Microsoft VP releases open-source GPU Taskman to track VRAM and AI workloads

Microsoft Vice President Scott Hanselman has introduced gputaskman, a free, open-source GPU task manager released on GitHub to help users understand how graphics-card memory is being consumed. Announced on November 3, 2026, the tool provides per-process visibility across Windows, Linux and macOS, reporting GPU utilization, VRAM allocation, active runtimes and other operational metrics. Unlike standard system monitors, gputaskman also separates AI and large-language-model workloads, offering users a clearer view of which processes are responsible for model loading, inference or training activity. The project is positioned as a diagnostic aid for developers, system administrators and power users who need to troubleshoot memory pressure, resource contention and GPU bottlenecks. Hanselman described the effort as a way to gain insight into where GPU memory is going and invited contributions on GitHub. For environments where AI applications increasingly compete for limited VRAM, the tool aims to make otherwise opaque driver and runtime allocations easier to trace and manage.