MOSFET Development for AI/ML Applications: Techniques, Comparisons and Challenges
摘要
There is a growing trend of miniaturization and making high performance semiconductor devices, which has ultimately led to the need for using advanced MOSFET technologies. Traditional MOSFETs like single gate MOSFETs and bulk MOSFETs face several challenges when device size is scaled down, mainly due to reduced gate control, leading to short channel effects, increase in leakage current, heat generation. The use of multi-gate MOSFETs such as FinFETs, GAA FETs (Gate-All-Around Field Effect Transistor), and SOI (Silicon on Insulator) technology serve as a potential solution to the issue. AI/ML applications require high computational power in order to accelerate model training and output generation. Modern MOSFETs like FinFETs and GAAFETs offer faster switching speed which is essential for computation tasks. Along with the advantage of fast switching speeds, advanced MOSFETs also contribute to achieving power efficiency, enabling robust AI hardware architecture. © 2017 Elsevier Inc. All rights reserved.