Along with 11 other research and industry partners, SynSense joins Ferro4EdgeAI, an EU-funded research consortium to develop an ultra-low power, scalable edge accelerator incorporating a memory-augmented neural network, based on low cost, high density, multi-level, Back End of Line (BEoL) integrated ferroelectric (FE) technology.  SynSense will apply their mixed-signal neuromorphic IC and non-volatile memory technologies to the new AI hardware. The project will achieve a 100-2500x gain in energy efficiency, breaking the POPS/W barrier of CMOS accelerators and predictions for other emerging technology AI hardware.  The outcome of the project will empower a wide range of applications such as consumer electronics, automotive, aerospace, and more.

With the rise of the Internet of Things, edge sensors and smart devices are generating up to 40 times more data than cloud/data centers. The annual growth rate of the edge AI hardware market has exceeded 40%, and it is projected to surpass $70 billion by 2026. Handling massively distributed quantity of data using pure cloud computing model is challenging. Innovative approaches involving embedded non-volatile memory (eNVM) arrays, which enable local computation and memory-augmented analogue AI accelerators are widely recognized as efficient strategies to achieve large-scale edge AI. 

Ferro4EdgeAI emerges within this trend. In recent years, FE technology has witnessed continuous breakthroughs, harnessing its spontaneous polarization that can be reversed in an external electric field and its non-volatile property. Based on the optimized FeFET-2 devices, the consortium partners will contribute expertise across the whole value chain, from device technology, circuit design, array manufacturing, to prototypes and systems simulations, aimed at maximizing the advantages of the technology – reliability, durability, high energy efficiency, low latency, low cost, and scalability.

This project represents a continuation of SynSense’s collaborative research efforts. SynSense has engaged in a range of EU artificial intelligence projects such as EdgeAI, TEMPO, MemSCALES, Andante, and SYNCH. This involvement has not only delved into non-volatile devices for neuromorphic inference ICs but also builds on the long expertise of SynSense in ultra-low power mixed-signal processing. This work leads to the mixed-signal device DYNAP™-SE2, as well as the software toolchain Rockpool for building and deploying Spiking Neural Networks (SNNs) on mixed-signal hardware.

“Our work will assist the Ferro4EdgeAI consortium to refine their HW designs, with a view to producing ultra-low-power commercial neuromorphic compute-in-memory and near memory devices,” said Dylan Muir, Vice President of Global Research Operations at SynSense.

SynSense will collaborate closely with partners including STMicroelectronics, CEA, NaMLab, TU Delft, ECL and Ferroelectric Memory Company. The optimized FE-based, ultra-low power analogue accelerators for neuromorphic AI will become a significant part of SynSense’s commercial neuromorphic solutions. These solutions will assist their customers in various fields such as wearable health devices, smart mobility, environmental monitoring, agriculture, smart city, to further enhance product value and embrace the revolution of edge intelligence.

About SynSense

SynSense | Neuromorphic Intelligence & Application Solutions

SynSense is a world-leading neuromorphic engineering company. SynSense provides custom-tailored, ultra-low-power silicon design solutions for industrial and consumer machine-learning inference applications. As a “full-stack” neuromorphic engineering company, SynSense delivers complete solutions, including custom IP, hardware, and software configurations to meet specific application needs. SynSense was founded in March 2017 in Zürich Switzerland, based on ground-breaking advances in neuromorphic computing hardware developed at the Institute of Neuroinformatics of the University of Zurich and the ETH Zurich.

SynSense@Linkedin Twitter

Media Contacts

Dylan Muir, dylan.muir@synsense.ai

Zurich, Switzerland

Nancy Huang,  juan.huang@synsense.ai, 028-62073881

Chengdu, China