topic Does KW45 supports AI/ML in Kinetis Microcontrollers

Does KW45 supports AI/ML

harisal — Tue, 10 Jun 2025 08:33:43 GMT

Dear Team,

May I know KW45 supports basic AI/ML applications.

By basic AI/ML I mean, Simple models (motion, sensor anomaly, binary classifier).

Also what is the current consumption of this Microcontroller

Regards

HARI

Re: Does KW45 supports AI/ML

Celeste_Liu — Fri, 13 Jun 2025 08:32:04 GMT

Hello @harisal ,

Thanks for your post. The KW45 doesn't feature with eIQ Neutron NPUs. Therefore, it does not support AI/ML in terms of hardware.

Please consider the MCX N94 and MCX N54: Arm® Cortex®-M33-Based Microcontrollers | NXP Semiconductors series or i.MX RT700 Crossover MCU with Arm® Cortex®-M33, NPU, DSP and GPU Cores | NXP Semiconductors series.

For power consumption, you can get it in datasheet. Take KW45 as an example, please refer to the KW45B (radio) and KW45Z (MCU) Data Sheet

BTY, if you want ask more questions about KW series, please create a post on Wireless MCU - NXP Community. That's the correct channel, the supporters there will help you better.

Hope it can help you.

BRs,

Celeste

--------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the "ACCEPT AS SOLUTION" button. Thank you!
--------------------------------------------------------------------------------------------------------------------

Re: Does KW45 supports AI/ML

Celeste_Liu — Fri, 13 Jun 2025 08:23:36 GMT

Hello @harisal ,

Please give me an opportunity to refine my response. If the requirement is to support AI/ML applications, the KW45 (with Cortex-M33 core and FPU) is feasible in theory. As detailed in Inferencing Deep Learning on Cortex M0/M0+ with the eIQTM Glow Inference Engine, eIQ Glow enables software-based deep learning inference on Cortex-M0/M0+ via model quantization. In practice, the KW45's FPU should deliver superior performance in floating-point operations compared to M0/M0+ devices, especially with its 1MB flash and 128KB SRAM specs that outperform the MCUs tested in the application note.

Note that solutions like NXP’s MCX Nx4x series and i.MX RT700 integrate eIQ Neutron NPUs for hardware accelerated AI inference, which can outperform software-based approaches in compute-intensive tasks.

Have a nice day.

BRs,

Celeste