The reporter learned from the official website of Peking University that recently, the university's research team has made significant breakthroughs in the field of next-generation chip technology, successfully developing the world's first tensor processor chip based on carbon nanotubes, which can achieve high-efficiency convolutional neural network operations. With the rapid development of artificial intelligence technology, especially the rise of large model applications such as ChatGPT, the future world's data will experience explosive growth, and the processing of massive amounts of data poses severe challenges to the computing power and energy efficiency of chips. However, the development of high-efficiency computing chips is facing two major bottlenecks: chip architecture and transistor performance. The traditional von Neumann architecture can no longer meet the high-speed and high bandwidth data handling and processing requirements. Future high-efficiency computing chips must innovate in hardware architecture to be suitable for tensor data operations in models such as neural networks. At the same time, the silicon-based complementary metal oxide semiconductor transistors used in chip construction are also facing challenges of size reduction and significant power consumption increase. There is an urgent need to develop ultra-thin and high carrier mobility semiconductors as channel materials, with the expectation of constructing transistors with better reducibility and higher performance than silicon-based CMOS transistors. Carbon nanotubes have excellent electrical properties and ultra-thin structures, and carbon nanotube transistors have demonstrated performance and power consumption potential beyond commercial silicon-based transistors. Therefore, they are expected to become the main device technology for building future high-performance computing chips. Only by achieving breakthroughs in both system architecture and underlying transistors can we maximize the computing power and energy efficiency of chips. At present, mature silicon-based device technology for computing chips mainly relies on architectural innovation, while research on new material electronic devices mainly focuses on improving the performance of transistors, and there is no research work combining the two. A research team from Peking University has successfully developed the world's first carbon nanotube based tensor processor chip based on the novel device technology of carbon nanotube transistors, combined with an efficient pulsating array architecture design. The chip adopts a new device technology and pulsating array architecture, integrating 3000 carbon nanotube transistors into a tensor processor chip, pushing carbon based electronics from device research to system demonstration, significantly improving the computational efficiency of convolutional neural networks, with extremely low power consumption and an accuracy rate of 88%. In addition, carbon based transistors exhibit comprehensive advantages such as better speed and power consumption than silicon-based CMOS transistors. Carbon based tensor processors have a threefold performance advantage at the 180 nanometer technology node and have the potential to continue to advanced technology nodes, which is expected to meet the demand for high-performance and high-efficiency chips in the era of artificial intelligence. (New Society)
Edit：Xiong Dafei Responsible editor：Li Xiang
Source：XinHuaNet
Special statement: if the pictures and texts reproduced or quoted on this site infringe your legitimate rights and interests, please contact this site, and this site will correct and delete them in time. For copyright issues and website cooperation, please contact through outlook new era email：lwxsd@liaowanghn.com
Sci-Tech
Recommended Reading Change it 
Home
