Maxim Integrated introduces a new neural network accelerator

     The new MAX78000 chip from Maxim Integrated. Based on a dual-core MCU, the MAX78000 combines ultra-low power deep neural network accelerators to provide the computing power needed for high-performance artificial intelligence (AI) applications, making it ideal for applications such as machine vision, facial recognition, object detection and classification, timing data processing, and audio processing.

     The Maxim MAX78000 integrates two MCU cores for system control, an Arm® Cortex®-M4 processor and a 32-bit RISC-V processor. Special features integrated with RISC-V processors enable fast loading of data into neural network accelerators with low power consumption. When configured and loaded with data, the MAX78000 442-KB Convolutional Neural network (CNN) accelerator runs AI inference 100 times faster than MCU solutions with less than 1% power consumption.

     The MAX78000 processor provides efficient power management to maximize the life of battery-powered Internet of Things (IoT) devices. Through dynamic voltage adjustment, the processor minimizes power consumption in the active core, with optional SRAM retention in low power mode. The processor combines low-power performance with efficient computing, reducing latency by a factor of 100, and is capable of performing AI reasoning at the iot edge.

     The MAX78000 comes with the MAX78000 Evaluation Kit. In addition to the MAX78000 processor, the evaluation kit includes a digital microphone, gyroscope/accelerometer, and a 3.5-inch touch color TFT display with support for a parallel camera module. We also supply the MAX78000FTHR, a development board in the Adafruit Feather form factor for rapid prototyping of small systems.