Energy- and Area-Efficient Memristor-based Ternary Content Addressable Memory for High-Speed and Low-Power Application

Abstract

The shrinking feature sizes in CMOS-based technologies pose challenges, prompting a need for more efficient computing architectures. Memristor-based ternary content addressable (MTCAM) offers a promising alternative to conventional MOSFET based TCAM by offering a promising avenue for high-speed data searching and matching with reduced power consumption. TCAM's ternary format enables flexible data searching, vital for networking, pattern recognition, and security applications. However, conventional TCAM suffers from high costs, chip size, and power consumption. MTCAM presents a solution, offering scalability and efficiency to meet modern computing demands. The work presents a novel MTCAM design aimed at high-speed and low-power applications. This MTCAM structure, comprising of five transistors and four memristors, offers a unique approach to data preservation. By integrating WRITE and SEARCH operations on the same lines, the design addresses potential data overwrite issues during SEARCH operations through a dedicated data recall unit. Moreover, the architecture employs a consistent voltage level for both WRITE and SEARCH operations, thereby enhancing efficiency. The study conducted using HSpice simulations for a 32-word array, each containing 144 MTCAM cells, using 32nm 0.9V strained Si technology models. Employing a precharge-low current race (CR) match-line scheme, the simulations yielded impressive results: a search time of 215ps and search energy of 0.556fJ/digit/search for worst-case mismatches at a supply voltage of 1.2V, along with a substantial voltage margin of 343mV. Comparative analysis with conventional MOSFET-based TCAM and other existing designs signifies the superiority of the proposed MTCAM. It exhibits remarkable reductions in search delay (83%), search energy consumption (61%), voltage margin (14%), peak power consumption (65%), and estimated area (53%) compared to conventional MOSFET-based TCAM. Furthermore, it boasts improvements in search delay (55%), search energy efficiency (18%), voltage margin (7%), and peak power consumption (10%) compared to the latest existing MTCAM design. Notably, the area footprint of the proposed MTCAM cell is significantly smaller than that of other designs, making it a viable option for integrated circuit applications.