Back in late 2018 Eta Compute Inc. (Westlake Village, Calif.) announced it was adding support for neural networks to its Cortex-M3 based microcontroller (see Eta adds spiking neural network support to MCU).
At the time the chip was called Tensai and included a Cortex-M3 operating at up to 100MHz clock frequency and an NXP CoolFlux DSP capable of accelerating machine learning.
Now manifested as the ECM3532 the key specifications are:
As low as 100microW active power consumption in always-on applications
Neural Development SDK with TensorFlow interface
16b dual-MAC DSP with 96kbytes dedicated SRAM for ML acceleration
Cortex-M3 processor with 256kbytes of SRAM, 512kbytes of flash memory
A 5mm by 5mm 81-ball BGA package.
Both cores feature the company’s patented Continuous Voltage Frequency Scaling (CVFS) and delivers power consumption of microwatts for many sensing applications.
Eta Compute has demonstrated the ECM3532 for image recognition and other edge sensing applications.
“This essentially eliminates battery capacity as a barrier to thousands of IoT consumer and industrial applications,” said Ted Tewksbury, CEO of Eta Compute, in a statement.
Related links and articles: