Renesas claims it will be the first to demo AI and ML on an MCU based on an Arm Cortex M85 processor, doing so at Embedded World 2023 in Germany this week. A year ago, Tokyo-based Renesas demonstrated working silicon with the M85 at the event.
The long-running embedded event is expected to feature multiple vendors in AI and IoT applications, including large US-based companies such as Qualcomm, AMD, Nvidia, Intel and Microsoft. Some will have dedicated booths, while others will provide tech at partner booths.
One Renesas demo at the Nuremberg event will showcase a people detection app made with Plumerai, a vision AI company. The app can identify and track people in a camera frame under various lighting. Such a capability could be extended to low-cost, low-power AI for IoT, Renesas said.
Another demo will showcase a motor control predictive maintenance use case with an AI-based unbalanced load detection app that uses Tensorflow Lite for microcontrollers with CMSIS-NN.
The Cortex-M85 delivers more than 6 CoreMark/MHz8 for better compute on DSP or ML, but from a single, easily programmed Cortex M processor, Renesas claimed. Cortex-M is advertised as offering deterministic operation and short interrupt response time, along with low-power. It is based on Helium, an acceleration technology.
Renesas is the industry’s largest supplier of 16- and 32-bit MCUs and ships more than 3.5 billion of them a year, with about half for the automotive industry and rest for industrial and IoT app and data center and communications. Cortex-M is a family of 32-bit RISC Arm processors that first emerged in 2004 with the M85 launched in 2022.
“Renesas is a big supplier of IoT and especially automotive chips so this demo isn’t an insignificant thing, but it’s also not really a breakthrough by Renesas itself since they are leverage what Arm has provided,” said Jack Gold, an analyst at J. Gold Associates. “Cortex M85 core if Arm tech targeted a lower power needs, especially IoT and available to license by anyone.”
Renesas faces competition from Texas Instruments and Qualcomm in the same market and multiple Chinese chip companies, Gold added. He said the Cortex M85 MCU concept will likely be matched by others.
Plumerai CEO Roland Nusselder said the Arm Helium technology for the MCUs with Cortex-M85 will accelerate the Plumerai inference engine, but did not offer any specs. The increased performance will enable customers to use larger and more accurate version of the Plumerai People Detection AI, with extended battery life.
“Our customers have an insatiable appetite for adding new and more accurate AI features that run on a microcontroller,” Nusselder said in a statement.
Generally, Renesas says in its product brochure that its RA family of MCUs using Cortex M offers variable performance with the RA8 offering more than 240MHz with 2MB flash memory. A low-end RA2 offers performance up to 60MHz for analog.
Also at Embedded World, Renesas will demonstrate a new AML100 analog machine learning chip application board from Aspinity that directly connects to Renesas Quick-Connect IoT platforms for battery-operated IoT products. Aspinity claimed the AML100 is the first fully analog machine learning chip.
On March 1, Renesas introduced an online, cloud-based IoT system design platform for developers to design hardware and software to validate product prototypes. The Quick-Connect Studio is design to allow engineers to develop hardware and software at the same time for greater savings. Quick-Connect and the other Renesas news will be demonstrated in Hall 1, Stand 234 at Embedded World March 14-16.
Embedded World, started in 2003, brought in 1,100 vendors and 30,000 visitors in 2019, then took a break for the pandemic for two years. Last year, it brought in 720 vendors with 18,000 visitors.
AI is a hot topic expected this year for many vendors, with IoT applications front and center. Mobility, industrial and medical applications are featured.