NXP's i.MX 93 Edge Machine Learning APU Is the First to Include an Arm Ethos-U65 microNPU

The first i.MX 9 application processor combines edge ML, security, and long-term support.

James Lewis
25 days agoMachine Learning & AI

NXP has introduced the first processor in their i.MX 9 product family. The i.MX 93 is an application processor targeting edge-powered machine learning applications with strengths in security, connectivity, cloud communication, and long-term availability.

The NXP i.MX 93 contains multiple domains and comes with different configurations. The main CPU is a single- or dual-core Arm Cortex-A55 that runs up to 1.7 GHz. In addition, the CPU supports external LPDDR4X DRAM.

An Arm Cortex-M33 runs the low-power real-time domain. This microcontroller is useful for time-critical operations. It has peripherals like a DMA controller, a watchdog timer, and a multitude of serial interfaces. In addition, this domain contains the EdgeLock Secure Enclave.

The flex domain expands connectivity with gigabit ethernet, USB, and an analog-to-digital converter. There are also camera and video interfaces for machine learning applications. Optionally, you can add an Arm Ethos-U65 microNPU. The i.MX 93 is the first application processor on the market to support this new Arm NPU.

Machine learning

The microNPU is a dedicated neural processing unit (NPU) for machine learning. For example, the combined interaction of the Cortex-M33 MCU and Ethos-U65 microNPU makes for a very low-power wake-on word detection solution.

NXP's eIQ Toolkit and eIQ Portal provide a simple way to develop, debug, optimize, and export machine learning models. This machine learning development kit supports the i.MX 93's Arm Ethos-U65 microNPU as an inference target.

Security and the cloud

The EdgeLock Secure Enclave contains on-die security and crypto features. While an obvious use for this enclave might be encrypted communications from the edge to the cloud, EdgeLock does more and acts like a "security HQ." It can manage critical security functions like root of trust, key management, and secure boot, to name a few.

All of these features fit into a larger ecosystem called EdgeVerse. NXP says EdgeVerse supported processors offer developers a simpler method to implement security in IoT devices.

NXP and Microsoft partnered to provide chip-to-cloud security. On the hardware side, the i.MX 93-CS variations contain a Microsoft Pluton module enabled in the secure enclave for built-in support for Microsoft Azure Sphere. Additionally, the two companies plan to support the platforms with security improvements for at least ten years.


Long-term support does not end with just the cloud applications.

As machine learning continues working its way into market segments like automotive and medical, there is a need for long-term component availability. As a result, NXP includes the i.MX 93 in their longevity support plan, which means NXP plans to support this processor for a minimum of 10 years.

More information

The i.MX 93 comes in two package sizes: a 9 x 9-millimeter package with 138 IO pins and an 11 x 11-millimeter package with 198 IO pins. The smaller package limits options to a single-core CPU and only provides parallel camera and display interfaces. NXP has not released details on pricing and shipping availability yet.

For more information, head over to the i.MX 93 product page.

James Lewis
Fan of making things that blink, fly, or beep. Host on element14 Presents, baldengineer.com, AddOhms, and KN6FGY.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles