NVMSurvey: Recent Advances and Comparative Analysis of Emerging Non-Volatile Memories (eNVMs)

Emerging non-volatile memories, specifically FeFET, PCM, RRAM, and STT-MRAM, promise significant advances in energy-efficient on-chip storage. Our research evaluates most recent (2020-2023) publications of such eNVMs against both optimistic and pessimistic projections sourced from 2016 to 2020 research using NVMExplorer, an open-source eNVMs simulator. Significant findings include FeFET’s low area, PCM’s improved energy efficiency and reduced read latency, and the evident advancements in STT-MRAM and RRAM. Moreover, we evaluate the potential benefits of recent eNVM solutions as memory resources for Deep Neural Network (DNN) accelerators and find that recent advances in RRAM and FeFET devices offer improved memory power and density for ResNet26 image processing. Overall, our study presents and analyzes continued research efforts on eNVM technologies as promising contenders to augment and replace conventional memory technologies.