Crossover MCU adds more punch for AI-Edge devices

“It brings increased compute performance and lower power performance making it ideal for AI-enabled edge devices, such as wearables, consumer medical devices, smart home devices and HMI platforms,” says Laurent Desseignés, i.MX RT MCU Marketing Director for NXP Semiconductors.

Featuring up to five powerful cores in a single device,the i.MXRT700 includes the first instantiation of the eIQNeutron NPU in a crossover MCU, accelerating AI workloads by up to 172x, while simultaneously reducing energy per inference by up to 119x. The i.MX RT700 crossoverMCUs also integrate up to 7.5MB ultra-lowpower SRAM and offer a 30-70% reduction in power consumption compared to previous generations.

The i.MX RT700 crossover MCU delivers on this need with a low-power, multicore design that ncorporates powerful graphics capabilities,AI hardware acceleration, advanced security and a sense compute subsystem. This empowers companies to create a full family of solutions with multimodal capabilities like presence sensing, gesture recognition,voice control, and more, all on a single unified platform.

Thei.MXRT700 crossover MCU family features up to five computing cores. This includes a primary Arm Cortex-M33 running at 325 MHz, with an integrated CadenceTensilica HiFi4 DSP for more demanding DSP and audio processing tasks.It also features NXP’s eIQ Neutron NPU, enabled with the eIQ machine learning software development environment. The family scales up to a best-in-class 7.5MB of ultra-low power SRAM with zero wait-state access.

The i.MX RT700 also includes an ultra-low power sense compute subsystem with a second Cortex-M33 and integrated Cadence Tensilica HiFi 1 DSP. This removes the need for an external sensor hub, reducing system design complexity, footprint and BOM costs.

To unlock the potential of AI on constrained devices like MCUs, the eIQ Neutron NPU efficiently accelerates AI inference by offloading the workload from the Cortex-M33.

This allows the eIQ Neutron NPU to accomplish tasks more quickly, such as up to 18x faster anomaly detection or up to 172x faster image classification, depending on the model, compared to running these tasks on the Cortex-M33 alone. With 7.5MB of onboard SRAM, i.MX RT700 can execute more complex multi-modal AI tasks that would be challenging for traditional microcontrollers.

“The i.MX RT700 family is designed to optimize power consumption, delivering a 30-70% improvement compared to previous generations,” says Desseignés. He cites a series of power architecture innovations, such as advanced adaptive deep sleep techniques, optimized clock architectures with low-power cache schemes, advanced wake/sleep cycle and more. As a result,end devices can either extend their battery life or can use smaller batteries, allowing more design flexibility.

Additionally, the i.MX RT700 family enhances system performance and flexibility with an integrated, high-efficiency DC-DC converter, basic memory management unit and supports improved analog peripherals. It also is NXP’s first crossover MCU to support the emerging eUSB standard,

Cyber security and privacy for smart connected consumer devices are now more important than ever, especially with upcoming cybersecurity regulations such as the U.S. Trust Mark and the European Cyber Resilient Act.Thei.MXRT700 family offers an accelerated compliance path to these regulations and other security standards for consumer devices, including ETSI 303 645.

Cyber resilience and protection of consumer’s data are at the core of the i.MXRT700 family. Desseignés points out an EdgeLock Secure Enclave (Core Profile), providing smart devices with advanced security functions, including secure boot with battery saving mode, secure update, seamless memory encryption and secure access to data. It also features strong device authentication with built-in physically unclonable function (PuF).

Thei.MXRT700 crossover MCU family is currently sampling for qualified early-access customers.