Microsoft VP releases open-source GPU task manager for VRAM usage insights

New Open-Source GPU Task Manager Released

Microsoft VP Scott Hanselman has introduced gputaskman, a new open-source GitHub project. This tool functions as a GPU-focused task manager, offering detailed insights into GPU resource allocation and usage across different processes.

gputaskman is compatible with Windows, Linux, and macOS operating systems. It provides users with information such as per-process GPU usage, VRAM consumption, active runtimes, and other statistical data related to GPU performance.

Addressing GPU Memory Visibility

Hanselman developed gputaskman to address the lack of detailed visibility into GPU memory usage. He noted that there was a need for a tool to understand where GPU memory was being consumed by various applications.

The tool's interface displays data in a terminal-like format, showing which applications are using the most VRAM. Screenshots demonstrate its ability to identify VRAM-hungry applications, including the desktop window manager and AI/LLM applications.

AI/LLM Specific Details

A notable feature of gputaskman is its ability to break out AI/LLM task-specific information. When an AI/LLM application like llama-server.exe is selected, the tool provides details such as the specific model being used (e.g., Qwen3-0.6B), its weights, KV cache, and the number of parameters.

This level of detail is intended to be informative for users working with AI and large language models, offering a clearer understanding of how these applications utilize GPU resources.

Availability and Future Outlook

gputaskman is released under the permissive MIT license, making its source code available for review and modification. Portable Windows downloads of version 0.3.0 are currently available.

Hanselman has also hinted at the possibility of expanded GPU monitoring capabilities being integrated into Windows as a standard feature in the future, suggesting a broader impact for this type of tool.