Syntiant , a provider of deep learning solutions making edge AI a reality for always-on applications in battery-powered devices, announced its new Syntiant® NDP102 Neural Decision Processor™ (NDP) for sensor applications that delivers 100x improvement in efficiency and performance compared to current MCU-based offerings.
Capable of performing sensor processing at under 100 microwatts, the NDP102 can operate as a standalone always-on component of a larger system. With a package size as small as 1.71 mm x 2.51 mm, the NDP102 can be used in a variety of edge AI applications, such as event detection, pressure sensing, gesture recognition, sensor fusion and other condition-based monitoring use cases.
“Our new NDP102 compliments our family of Neural Decision Processors by providing an easily deployed machine learning solution for always-on sensor processing in even the smallest of devices,” said Kurt Busch, CEO of Syntiant. “Whether used in wearables such as personal fitness trackers or with condition-based monitoring systems in the home or factory, the NDP102 serves as a powerful AI interface for sensor processing in a tiny package with near-zero power consumption.”
Syntiant’s architecture enables machine learning workloads that are typically run on cloud servers to be performed on-device in the always-on domain. The NDP102 allows critical and time-sensitive decisions to be made faster, more reliably and with greater security, such as remote patient monitoring in next-generation medical devices, or event detection in smart homes. Syntiant’s NDP102 can continuously monitor vibrations or temperature to detect and act on any anomaly before costly downtime failures occur.
“We are very excited at how easy it is for developers to incorporate AI into their products with the NDP102,” said Jue Chang, IoT product manager for Murata Manufacturing Co., the leading global manufacturer of wireless connectivity modules. “By adding Murata wireless modules and sensors, we can offer customers a complete solution for a range of low-power IoT applications, including smart home and smart appliances.”
The NDP102 provides a self-contained solution for always-on sensor processing. Leveraging the Syntiant Core 1™ architecture, the NDP102’s native neural network processing capabilities eliminate the need for intermediate compilers and works seamlessly with popular machine learning frameworks such as TensorFlow, which greatly reduces time to market and helps assure expected performance.
Key features include:
– Under 100μW active power consumption in always-on sensor applications
– 1.71 mm x 2.51 mm 20-pin eWLB package (0.4 mm ball pitch)
– Syntiant Core 1™ neural processor
– Hardware feature extraction up to 100Hz frame rate and up to 40 frequency bins per frame
– Deeply embedded Arm Cortex-M0 processor with 112KB SRAM
– Six GPIO pins with programmable direction and drive strength
– Fully connected network, supporting up to 589k parameters
– I2C serial interface option for sensor applications
– On-chip programable clock divider
– External serial flash boot
The Syntiant NDP102 is sampling now and is expected to start shipping in production volumes in Q4 2021. Pricing for 10Ku quantities is $3 per unit.
