New Linux kernel patches currently undergoing review will allow AMD Ryzen AI NPU power metrics to be exposed under Linux. In turn this is useful for helping to gauge the utilization of the neural processing unit and also helping to evaluate the actual power efficiency of leveraging the AMD Ryzen AI NPU.
Linux enabling the AMD Ryzen AI NPU was off to a slow start in the first place with the the AMD XDNA Linux driver code having not been posted until nearly a year after the Ryzen 7040 series launch, several months after that when it began the kernel review process, and then only at the start of the 2025 calendar year when it was submitted for the mainline Linux kernel as part of Linux 6.14. The upstream AMDXDNA kernel “accel” driver continues to be worked on in the mainline tree while there are also varying efforts happening in user-space to leverage that kernel accelerator driver.
The latest work to note in kernel space for the Ryzen AI NPUs is power metrics reporting. This patch series working its way to the Linux kernel hooks into the AMD Platform Management Framework “PMF” driver code to integrate power data collection and reporting for the neural processing unit. Real-time NPU power data is then exposed to user-space for those curious about the power consumption of the NPU.
This will certainly be interesting and useful as more user-space software becomes available to actually utilize the Ryzen AI NPU under Linux for then helping to evaluate the impact on power efficiency of AMD’s NPU compared to the workloads on just the GPU or CPU cores.
