AI-Z

Terminal app for real-time monitoring of CPU/NPU/GPU performance timelines and AI related benchmarks. Supports AMD/Intel/NVIDIA. Provides many metrics all in one app to provide a holistic view including PCIE bandwidth, GPU Memory Controller usage, hottest core utilization, disk, network. Benchmarks CUDA, Vulkan, OpenCL and ONNX

N/GPU	Linux	Windows
NVIDIA	✓	✓
AMD	✓*	✓*
Intel	✓*	✓*

* some metrics are missing

Timeline screenshot

Install

Ubuntu/Debian

echo "deb [trusted=yes] https://www.ai-z.org/ stable main" | sudo tee /etc/apt/sources.list.d/ai-z.list
sudo apt update
sudo apt install ai-z

Windows 10/11

Download either the portable ZIP or the installer from the latest GitHub Release:

Latest GitHub Release

Direct downloads (always point at the latest release):

Portable ZIP (x64)
Installer (x64)
Portable: unzip and run ai-z.exe
Installer: run ai-z-windows-x64-setup.exe

Intel GPU metrics

On Intel Arc/Iris GPUs, metrics are queried via Intel Graphics Control Library (IGCL) and Windows D3DKMT API:

Supported metrics: - Temperature (°C) - Power draw (W) - via D3DKMT - GPU utilization (%) - GPU/Memory clock speed (MHz) - Fan speed (RPM) - VRAM temperature (°C) - Throttle state (power/thermal limiting)

Known limitations: - PCIe link info: The IGCL ctlPciGetState() API may return error 0x40000019 on some systems. This is a known Intel issue. The app attempts to fall back to Windows SetupAPI, but some Intel drivers don't report accurate PCIe link speed/width. - If you see PCIE:-- in the UI, this is expected on affected systems.

Troubleshooting: 1. Update to the latest Intel GPU driver. 2. Run ai-z.exe --diag-igcl-full to verify IGCL is working and see available metrics.

Build

Linux

sudo apt-get install -y build-essential cmake ninja-build pkg-config git
git clone https://github.com/microsoft/vcpkg.git
./vcpkg/bootstrap-vcpkg.sh
cmake -S . -B build -DCMAKE_TOOLCHAIN_FILE=./vcpkg/scripts/buildsystems/vcpkg.cmake
cmake --build build -j
./build/ai-z

Optional (developer): enable extra compiler warnings:

cmake -S . -B build -DAI_Z_ENABLE_WARNINGS=ON
cmake --build build -j

Build a `.deb`

Prereqs:

sudo apt-get install -y build-essential cmake dpkg-dev

Build + package:

cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j
cpack --config build/CPackConfig.cmake -G DEB

Note: If you want the resulting .deb to install on Ubuntu 22.04, build the package on Ubuntu 22.04 (or in a 22.04 container/chroot). Building on newer distros (e.g. 24.04+) will produce a package that depends on newer libc6 and libstdc++6 versions and will not be installable on 22.04.

This produces something like ai-z_0.1.0_amd64.deb in the build directory.

Install locally

For local installs (or downloading from a release page), users can install via apt:

sudo apt install ./ai-z_0.1.0_amd64.deb

Windows

Prerequisites: - Visual Studio 2022 with C++ development tools - CMake 3.16+ - Git

Build:

Clone the vcpkg package manager and bootstrap it:

git clone https://github.com/microsoft/vcpkg.git
.\vcpkg\bootstrap-vcpkg.bat

Configure and build ai-z with Visual Studio:

cmake -S . -B build -G "Visual Studio 17 2022" -A x64 `
  -DCMAKE_TOOLCHAIN_FILE=.\vcpkg\scripts\buildsystems\vcpkg.cmake
cmake --build build --config Release

Run:

.\build\Release\ai-z.exe

The TUI application will launch in your terminal. Use: - Arrow keys or vim-style keys to navigate - q or Ctrl+C to quit - Check the UI for additional keyboard shortcuts