Here’s another high-performance microcontroller from Embedded World, again, running an AI face tracking and detection demo. But there is a difference! On the Arrow booth, Alif Semiconductor’s Henrik Flodell shows us the Ensemble family of fusion processors, with a visual focus on the AI-accelerated E3 dual-core MCU as it runs the popular machine-vision demo at over 142 inferences per second.
Indeed, an MCU is doing that! Alif’s Ensemble family of fusion processors includes single- and dual-core Cortex M-55 MCUs accelerated with Arm Ethos-U55 microNPUs. This combination enables these microcontrollers to perform high-performance AI inferencing at increased speed and reduced power consumption when compared to previous Cortex M generations or an M-55 MCU alone. Other Alif Ensemble devices incorporate Cortex-A32 MPU cores in addition to the MCU cores, creating devices capable of running Linux at the same time as the MCU real-time operating system.
This all sounds quite power-hungry, but Alif addresses this issue with clever power management protocols that shut off parts of the chip that are not required for a given task. This means that a current draw comparable to that of purpose-built, low-power MCUs is possible when only this simpler functionality is required by the device. With this, Alif has created a family of devices that bring high-performance AI/ML capabilities to battery operated IoT devices.