芯粒技术在人工智能芯片设计中的应用探索  

Exploration on the Application of Core Particle Technology in Artificial Intelligence Chip Design

作　　者：张衡 袁杰 ZHANG Heng;YUAN Jie(Shenzhen Intellifusion Technologies Co.,Ltd.,Shenzhen Guangdong 518057,China)

机构地区：[1]深圳云天励飞技术股份有限公司,广东深圳518057

出　　处：《信息与电脑》2024年第7期37-39,共3页Information & Computer

摘　　要：随着人工智能技术的飞速发展,人工智能(Artificial Intelligence,AI)芯片的需求呈指数级增长。为了满足AI芯片的高算力需求,基于芯粒技术(Chiplet)的片上系统(System on Chip,SoC)设计方法逐渐受到关注。Chiplet技术可以灵活地将多种功能模块进行组合,从而实现不同的芯片系统。基于Chiplet技术的AI芯片设计可以带来更高的算力、更好的灵活性和更低的功耗。本文详细探讨基于Chiplet技术的人工智能芯片设计方法,并分析其技术特点、设计流程以及面临的挑战。With the rapid development of artificial intelligence technology,the demand for artificial intelligence chips has been growing exponentially.In order to meet the high computational requirements of AI chips,the design methods of System on Chip based on Chiplet technology is gradually gaining attention.Chiplet technology can flexibly combine multiple functional modules to realize different chip systems.The design of AI chip based on Chiplet technology can lead to higher computing performance,better flexibility,and lower power consumption.In this paper,the design methods of artificial intelligence chip based on Chiplet technology is discussed in detail,and its technical characteristics,design process and challenges are analyzed.

关 键 词：Chiplet技术 AI芯片 人工智能 SOC设计 

分 类 号：TM76[电气工程—电力系统及自动化]