Windows Task Manager Gains NPU and GPU Neural Engine Columns

Per-process activity for the neural processing unit and the GPU's dedicated neural engine now appears on the Processes tab, alongside the familiar CPU and standard GPU columns, according to Microsoft's Windows IT Pro Blog post on the feature. The Performance tab adds matching overall utilization views for NPU and GPU neural engine, so admins can watch live usage graphs the same way they already do for CPU, memory, disk and network.

Microsoft says the underlying columns became available on some newer devices as part of the May and June 2026 non-security updates, per Message Center post MC1457685, which had a release start date of August 19, 2026. The company distinguishes standard GPU compute from GPU neural engine, describing the latter as dedicated AI matrix hardware built into the newest graphics processors rather than the rendering pipeline used for graphics workloads.

On the Processes tab, admins right-click the column header area and select NPU, NPU engine, GPU, or GPU engine to add them to the view. The Details tab offers a similar Select columns option, plus optional Dedicated NPU memory and Shared NPU memory columns for tracking model weight allocation, according to the Windows IT Pro Blog walkthrough. On the Performance tab, selecting NPU or GPU from the left-hand list, or right-clicking the graph and choosing View, brings up dedicated live utilization charts.

• NPU engine values appear as NPU, Compute or NPU, Neural when a workload is actively using the NPU.
• GPU engine values appear as GPU, Neural when a workload uses the GPU's dedicated AI matrix hardware.
• A workload stuck at 0% NPU with no NPU engine value, while CPU or standard GPU usage climbs, indicates a fallback to general compute.

Microsoft frames three main audiences for the feature in its Message Center post. IT professionals supporting helpdesk tickets can use the Processes and Performance tabs to see exactly how a background task interacts with processing engines when a user reports a problem with an AI-enabled app. Developers building on frameworks such as ONNX Runtime or Windows DirectML can confirm whether their execution provider bindings actually reach the NPU or GPU neural engine, since drivers can silently fall back to a higher-overhead pipeline on an unsupported data type or driver mismatch.

Power users get a simpler benefit: understanding why an app drains battery faster than expected. Microsoft's blog post contrasts a CPU-and-standard-GPU execution path, which it describes as capable of draining a battery in about two hours under heavy load, against an NPU-and-neural-engine path that uses far less power for the same AI task.

As AI-capable hardware spreads across the Windows fleet, IT teams have lacked a built-in way to confirm that AI-enabled line-of-business apps and vendor agents actually use the dedicated silicon vendors advertise. Microsoft's Message Center post rates both admin impact and user impact as low, and states the change requires no configuration action or service disruption, since it is a visibility addition rather than a behavioral change to any workload.

For example, an IT helpdesk investigating a complaint about a slow AI transcription tool can now open Task Manager's Processes tab, add the NPU and NPU engine columns, and see within seconds whether the app is running on the NPU or has fallen back to the CPU. That kind of on-device check previously required specialized profiling tools instead of a built-in system utility.

Microsoft's recommendation is straightforward: explore the new Task Manager metrics to evaluate application performance, troubleshoot resource bottlenecks, and understand how devices use AI-capable hardware, per the Message Center guidance. No compliance considerations are identified for this change, and Microsoft did not flag any required rollout action beyond reviewing the linked Windows IT Pro Blog article for use-case guidance.

Microsoft describes the current Task Manager additions as a foundation rather than a finished feature set, noting in its blog post that these capabilities provide a base for future hardware innovation across the Windows ecosystem. Microsoft's available materials do not detail specific further Task Manager or Performance Monitor additions beyond the NPU and GPU neural engine columns covered in this rollout.

Because the feature depends on newer NPU-equipped silicon and driver support, availability varies by device, and the available Microsoft materials reviewed do not include a full list of qualifying hardware. IT teams standardizing on AI PCs should confirm on a device-by-device basis whether the NPU, NPU engine, and neural engine columns appear before relying on them for fleet-wide troubleshooting.

The update gives IT staff, developers, and power users native visibility into whether AI workloads use the NPU or GPU neural engine instead of silently falling back to CPU or standard GPU compute.

Microsoft added per-process NPU and GPU neural engine columns to Windows Task Manager, documented in Message Center post MC1457685 and a Windows IT Pro Blog article.