A RRAM Based 9T1R NVSRAM for Low-Power Computing in Memory

With the rapid advancement of AI technology, the amount of data has been exponential growth, necessitating memory solutions that offer high speed, low energy consumption, non-volatility, and high reliability in memory cell. In this brief, we propose a 9T1R non-volatile SRAM (NVSRAM) based on Resistive Random Access Memory (RRAM) for low-power computing in memory (CIM). This cell combines the advantages of SRAM and RRAM to efficiently execute CIM. Additionally, it can power down during idle periods and back up data to RRAM, reducing SRAM standby power consumption. We provide a detailed introduction to the operating principles of three distinct modes: SRAM mode, NVSRAM mode, and CIM mode. Circuit simulations for each mode are presented to validate the functionality and performance of the design. Comprehensive performance evaluation tests reveal that the proposed unit achieves excellent results in terms of read static noise margin, energy consumption, and operation latency, underscoring its potential for advanced memory applications.