tool that extends and enriches what nvidia-smi does  with live updates, color coding, interactive filtering, and more. It’s perfect for developers, engineers, researchers, and admins who need real-time insight into how GPU resources are being used.

What Makes nvitop Different?

Unlike nvidia-smi, which outputs a static snapshot, nvitop offers a fundamentally different experience:

  • Interactive Real-Time Monitoring — Runs as a live monitor that refreshes continuously, similar to htop for CPUs
  • Color-Coded Output — Intuitive visuals make it easy to spot heavy GPU usage and memory pressure at a glance
  • Process Management — Sort, filter, and manage processes directly with keyboard controls
  • Rich Data Display — GPU metrics including utilization, memory, temperature, power, and processes in a compact, organized view
  • Cross-Platform — Works on both Linux and Windows
nvitop monitor mode interface
Monitor mode of nvitop showing GPU utilization, memory usage, and running processes

Side-by-Side Comparison with nvidia-smi

The difference becomes immediately apparent when you compare nvitop’s output with the traditional nvidia-smi tool:

nvitop compared to nvidia-smi
nvitop (top) vs nvidia-smi (bottom) — notice the richer information density and visual clarity

Installation

Installing nvitop is straightforward. If you have Python installed, you can use pip:

pip install --upgrade nvitop

Or, if you prefer conda:

conda install -c conda-forge nvitop

For isolated environments using uvx or pipx:

uvx nvitop
# or
pipx run nvitop

Once installed, simply run:

nvitop

This launches the interactive monitor showing your NVIDIA GPUs, their current utilization, and the processes using them — all updating live.

What You’ll See

Running nvitop in your terminal displays a live dashboard with comprehensive GPU stats:

  • GPU temperature and fan speed
  • Memory usage with bar charts
  • GPU and memory utilization percentages
  • Power consumption
  • Process list with GPU memory and compute usage per process
  • History graphs for utilization trends

It’s like combining the best of nvidia-smi, top, and htop — but specifically designed for your GPUs.

nvitop process filtering
Process filtering and colorful interface — easily identify resource-heavy processes

Process Metrics and Management

One of nvitop’s standout features is the ability to dive deep into individual process metrics. Select a process and press Enter to see detailed live graphs:

nvitop process metrics screen
Process metrics screen — watch detailed GPU usage for a specific process

You can also manage processes directly from the interface:

  • Ctrl+C or I — Send interrupt signal (SIGINT)
  • T — Terminate process (SIGTERM)
  • K — Kill process (SIGKILL)
  • t — Toggle tree-view to see process hierarchies
  • e — View environment variables

Windows Support

Unlike many GPU monitoring tools that are Linux-only, nvitop works natively on Windows as well:

nvitop running on Windows
nvitop running on Windows with PowerShell in Windows Terminal

Built-in Help

Press h at any time to access the comprehensive help screen with all available keybindings:

nvitop help screen
The built-in help screen — press h to access

Why This Matters

Real-time GPU visibility is crucial in many modern workloads:

  • Deep Learning Training — See which models or data pipelines are consuming your GPU resources and identify bottlenecks
  • HPC / Multi-User Servers — Quickly identify who is using GPUs and how much, essential for shared compute environments
  • Debugging — Spot processes consuming excessive memory or identify stuck jobs that need intervention
  • DevOps Monitoring — Integrate with larger monitoring stacks using nvitop’s Python API or the new nvitop-exporter for Grafana dashboards

Key Features Summary

Feature Description
Live Monitoring Continuous updates with configurable refresh intervals
Process Management Kill, terminate, or interrupt processes directly
Tree View See process hierarchies and parent relationships
Device Selection Includes nvisel tool for CUDA device selection
Python API Full programmatic access for custom monitoring tools
MIG Support Works with NVIDIA Multi-Instance GPU configurations
Grafana Integration Export metrics via nvitop-exporter for dashboards

Common Usage Examples

# Basic monitoring (auto display mode)
nvitop

# One-shot query (like nvidia-smi)
nvitop -1

# Full display mode
nvitop -m full

# Compact display mode
nvitop -m compact

# Only show specific GPUs
nvitop -o 0 1

# Only show CUDA visible devices
nvitop -ov

# Colorful spectrum-like bar charts
nvitop --colorful

# For light terminal themes
nvitop --light

Summary

If you’re into HPC and GPU diagnostics, this is a cool tool to learn and play with. Nick Tailor tech choice award for sure.

Check out the official repository on GitHub: https://github.com/XuehaiPan/nvitop

Full API documentation is available at: https://nvitop.readthedocs.io

January 21, 2026 HPCNo Comments »