Silicon Labs Promises to "Connect Just About Anything" with Its Edge AI-Capable Series 3 Parts

Silicon Labs has offered up a teaser for its upcoming Series 3 Internet of Things (IoT) product range, promising "dozens of products covering all major protocols and frequency bands, to connect just about anything" — which will launch with a triple-concurrency part designed as "the world's most flexible IoT modem" equipped module.

"AI [Artificial Intelligence] is rapidly becoming a key growth catalyst that will enable the number of IoT devices to reach over a 100 billion in the next decade,," claims Silicon Labs president and chief executive officer Matt Johnson. "Our upcoming Series 3 platform's unparalleled capabilities and productivity will unlock new applications and new capabilities across a vast range of industries, from manufacturing and retail to transportation, healthcare, energy distribution, fitness, and agriculture, helping transform each sector in remarkable ways."

The company has been slowly teasing its Series 3 range, which is the unsurprisingly-named successor to its current Series 2 family, for a while — but has now offered additional details as part of Johnson's joint keynote with chief technology officer Daniel Cooley at Embedded World North America this week. Among the snippets: the promise that the first Series 3 device " is designed to include the world's most flexible IoT modem, capable of true concurrency on three wireless networks with micro-second channel switching."

The Series 3 family, the company has confirmed, will be based on a multi-core architecture that pairs Arm Cortex-M cores with "dedicated coprocessors" for radio and security subsystems. Full specifications have yet to be announced, though SiLabs has stated that parts will range from a single-core Arm Cortex-M33 running at 133MHz to dual-core Cortex-M55 parts running at over 200MHz. All will support the company's Secure Vault High subsystem, including authenticated execute-in-place, and will feature "the world's most secure memory interface" and NIST's latest post-quantum encryption standards.

For those wondering what any of the above has to do with artificial intelligence, Johnson and Cooley offered an answer: selected parts, presumably situated at the higher end of the range, will include SiLabs' in-house second-generation Matrix Vector Processor neural network accelerator. This, the company claims, can boost machine learning and artificial intelligence performance by two orders of magnitude — delivering on-device edge AI on even battery-powered devices.

What SiLabs has yet to announce is availability and pricing for the new range — but it is expected to make some announcements during its Works With 2024 virtual developers' conference on November 20th-21st, with interested parties invited to register for free attendance.