Research and Application of Compute-in-Memory Architectures: RRAM, MRAM, and FeRAM

With the rise of data-intensive applications such as artificial intelligence, big data analytics, and edge computing, the performance constraints of traditional memory technologies have become noticeable. Next-generation non-volatile memory devices, including resistive random-access memory (RRAM), magneto resistive random-access memory (MRAM), and ferroelectric random-access memory (FeRAM), are being actively explored to address demands for higher speed, lower power consumption, and greater integration density, especially within compute-in-memory (CIM) architectures. This study offers an overview of the operating principles, material mechanisms, performance characteristics, and reliability difficulties of RRAM, MRAM, and FeRAM. Furthermore, it covers their recent technological developments and illustrative applications in neuromorphic computing, embedded memory, and edge artificial intelligence devices. A comparative examination emphasizes each technology’s advantages: RRAM has high density and multilevel cell potential with cell dimensions down to 4F² and multilevel storage achieving over 7 bits per cell in some prototypes; MRAM features outstanding endurance exceeding 1012 white cycles and write/read energies as low as a few picojoules per bit; FeRAM is famous for its excellent reliability about 50-100ns and low power consumption. The research additionally examines development patterns and future potential for developing memory in advanced information processing systems.