Infineon Targets On-Device Artificial Intelligence, Machine Learning Workloads with the PSoC Edge

Infineon has extended its PSoC family of microcontrollers with a new range designed specifically for running machine learning and artificial intelligence (ML and AI) workloads at the edge: PSoC Edge.

"Infineon’s new PSoC Edge portfolio is built on years of experience delivering solutions to intelligently control, interact, and connect our world with highly usable and powerful MCU-based systems," claims Infineon's Steve Tateosian in an announcement brought to our attention by CNX Software. "They offer an ideal mix of scalable power and performance to support emerging AI/ML requirements, extensive HMI capabilities and security features required for next-generation applications. All supported with a deep ecosystem of software and tools, fit for task."

The new PSoC edge chips are built around Arm Cortex-M55 core with Helium vector processing extensions and Ethos-U55 neural processing unit (NPU) coprocessor — meaning that they're tailored for "tinyML" workloads. That doesn't mean they do away with energy efficiency, though: the company is also offering a lower-power Cortex-M33 core with its own NNLite proprietary hardware accelerator, for jobs where the full might of the M55-and-U55 combo isn't required.

The company is positioning the new parts as ideal for everything from wearables and smart home devices to use in security systems, human machine interfaces, and robotics — using its Infineon Edge Protect platform to deliver safety and privacy functionality where required. On the software side, developers can use Infineon's ModusToolbox or Imagimob Studio — now a wholly-owned Infineon subsidiary.

Infineon has yet to release full specifications and pricing for the new parts, with what details it has published available on the official product page.