Ultra Low Power Associative Computing with Spin Neurons and Resistive Crossbar Memory

摘要

Emerging resistive-crossbar memory (RCM) technology can be promising forcomputationally-expensive analog pattern-matching tasks. However, the use ofCMOS analog-circuits with RCM would result in large power-consumption and poorscalability, thereby eschewing the benefits of RCM-based computation. Wepropose the use of low-voltage, fast-switching, magneto-metallic spin-neuronsfor ultra low-power non-Boolean computing with RCM. We present the design ofanalog associative memory for face recognition using RCM, where, substitutingconventional analog circuits with spin-neurons can achieve ~100x lower power.This makes the proposed design ~1000x more energy-efficient than a 45nm-CMOSdigital ASIC, thereby significantly enhancing the prospects of RCM basedcomputational hardware.