Featured image of post GPU Task Manager: terminal GPU monitoring tool released in v0.4.2 and v0.4.3

GPU Task Manager: terminal GPU monitoring tool released in v0.4.2 and v0.4.3

GPU Task Manager (gputaskman) has been released in versions v0.4.2 and v0.4.3.

The experimental, open-source terminal tool is an “htop for your GPU”: it shows per-process GPU memory and engine activity and tries to identify loaded LLMs with their memory breakdown. It runs on Linux (NVIDIA plus AMD and Intel), Windows, WSL and macOS.

Version 0.4.2 brings vLLM and llama-server metrics, AMD support in WSL and offline playback of recorded GPU sessions, among other things. On Windows, version 0.4.3 replaces the “N/A” line for Intel and Qualcomm GPUs with the GPU utilization.

Changes in version 0.4.2 (paraphrased):

  • Shows vLLM request and KV-cache gauges from /metrics, as the runtime reports them
  • Reports KV-cache usage and request concurrency from llama-server /metrics
  • Adds AMD GPU support in WSL via ROCm AMD SMI
  • Adds a gputaskman summary command for record files
  • Adds interactive offline playback of a recorded GPU session
  • Adds Xe per-process fdinfo memory and cycle accounting
  • Reads Xe primary-GT frequency and throttle sysfs values
  • Normalizes Windows backslashes when formatting Handy model paths
  • Changes report output to integer percentages and shortens the Intel VRAM caveat

Changes in version 0.4.3 (paraphrased):

Windows Intel/Qualcomm: replace the all-N/A device line with GPU % and one reason (Windows reports no temperature, power, clocks or PCIe for these GPUs). JSON unchanged.

GPU Task Manager

Source: GitHub

PlayingTux – Playing Games on Linux - since 1995.