Recent Papers

ABSTRACTThe development of neuromorphic devices has been fueled by the increasing need for computer systems that can mimic human intellect. With an emphasis on parameters like power consumption, memory retention and computational speed, this study assesses how well ion-based neuromorphic memory devices perform compared to traditional computing systems. Conventional systems, like Dynamic random-access memory (DRAM) and static random-access memory (SRAM), rely on electronic transport for data storage and processing, whereas ion-based devices use ionic transport mechanisms to mimic neural plasticity. Current-voltage (I-V) analyses and pulse stress tests were used to evaluate the devicescapacity to simulate biological processes such long-term depression (LTD) and long-term potentiation (LTP). In comparison to conventional DRAM and SRAM, the results showed better memory retention, dynamic adaptability, and low energy consumption (0.81.2 mW). Additionally, the devices demonstrated scalability and suitability for three-dimensional (3D) and nanoscale integration, making them attractive options for applications involving neuromorphic computing. However, issues like device stability under different voltages and environmental resilience were found, highlighting the necessity for additional optimization. With substantial potential for use in edge computing, robotics, and artificial intelligence, this research lays the groundwork for the development of ion-based neuromorphic memory systems.INTRODUCTION The von Neumann architecture, which divides memory and processing units, is the foundation of conventional computer systems (Kimovski, et al., 2023). Although this design works well for many applications, it contains built-in flaws that make it difficult to perform in contemporary, data-intensive activities. One significant issue is the memory wall, where bottlenecks are created by the necessity of data transfers between memory and processors, raising latency and decreasing overall efficiency (Schenk, Pei, Slesazeck, Schroeder, & Mikolajick, 2020). Furthermore, these systems are energy-inefficient since they constantly transport data back and forth, which uses a lot of power, especially when workloads call for frequent memory access (AlTwaijiry, 2021). These problems are made worse by scaling limitations; as memory components get smallerto sub-10nm dimensionstechnologies like DRAM and SRAM experience higher power consumption and leakage currents, which restricts their scalability for sophisticated applications. The two most popular memory types in traditional systems are DRAM and SRAM (Mittal, et al., 2021; Mutiu et al., 2022). Although DRAM's great density makes it valuable, its periodic refresh cycles are necessary to preserve data integrity, which results in energy inefficiency and retention constraints. SRAM, on the other hand, stores data using bistable flip-flops, which eliminates the need for refresh cycles and provides faster access times. However, it is more costly and less scalable than DRAM due to its greater cell size and higher power consumption (Mutlu, Ghose, Gmez-Luna, & Ausavarungnirun, 2022). Due to these drawbacks, DRAM and SRAM are unable to satisfy the increasing need for memory solutions that are scalable, flexible, and energy-efficient for contemporary computing requirements (Mittal, Verma, Kaushik, & Khanday, 2021). Inspired by the effectiveness and adaptability of the human brain, neuromorphic computing has become a potential paradigm to address these issues. In light of contemporary processing demands, the shortcomings of conventional von Neumann computer systems have become more apparent. These systems include inefficiencies including high latency and energy consumption, which are sometimes referred to as the "memory wall," because they rely on separate memory and processing units. On the other hand, by combining memory and computation into a single framework, neuromorphic computingwhich draws inspiration from the architecture and operation of organic neural systemsoffers an alternative. Through ionic transport, ion-based neuromorphic devices mimic synaptic characteristics, allowing for flexibility and adaptability. Because of this, they are ideal for applications that need low power consumption and dynamic learning. Research is still ongoing to determine how well these technologies function in comparison to more traditional systems like DRAM, SRAM, and Flash memory. Therefore, this study aims to analyze and compare the performance of ion-based neuromorphic memory devices and conventional computing systems across key metrics: power consumption, memory retention, computational speed, and adaptability.

Copyright © 2025 ABIOLA ODUTAYO ODUMERU. This is an open access article distributed under the Creative Commons Attribution License.