Scott Hanselman, a vice president at Microsoft and GitHub, has released an open-source GPU monitoring tool called gputaskman. The software aims to provide detailed insights into how specific processes utilize graphics hardware, addressing limitations in standard task managers.
The tool tracks metrics including GPU utilization, memory usage, temperature, power consumption, clock speeds, encoder and decoder activity, and PCIe traffic. It is designed to work across Windows, Linux, WSL, and Apple Silicon Mac platforms.
For users running local large language models, gputaskman identifies LLM processes and displays specific data such as model names, runtimes, weight quantization, layer counts, and memory allocation. This helps users monitor the impact of AI tasks on hardware.
Currently in an early experimental stage, the software has known limitations. Some users have reported issues with temperature data failing to refresh, and the developer notes that performance and functionality may vary depending on the specific hardware and operating system.
