Search

We present TAICHI, a general in-memory computing deep neural network accelerator design based on RRAM crossbar arrays heterogeneously integrated with local arithmetic units and global co-processors to allow the system to efficiently map different models while maintaining high energy efficiency and throughput. A hierarchical mesh network-on-chip is implemented to facilitate communication among clusters in TAICHI to balance reconfigurability and efficiency. Detailed deployment of the different circuit components is discussed, and the system performance is estimated at several technology nodes. The heterogeneous design also allows the system to accommodate models larger than the on-chip storage capability. [ABSTRACT FROM AUTHOR]

A comprehensive analysis of write operations (SET and RESET) in a resistance-change memory (resistive random access memory) crossbar array is carried out. Three types of resistive switching memory cells-nonlinear, rectifying-SET, and rectifying-RESET-are compared with each other in terms of voltage delivery, current delivery, and power consumption. Two different write schemes, V/2 and V/3, were considered, and the V/2 write scheme is preferred due to much lower power consumption. A simple numerical method was developed that simulates entire current flows and node voltages within a crossbar array and provides a quantitative tool for the accurate analysis of crossbar arrays and guidelines for developing reliable write operation. [ABSTRACT FROM AUTHOR]