⚡ 本页包含 AI 生成的分析内容，仅供参考
该论文提出了一种基于22nm工艺的8Mb STT-MRAM近存计算宏，支持8位精度和46.4-160.1TOPS/W的能效，适用于边缘AI设备。通过非易失性存储器实现神经网络参数断电存储，并快速响应设备唤醒，解决了边缘AI低功耗和高能效需求。
Yu-An Chien1, Guan-Yi Lin1, Po-Jung Chen1, Tsen-Hsiang Pan1, De-Qi You1, Fang-Yi Chen1, Andrew Lee1, Chung-Chuan Lo1, Ren-Shuo Liu1, Chih-Cheng Hsieh1, Kea-Tiong Tang1, Yu-Der Chih3, Tsung-Yung Chang3, Meng-Fan Chang1,2 National Tsing Hua University, Hsinchu, Taiwan TSMC Corporate Research, Hsinchu, Taiwan 3 TSMC, Hsinchu, Taiwan *Equally Credited Authors (ECAs) 1 2 Nonvolatile-memory-based computing in memory (nvCIM) [1-6] is ideal for low-power edge-AI devices requiring neural network (NN) parameter storage in the power-off mode, a rapid response to device wake-up, and high energy efficiency for MAC operations (EFMAC). Current analog nvCIMs impose a tradeoff between the signal margin (SM) and the number of accumulations (NACU) per cycle versus EFMAC and computing latency (TCD-MAC). Near-memory computing (NMC), with high precision for inputs (IN), weights (W), and outputs (OUT), and a high NACU is a trend to improve EFMAC, TCD-MAC, and
Yen-Cheng Chiu*1, Win-San Khwa*2, Chung-Yuan Li1, Fang-Ling Hsieh1,