The Windows task manager got a new opportunity: the system utility now shows which programs the neuroprocessor and specialized neural blocks of the graphics card use. The new metrics help to understand where Windows is sending AI computing, and to notice a situation where the application instead of an energy-efficient accelerator loads the central processor or GPU. Microsoft explained that the extended statistics are primarily for the analysis of local AI loads.
The main change concerns the NPU, or neuroprocessor, of a specialized unit for performing neural network operations. In the tabs “Processes”, “Users” and “Details” appeared optional columns NPU and NPU Engine. The first shows the loading of the accelerator by a specific process, the second allows you to determine the used computing unit. In the “Details” tab, NPU indicators of dedicated and general memory are also available.
The task manager has learned to take into account the neural units built into the GPUs. In the “Productivity” tab, Windows now shows similar accelerators separately, so the user can see not only the overall loading of the GPU, but also the activity of the equipment intended for neural network operations. New options can be enabled through the context menu of column headers.
Microsoft began testing advanced statistics on March 30 in pre-assemblies of Windows 11. In late May, features were included in the update KB5089573 for Windows 11 24H2 and 25H2. The package updated the systems to builds 26100.8524 and 26200.8524, and the June cumulative update included changes from the May issue.
Advanced statistics will be useful for developers of applications using local neural networks. The program can be designed to execute the model through the NPU, but an unsupported operation, driver or format of computing can force the system to use another hardware unit. Without additional diagnostics, it is difficult for the developer to notice switching, especially if the program continues to work, but begins to consume more energy or performs tasks more slowly.
The benefits are not limited to development. Background blur during video calls, image processing, generative functions, speech recognition, and other local AI tasks can be performed on different components of the computer. NPUs were created for such calculations and usually reduce energy consumption compared to the constant load on the CPU or GPU. Task Manager now helps to determine the load source without installing a separate diagnostic software.
Microsoft’s rise to NPU’s focus coincides with the proliferation of Copilot+PC computers. To obtain the status of Copilot+, the device must have a neuroprocessor with a capacity of at least 40 trillion operations per second, at least 16 GB of RAM and a drive of 256 GB. Microsoft uses NPUs to locally perform part of the functions of artificial intelligence in Windows.
Together with NPU statistics, Microsoft has added an optional column, “Isolation” in the “Processes” and “Details” tabs. The parameter shows the applications running inside the AppContainer. The change does not directly relate to AI, but is included in the same set of task Manager improvements.
The expansion of capabilities is gradually transforming the Task Manager from a means to complete the hovering programs into a full-fledged monitoring tool for modern equipment. Along with the central processing unit, memory, drives, network and video card, Windows now allows you to track hardware AI accelerators up to individual processes, which is especially useful on new laptops with dedicated NPUs.
The main change concerns the NPU, or neuroprocessor, of a specialized unit for performing neural network operations. In the tabs “Processes”, “Users” and “Details” appeared optional columns NPU and NPU Engine. The first shows the loading of the accelerator by a specific process, the second allows you to determine the used computing unit. In the “Details” tab, NPU indicators of dedicated and general memory are also available.
The task manager has learned to take into account the neural units built into the GPUs. In the “Productivity” tab, Windows now shows similar accelerators separately, so the user can see not only the overall loading of the GPU, but also the activity of the equipment intended for neural network operations. New options can be enabled through the context menu of column headers.
Microsoft began testing advanced statistics on March 30 in pre-assemblies of Windows 11. In late May, features were included in the update KB5089573 for Windows 11 24H2 and 25H2. The package updated the systems to builds 26100.8524 and 26200.8524, and the June cumulative update included changes from the May issue.
Advanced statistics will be useful for developers of applications using local neural networks. The program can be designed to execute the model through the NPU, but an unsupported operation, driver or format of computing can force the system to use another hardware unit. Without additional diagnostics, it is difficult for the developer to notice switching, especially if the program continues to work, but begins to consume more energy or performs tasks more slowly.
The benefits are not limited to development. Background blur during video calls, image processing, generative functions, speech recognition, and other local AI tasks can be performed on different components of the computer. NPUs were created for such calculations and usually reduce energy consumption compared to the constant load on the CPU or GPU. Task Manager now helps to determine the load source without installing a separate diagnostic software.
Microsoft’s rise to NPU’s focus coincides with the proliferation of Copilot+PC computers. To obtain the status of Copilot+, the device must have a neuroprocessor with a capacity of at least 40 trillion operations per second, at least 16 GB of RAM and a drive of 256 GB. Microsoft uses NPUs to locally perform part of the functions of artificial intelligence in Windows.
Together with NPU statistics, Microsoft has added an optional column, “Isolation” in the “Processes” and “Details” tabs. The parameter shows the applications running inside the AppContainer. The change does not directly relate to AI, but is included in the same set of task Manager improvements.
The expansion of capabilities is gradually transforming the Task Manager from a means to complete the hovering programs into a full-fledged monitoring tool for modern equipment. Along with the central processing unit, memory, drives, network and video card, Windows now allows you to track hardware AI accelerators up to individual processes, which is especially useful on new laptops with dedicated NPUs.
