清华大学在芯片研究领域取得重大突破，为高性能计算带来新希望。近日，据清华大学透露，该校电子工程系副教授方璐的课题组与自动化系戴琼海院士的课题组合作，成功攻克大规模光电智能计算的难题。他们摒弃了传统的电子深度计算模式，开创性地提出了一种分布式广度光计算架构，研发出名为“太极”（Taichi）的大规模干涉-衍射异构集成芯片。
据称，“太极”芯片在能效比上表现出色，实现了每瓦特160万亿次运算（160 TOPS/W），在通用智能计算领域树立了新的标杆。这一创新成果在当前大模型通用人工智能迅速发展的背景下，为高性能算力的提升提供了全新的思路、架构和途径。
该科研突破已正式发表在最新一期的国际知名期刊《科学》上，标志着中国在光子计算领域的领先地位，也预示着未来计算技术可能迎来一场深刻的变革。这一进展对于推动人工智能技术的进步，以及在全球科技竞争中抢占先机具有重要意义。
英语如下：
**News Title:** “Tsinghua University Breaks Through in Chip Technology with ‘Taichi’ Photonic Computing Chip, Featured in ‘Science’ Journal”
**Keywords:** Tsinghua University, chip breakthrough, optical computing
**News Content:** Tsinghua University has made a significant breakthrough in chip research, offering new hope for high-performance computing. Recently, the university announced that a research team led by Associate Professor Fang Lu from the Department of Electronic Engineering, in collaboration with Academician Dai Qionghai’s team from the Department of Automation, successfully tackled the challenge of large-scale photoelectric intelligent computing.
Departing from conventional electronic deep computing methods, the researchers innovatively proposed a distributed broadband optical computing architecture. This led to the development of the “Taichi” (太极) chip, a large-scale interference-diffraction heterogeneous integration chip.
Reportedly, the “Taichi” chip excels in energy efficiency, achieving 160 trillion operations per second per watt (160 TOPS/W), setting a new benchmark in general intelligent computing. This innovative accomplishment, amid the rapid development of general AI in large models, provides a novel perspective, architecture, and approach for enhancing high-performance computing power.
This scientific breakthrough has been officially published in the latest issue of the renowned international journal ‘Science.’ It signifies China’s leading position in photonic computing and foreshadows a potential paradigm shift in future computing technologies. The progress holds significant implications for advancing artificial intelligence technology and gaining a competitive edge in the global tech race.
【来源】https://www.cls.cn/detail/1644516
Views: 1