xpu-top

Lightweight terminal dashboard for Intel XPUs that mimics the experience of nvidia-smi. It wraps xpumcli or xpu-smi (whichever is available) and renders colorful memory, power, and temperature bars in-place inside your terminal.

Features

• Auto-discovers xpumcli or xpu-smi, or accept a manual --cmd override.
• Shows device name, memory usage/utilization, power draw, and temperature per card.
• Smooth single-screen refresh using the terminal's alternate buffer (no scrollback spam).
• Handy xpu-top launcher script so you can run the monitor with one command from anywhere.

Requirements

• Python 3.8+
• Intel XPUM toolkit providing either xpumcli or xpu-smi on your $PATH
• A POSIX-like shell (for the xpu-top helper) and a terminal that supports ANSI escape sequences

Installation

1. Clone the repository and enter it:

   git clone https://github.com/Epsilon617/xpu-top.git
   cd xpu-top

2. Make sure the helper script is executable (should already be, but just in case):

   chmod +x xpu-top

3. (Optional) Add the repo root to your PATH so xpu-top can be invoked from anywhere:

   export PATH="$PWD:$PATH"
   # To persist, append the line above to ~/.bashrc or your shell profile

Usage

Run the dashboard via the helper script:

xpu-top

This launches xpu_monitor.py which by default tries xpumcli first and falls back to xpu-smi. The output stays within a single screen and updates until you press Ctrl+C.

Sample output

Intel XPU Monitor — Last update: 23:36:50.211
Press CTRL+C to exit. Source: xpumcli dump

Intel(R) Data Center GPU Max 1550 (ID 0)
  Mem  22.78 GiB / 128.00 GiB | ████████████████████████████████████████████████████████████  17.8%
  Power  355.7 W | Temp  49.1 °C
————————————————————————————————————————————————————————————————————————————————————————————————————

Intel(R) Data Center GPU Max 1550 (ID 1)
  Mem  25.52 GiB / 128.00 GiB | ████████████████████████████████████████████████████████████  19.9%
  Power  378.8 W | Temp  50.5 °C
————————————————————————————————————————————————————————————————————————————————————————————————————

Intel(R) Data Center GPU Max 1550 (ID 2)
  Mem   2.67 GiB / 128.00 GiB | ████████████████████████████████████████████████████████████   2.1%
  Power  306.1 W | Temp  47.0 °C
————————————————————————————————————————————————————————————————————————————————————————————————————

Intel(R) Data Center GPU Max 1550 (ID 3)
  Mem  57.15 GiB / 128.00 GiB | ████████████████████████████████████████████████████████████  44.6%
  Power  295.5 W | Temp  48.0 °C

Useful flags

• xpu-top --cmd xpu-smi — explicitly pick the backend executable.
• xpu-top --metrics 18,5,1,3 — pass a comma-separated list of XPUM metric IDs to match your toolkit.
• xpu-top --bar-width 60 — override the auto-detected bar width when you want wider graphs.

You can also call the Python entry point directly:

python xpu_monitor.py --cmd xpumcli

Environment variables

• XPUM_MONITOR_CMD — set to an absolute path or binary name (e.g. xpu-smi) to control the backend without passing --cmd every time.