Demonstration of Differential Mode FeFET-Array for multi-precision storage and IMC applications

Abstract

Harnessing multibit precision in non-volatile memory (NVM) based synaptic core can accelerate multiply and accumulate (MAC) operation of deep neural network (DNN). However, NVM-based synaptic cores suffer from the trade-off between bit density and performance. The undesired performance degradation with scaling, limited bit precision, and asymmetry associated with weight update poses a severe bottleneck in realizing a high-density synaptic core. In this work, we demonstrate: (i) implementation of novel differential mode ferroelectric field effect transistor (FeFET) (DM-FeFET) based multibit crossbar array of 12 Kbit size. (ii) bit density of 223Mb/mm2, which is ∼2x improvement compared to conventional FeFET array; (iii) 196 TOPS/W energy efficiency for VGG-8 network and (iv) superior bit error rate (BER) resilience showing ∼94% training and 88% inference accuracy with 1% BER.