Automotive MCUs Evolve with On-Chip AI and Zonal Support
The automotive sector is seeing the introduction of its first microcontrollers with dedicated AI acceleration, enabling more powerful on-device processing for tasks like anomaly detection. Concurrently, new MCU families like the S32K5 are being released to support the industry's shift to software-defined vehicles (SDVs) and zonal architectures.
- The STMicroelectronics Stellar P3E is the first MCU with an on-chip Neural-ART Accelerator, a neural processing unit (NPU) that provides up to 30 times greater AI efficiency than traditional MCU cores for microsecond-level inference latency. - NXP's S32K5 family, the industry's first 16nm FinFET MCUs with embedded MRAM, features Arm Cortex CPU cores running up to 800 MHz and a dedicated eIQ Neutron NPU to offload machine learning tasks. - Zonal architecture groups electronic control units (ECUs) by physical location rather than by function, a shift from traditional domain-based designs. This approach can dramatically reduce the length and weight of wiring harnesses; for instance, Tesla's Model 3 reduced its cabling from 3 km to 1.5 km. - The integration of on-chip AI accelerators allows for low-power, real-time processing of sensor data at the vehicle's edge, enabling applications like predictive maintenance and "virtual sensors" without adding costly, separate hardware. - This evolution in MCU design is a foundational element for the industry-wide transition to Software-Defined Vehicles (SDVs), which decouple hardware and software to allow for over-the-air (OTA) updates, new feature deployments, and subscription-based services. - To support the growing software complexity and frequent OTA updates in SDVs, these new MCUs incorporate advanced non-volatile memory technologies like phase-change memory (PCM) in ST's Stellar P3E and magnetic RAM (MRAM) in NXP's S32K5, offering higher density or faster write speeds than traditional embedded flash.
