NeuRAM3 · NEUral computing aRchitectures in Advanced Monolithic 3D-VLSI nano-technologies

We propose to fabricate a chip implementing a neuromorphic architecture that supports state-of-the-art machinelearning algorithms and spike-based learning mechanisms. With respect to its physical architecture this chip willfeature an ultra low power, scalable and highly configurable neural architecture that will deliver a gain of a factor 50xin power consumption on selected applications compared to conventional digital solutions; and fabricated in Fully-Depleted Silicon on Insulator (FDSOI) at 28nm design rules. In parallel the project will be validating the modules torealise RRAM synapses both planar and in a 3D monolithic structure.We will complete this vision and develop complementary technologies that will allow to address the full spectrumof applications from mobile/autonomous objects to high performance computing coprocessing, by realising (1) atechnology to implement on-chip learning, using native adaptive characteristics of electronic synaptic elements;and (2) a scalable platform to interconnect multiple neuromorphic processor chips to build large neural processingsystems.The neuromorphic computing system will be developed jointly with advanced neural algorithms and computationalarchitectures for online adaptation, learning, and high-throughput on-line signalprocessing, delivering1. an ultra-low power massively parallel non von Neumann computing platform with non-volatile nano-scale devicesthat support on-line learning mechanisms2. a programming toolbox of algorithms and data structures tailored to the specific constraints and opportunities of thephysical architecture;3. an array of fundamental application demonstrations instantiating the basic classes of signal processing tasks.The neural chip will validate the concept and be a first step to develop a European technology platform addressingfrom ultra-low power data processing in autonomous systems (Internet of Things) to energy efficient large dataprocessing in servers and networks.