Neural processing units (NPUs) are now common in laptop processors and appear in some desktop chips, and more Windows apps use them for things like background blur, transcription and image effects. The trouble is knowing whether an app is really using the NPU. If it isn't, the same job may be running on the CPU or the graphics card's normal rendering pipeline, using more power and making more noise.
Microsoft has expanded Task Manager to show this. Forge PC hasn't tested these views on every processor; the steps below follow Microsoft's Windows IT Pro blog post on Task Manager's AI workload features (published 19 August 2026, updated 19 October 2026). Microsoft says the per-app data appears "on some newer devices", so what you see depends on your hardware and drivers. For a primer on what an NPU is, see what is an NPU?.
Press Ctrl + Shift + Esc to open Task Manager.
Open the Performance tab.
Select NPU in the list on the left, if your PC has one. Select GPU to see the graphics card's engines, including the neural engine on GPUs that have dedicated AI matrix hardware.
You can also right-click the graph area and choose View to switch between CPU, Memory, Disk, Network, GPU and NPU. If there's no NPU entry at all, your processor may not have one, or its driver isn't installed.
See which app is using it: the Processes tab
Open the Processes tab.
Right-click any column heading and tick NPU, NPU engine, GPU and GPU engine.
Find the app in the Name column and run the AI feature you want to check.
According to Microsoft, an app using the NPU shows NPU – Compute or NPU – Neural in the NPU engine column (the label depends on your processor and driver). An app using a GPU's AI hardware shows GPU – Neural in the GPU engine column.
More detail: the Details tab
The Details tab offers the same NPU, NPU engine, GPU and GPU engine columns: right-click a column heading and choose Select columns. It also has optional memory columns:
Dedicated NPU memory is memory that belongs to the NPU alone. Microsoft says this column is empty on integrated NPUs, which have no separate pool.
Shared NPU memory is system memory the NPU borrows. On a system with both, Microsoft says rising shared NPU memory means dedicated memory is full and allocations are spilling over, which can reduce performance.
The equivalent Dedicated GPU memory and Shared GPU memory columns work in the same way for the graphics card; our guide to checking VRAM usage explains those.
Signs an app is falling back to the CPU
Microsoft lists these symptoms of an app that isn't using the dedicated AI hardware:
The CPU column shows high usage while the NPU column is idle.
The NPU column stays at 0% and the NPU engine column is blank.
The GPU engine column doesn't show GPU – Neural.
Microsoft also suggests comparing the Power usage and Power usage trend columns, which can confirm the extra cost of a fallback.
Why your NPU might show no activity
The app doesn't use it. Intel's support guidance points out that not all applications use the NPU, and those that do may not use it all the time. Many local AI tools, including Ollama and LM Studio, run models on the graphics card or CPU instead; see NPU vs GPU for local AI.
Missing or old driver. Microsoft names a missing driver as one main reason for fallback. Intel advises updating the NPU driver to the latest version.
BIOS settings. Intel also says to update the BIOS and make sure the NPU is enabled in it.
Unsupported model format. For developers, Microsoft says a data type mismatch, such as trying to run an FP32 model on an INT8-optimised NPU, or an operation the NPU doesn't support, can cause a silent fallback.
Not powerful enough for some features. Microsoft's Copilot+ PC guidance says many of the new Windows AI features require an NPU capable of 40+ TOPS, so a less powerful NPU may sit idle while those features are unavailable.
Our verdict
To check NPU usage in Windows 11, open Task Manager, look for NPU on the Performance tab, then add the NPU, NPU engine, GPU and GPU engine columns on the Processes or Details tab. "NPU – Compute", "NPU – Neural" or "GPU – Neural" means the AI hardware is working; high CPU use with an idle NPU means the app is falling back. If nothing appears, update the NPU driver and BIOS and check the NPU is enabled. Forge PC hasn't tested every processor; these steps follow Microsoft's and Intel's guidance.