A 16nm 96Kb Integer/Floating-Point Dual-Mode-Gain-CellComputing-in-Memory Macro Achieving 73.3-163.3TOPS/W and 33.2-91.2TFLOPS/W for AI-Edge Devices

📋 论文概要

该论文提出了一种采用16nm工艺的96Kb整数/浮点双模式增益单元计算存储宏（CIM），解决了先进AI边缘芯片对计算灵活性和高能效的需求。该宏支持整数和浮点双模式运算，实现了73.3-163.3 TOPS/W的能效。

💡 主要创新点

• 提出双模式增益单元（Gain-Cell）计算存储宏，同时支持整数和浮点运算，增强了计算灵活性。
• 通过优化存储单元和计算路径，在16nm工艺下实现了高能效的CIM设计，适用于复杂的神经网络推理。

🏷 关键词

📄 原文摘要

Zhao-En Ke2, Ting-Chien Chiu2, Jun-Ming Hsu2, Chiao-Yen Cheng2, Yu-Chen Chen2, Chung-Chuan Lo2, Ren-Shuo Liu2, Chih-Cheng Hsieh2, Kea-Tiong Tang2, Meng-Fan Chang1,2 TSMC Corporate Research, Hsinchu, Taiwan National Tsing Hua University, Hsinchu, Taiwan 3 Industrial Technology Research Institute, Hsinchu, Taiwan *Equally Credited Authors (ECAs) 1 2 Advanced AI-edge chips require computational flexibility and high-energy efficiency (EEF) with sufficient inference accuracy for a variety of applications. Floating-point (FP) numerical representation can be used for complex neural networks (NN) requiring a high inference accuracy; however, such an approach requires higher energy and more parameter storage than does a fixed-point integer (INT) numerical representation. Many compute-in-memory (CIM) designs have a good EEF for INT multiply-and-accumulate (MAC) operations; however, few support FP-MAC operations [1-3]. Implementing INT/FP

👥 作者与机构

Win-San Khwa*1, Ping-Chun Wu*2, Jui-Jen Wu1, Jian-Wei Su2,3, Ho-Yu Chen2,