Microsoft software engineer Scott Hanselman has launched a new free tool called gputaskman, designed to provide detailed and deep insight into how applications and local models consume graphics processing unit (GPU) memory on personal computers running Windows 11, going beyond the capabilities available in the traditional Task Manager.
Although the default Task Manager in Windows displays GPU and memory consumption at the level of individual processes, it fails to provide precise details about the nature of actual usage. The gputaskman tool bridges this gap by clarifying whether GPU memory is being consumed by a local AI model, a web browser, or the desktop compositor, along with displaying the GPU temperature.
Why this news matters
As tech companies like Microsoft lean toward hybrid AI initiatives that integrate local device processing with cloud services, the need for developers and advanced users to precisely understand the causes of hardware resource depletion is growing. Developer inquiries that inspired the tool point to the ambiguity surrounding memory consumption, such as tracking large language model (LLM) weights, the key-value (KV) cache, and how workloads are distributed between system RAM and GPU memory.
Current status and tool availability
The tool is currently in its early stages of development, supporting PCs powered by graphics cards from AMD and NVIDIA. The tool's developer has invited the tech community to contribute via GitHub, noting that future plans include expanding support to include devices such as the Surface Laptop Ultra and other hardware architectures based on Windows on Arm processors.