Task Manager adds visibility into AI workloads

What changed

The Processes, Details and Performance tabs add NPU, NPU engine, GPU engine and memory views. They expose whether an AI workload runs on the intended accelerator or falls back to CPU or standard GPU compute.

Why it matters

Endpoint support and application teams can diagnose AI workload placement, performance, power use and driver or model compatibility without a separate monitoring tool.

Known limitations

• The additional metrics require supported newer devices and compatible hardware and drivers.

Now check

• Enable the NPU, NPU engine, GPU and GPU engine columns on representative devices
• Confirm managed AI applications use the intended execution provider
• Review unexpected CPU fallback, shared-memory growth and battery impact