A review on selective in-memory computing processors: Potential alternative to AI-driven applications

• Brief overview of the alternate computing concepts overcoming the Von-Neumann bottleneck. • An attempt to make a fair and systematic evaluation between the Non-Von-Neumann computational concepts. • The review encapsulates the importance of In-memory processor architectures in the Data-Intensive applications. • A detailed comparison of a selective In-memory processors by effective visual data representation. • Appropriate applications of the processors by considering the merits and demerits. In-memory computing (IMC) is a paradigm-shifting approach to data processing that eliminates the sluggishness of transferring data between memory and processing units. By integrating computation directly within the memory, IMC accelerates performance for data-dominant applications like artificial intelligence (AI), machine learning, big data analytics, and edge computing. This approach exploits the parallelism and proximity of memory elements to achieve lower latency, higher energy efficiency, and greater scalability compared to traditional Von Neumann architectures. Emerging memory technologies, including resistive RAM (ReRAM), phase-change memory (PCM), and spintronic devices, are key enablers of IMC, providing the foundation for non-volatile and densely packed memory arrays capable of performing arithmetic and logic operations. This work highlights the potential of IMC to transform computational paradigms by addressing the memory wall and energy constraints, driving innovation in both hardware design and software optimization to meet the demands of future computing workloads. This paper presents all the Non-Von-Neumann computational paradigm computing concepts, distinct memory devices and the IMC architecture suitable for AI-driven applications. The established outcome based on detailed review and analysis of IMC processors is that the potential alternative for data-dominant applications is In-memory computing due to its capability and efficiency of handling huge data.