15.3 A 351TOPS/W and 372.4GOPS Compute-in-Memory SRAM Macro in 7nm FinFET CMOS for Machine-Learning Applications.
Qing DongMahmut E. SinangilBurak ErbagciDar SunWin-San KhwaHung-Jen LiaoYih WangJonathan ChangPublished in: ISSCC (2020)
Keyphrases
- random access memory
- machine learning
- cmos technology
- power consumption
- nm technology
- low voltage
- embedded dram
- low power
- power dissipation
- dynamic random access memory
- design considerations
- low cost
- silicon on insulator
- metal oxide semiconductor
- pattern recognition
- leakage current
- machine learning methods
- active learning
- data mining
- text mining
- learning algorithm
- artificial intelligence
- computer vision
- memory access
- machine learning algorithms
- memory requirements
- knowledge acquisition
- main memory
- supervised learning
- high speed
- feature selection
- power reduction
- chip design
- natural language processing
- memory usage
- text classification
- bit rate
- power management
- data analysis
- parallel processing
- memory space