Purpose-Built Hardware for Physical AI
High-performance processing on power-constrained devices.
Neural Decision Processors
Our family of industry-leading Neural Decision Processors (NDPs) are specifically designed to run deep learning models, providing 100x the efficiency and 10/30x the throughput of existing low-power MCUs. From acoustic event detection for security applications to video processing in teleconferencing, our hardware can equip almost any device with powerful advanced intelligence, enabling real-time data processing and decision making with near-zero latency.
Seamless Integration
Our core architectures process neural network layers directly without any compilers, enabling larger networks at significantly lower power, greatly increasing battery life, privacy and responsiveness, while decreasing infrastructure costs.
With I/O support for a variety of sensor modalities, including vision, audio and motion, our NDPs bring physical AI to devices across domains, from earbuds and doorbells to automobiles and manufacturing machinery.
Neural Decision Processor Portfolio
| Status | Primary Modality | Package | Syntiant DNN* Core | MCU | DSP | ||||||
| Memory | Clock | Features | Memory | Clock | Memory | Clock | Features | ||||
| ⬇ NDP115 | Mass Production | Audio, Sensors | 2mm x 2mm 25-ball WLBGA | pmem: 448kB dmem: 64kB | Same as DSP | All Networks, Fixed Point | M0, 48kB | Same as DSP | imem: 64kB dmem: 144kB | 0.9V: < 22MHz 1.0V: < 60MHz 1.1V: < 100MHz | HiFi3: Fixed-point |
| ⬇ NDP120 | Mass Production | Audio, Sensors | 3.1 mm x 2.5 mm 42-ball WLBGA & 5mm x 5mm 40-pin QFN | pmem: 896kB dmem: 128kB | Same as DSP | All Networks, Fixed Point | M0, 48kB | Same as DSP | imem: 96kB dmem: 192kB | 0.9V: < 22MHz 1.0V: < 60MHz 1.1V: < 100MHz | HiFi3: Fixed-point |
| ⬇ NDP200 | Mass Production | Computer Vision | 5mm x 5mm 40-pin QFN | pmem: 896kB dmem: 128kB | Same as DSP | All Networks, Fixed Point | M0, 48kB | Same as DSP | imem: 96kB dmem: 192kB | 0.9V: < 22MHz 1.0V: < 60MHz 1.1V: < 100MHz | HiFi3: Fixed-point |
| ⬇ NDP250 | Sampling | Computer Vision, Sensor, Speech | 6.1mm x 5.1mm 120-ball eWLB | 6MB | Same as DSP | All Networks, Fixed Point | 0.5MB | Same as DSP | 1.5MB | 0.8V: < 80MHz 0.9V: < 120MHz | HiFi3: Fixed-point |
* Deep Neural Network
AISonic™ Audio Edge Processors
Our AISonic™ Audio Edge Processors are optimized for machine learning, allowing for high compute and low power operation for voice activation, hands-free control and contextual audio processing in a range of mobile devices, TWS headsets, portable speakers, TVs and other IoT devices.
Syntiant Powered Platforms
Syntiant Evaluation Kits
Syntiant evaluation kits are designed to help customers bring up the solution quickly and access features easily. The evaluation kit consists of a shield, featuring our Neural Decision Processor, connected to a Raspberry Pi Model 3B+ single-board.
RASynBoard
Developed in collaboration with Avnet and Renesas, the RASynBoard is a small, ultra-low power, edge AI/ML board, featuring a Syntiant NDP120, a Renesas RA6M4 host MCU plus a power-efficient DA16600 Wi-Fi/BT combo module.
Nicla Voice
Built in collaboration with Arduino and powered by the Syntiant NDP120, the Nicla Voice platform allows for easy implementation of always-on speech recognition at the edge, enabling multiple concurrent AI models.
Eoxys ML SOMs
Eoxys’ XENO+ ML (Machine-Learning) SOM (System-On-Module) with our NDP120 with Syntiant’s second-generation Core 2 neural network engine bring highly accurate, low-power inference to edge devices for both consumer and industrial segments. The XENO+ ML SOMs bring together AI and Machine-Learning capabilities with Wi-Fi/ BLE/ LTE wireless connectivity.