intel_gpu_top

intel_gpu_top [-d SECONDS] [-J] [-l] [-p PID] [-m]

-d, --delay=SECONDS
    Set minimum delay between screen updates (default: 1 second)

-J, --json
    Output stats in JSON format instead of curses UI

-l, --list
    List all available Intel GPUs and exit

-p, --pid=PID
    Monitor stats for a specific process ID only

-m, --minimal
    Show only GPU busy percentage (compact mode)

intel_gpu_top is a command-line utility for monitoring Intel integrated GPUs in real-time, similar to top(1) for CPUs. It displays key metrics including GPU utilization, engine loads (Render/3D/0, Video, Blitter, VideoEnhance), request counts, actual and requested frequencies, power draw, temperatures, voltage, and fabric clock. The interface uses ncurses for a dynamic, updating view in the terminal.

Ideal for developers, gamers, and system admins troubleshooting graphics performance, overheating, or power issues on Linux systems with Intel Arc, Iris Xe, UHD Graphics, or older iGPUs. Output refreshes at configurable intervals, and JSON mode enables scripting or logging. It relies on the DRM/i915 kernel driver for data via sysfs/debugfs.

Requires Intel GPU with i915 DRM driver; full access may need root or render group membership. No support for non-Intel GPUs. Data accuracy depends on kernel version and driver features.

On Debian/Ubuntu: sudo apt install intel-gpu-tools
On Fedora: sudo dnf install intel-gpu-tools

intel_gpu_top -d 0.5 for faster updates.
intel_gpu_top -J > gpu.json for scripted logging.

Developed by Intel as part of intel-gpu-tools package, first released around 2013-2014 alongside Linux kernel graphics improvements. Actively maintained for modern Intel GPUs like Tiger Lake, Alder Lake, and Arc discrete cards.

top(1), htop(1), radeontop(1)