Syntiant's NDP120 Deep Learning Chip Offers a 25x Performance Boost for Always-On Applications

New high-performance "Neural Decision Processor" aims to bring desktop-class performance to battery-powered devices.

Deep learning startup Syntiant, which recently celebrated shipping its millionth high-efficiency "Neural Decision Processor," has launched a new model in its accelerator family: the NDP120, designed for always-on audio and sensor applications in battery-powered devices.

"The NDP120 is the first of a family of semiconductors using our next generation Syntiant Core 2 tensor processor platform that brings performance levels previously found in plugged-in devices to a power level suitable to run on batteries," claims Syntiant CEO Kurt Busch of the launch. "We took years of real world, low-power edge deep learning experience to develop this architecture into a scalable design optimized to bring neural processing to power constrained deployments."

The NDP120 is built around the company's Syntiant Core 2 neural network inference engine, packaged alongside a highly-configurable front-end interface for audio work and hardware for multi-modal sensor fusion. Compared to the Syntiant Core 1 powering Syntiant's earlier NDP100 and NDP101 parts, the company claims the Syntiant Core 2 offers a twenty-fivefold increase in tensor throughput.

The ND120 includes support for up to seven audio streams, I2S and TDM audio output capabilities, a HiFi-3 digital signal processor (DSP), an Arm Cortex-M0 microcontroller with 48kB static RAM (SRAM), Quad SPI target and controller interfaces, I2C, up to 26 general-purpose input/output (GPIO) pins, flexible clock generation, and an onboard security system for firmware decryption and authentication tasks.

Syntiant is sampling the NDP120 now with a training development kit supporting bit-exact simulations within high-level languages including TensorFlow and Keras and support for all major frameworks. Volume production is scheduled for summer 2021, with pricing set at $6 per chip in 10,000-unit tray quantities. More information is available on the company's website.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire:
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles