静态随机存取存储器
计算机科学
电容器
利用
电气工程
Boosting(机器学习)
互连
电压
电子工程
工程类
电信
人工智能
计算机安全
作者
Rajiv Joshi,Matthew M. Ziegler
标识
DOI:10.23919/mixdes.2019.8787172
摘要
This paper reviews key developments and the continuation of low power techniques needed from the Moore to AI eras. SRAM with a wider range of operation, from extreme low to high voltages, is enabled using novel circuit techniques and demonstrated for edge or data centric accelerators. These techniques exploit interconnect as well as inductor and capacitor coupling for boosting on demand. Several chips fabricated in 14nm SOI technology show functional 8T SRAM down to 0.24V-0.30V These new techniques can lead to lower voltage operation of cognitive and neural network for IoT and data centric applications.
科研通智能强力驱动
Strongly Powered by AbleSci AI