Ubuntu is a common environment for NVIDIA GPU workloads, so Nvitop fits naturally into development servers, workstations and remote SSH sessions. The safest setup order is driver first, Python second, package third.
Verify the NVIDIA layer
Run nvidia-smi to confirm that the kernel driver and userspace tooling can communicate with the GPU. Nvitop depends on NVML, which is supplied with the NVIDIA driver stack.
Check Python and terminal support
Use Python 3.8 or newer. Unix-like systems include curses support through the standard environment, with wide-character ncurses support expected by the project.
nvidia-smi python3 --version python3 -m venv .venv # activate .venv pip3 install --upgrade nvitop nvitop
Install Nvitop
Install the package with pip in a virtual environment, or use conda-forge, pipx or uvx if those match your package-management strategy.
Run locally or over SSH
Start with nvitop. Remote sessions can work well as long as the remote machine has the driver and NVML available and your terminal provides the required capabilities.
Separate driver installation from app installation
Do not treat a Python package reinstall as a fix for a broken NVIDIA driver. Solve driver visibility first, then address Python package or PATH issues.
Next steps
Use the official project documentation and your own environment records as the final authority for commands that can change after this guide was published. If you are downloading source through this site, the homepage explains exactly which archive the main button uses.