Microsoft announced that the Windows 11 Task Manager will be updated to track Neural Processing Unit (NPU) usage for each process, following the existing monitoring standard for CPU, disk, and memory.
The NPU is a specialized component designed for executing machine learning and artificial intelligence tasks, and it is a requirement for Copilot+ PC certification on newer computers. In some devices, the graphics card (GPU) may include a neural engine for similar functions, in which case the Task Manager will also display data about its activity.
Although the announcement was made on the Microsoft Windows IT Pro Blog, no specific release dates or operating system versions where the feature will be available were provided; the company mentioned that it will arrive on 'some newer devices' without detailing which ones.
For users interacting with AI services such as ChatGPT, Claude, or Gemini, there is no cause for concern, as the processing of these activities occurs on the respective companies' servers, outside the local equipment, and therefore will not appear in the Task Manager.
Microsoft identified three main groups for whom this new monitoring will be beneficial: IT professionals who need to diagnose and report performance failures; developers looking to verify if models are utilizing the appropriate hardware resources; and advanced users interested in optimizing computer performance and battery life.
In the context of technical support, a professional can check in the Processes tab whether an AI software is using the NPU or the GPU's neural engine instead of the CPU or GPU, using the NPU, NPU engine, GPU, and GPU engine columns. Additionally, the Performance tab allows evaluating usage over time.
The information is even more detailed in the Task Manager Details tab, which displays columns showing the dedicated and shared memory for both the NPU and the GPU. This makes it possible to identify the source of unsatisfactory performance in an AI model, whether due to data type incompatibility or the absence of a driver.
For the advanced or curious user, the functionality helps determine whether a specific AI application is using the NPU—which is more efficient for such tasks—or the CPU, which is slower and consumes more energy.
However, the implementation of the new features may present complexities. As pointed out in a Windows Forum thread, various configurations, such as workstations with dedicated GPUs and large VRAM, integrated Intel or AMD platforms, and ARM laptops with Qualcomm NPUs, may exhibit different memory patterns, even when tasks are correctly optimized for the devices.
Furthermore, the diagnostics required for developers and IT professionals will still require more comprehensive tools. Nevertheless, the Task Manager establishes itself as a valuable starting point for investigating AI-related issues.