The release is said to mark a significant leap forward in energy-efficient AI for sensor-edge applications, according to the company.
Innatera used CES 2024 to showcase the Spiking Neural Processor T1 through live demonstrations of radar and audio applications from Innatera's partners.
The processor is a versatile neuromorphic processing technology that is able to mimic the brain's mechanisms for processing sensory data. It uses a breakthrough analogue-mixed signal computing architecture purpose built to implement spiking neural networks (SNN), a unique breed of event-driven neural networks that are inherently well suited for pattern recognition and signal processing in noisy, time-series data.
Innatera said that the T1 system-on-chip could revolutionise the processing of sensor data at the edge, opening up new possibilities for applications in wearables, smart home and IoT devices.
With the device’s ultra-efficient, event-driven processing capabilities, the T1 enables significant power-performance gains for always-on sensing use-cases. With its analogue-mixed signal neuromorphic computing technology at its heart, the T1 also blends a RISC-V processor and support for accelerating traditional CNN models.
The Spiking Neural Processor is complemented by a powerful software development kit - Talamo. Integrated with the industry-standard PyTorch framework, Talamo provides a streamlined platform for the development and deployment of spiking neural network applications.
"Neuromorphic computing is here and will redefine intelligence at the sensor-edge. We're excited to unveil the Spiking Neural Processor and announce the availability of the T1 to customers for pre-production trials.", said Sumeet Kumar, CEO at Innatera.
T1 evaluation kits are available as part of Innatera's early access program, ahead of mass production later this year.