This time, groundbreaking news came from China in the world of science and technology. China has developed the world’s first carbon nanotube-based tensor processor chip (TPU). The team led by Peng Lianmao and Zhang Zhiyong from the Carbon-Based Electronics Research Center of Peking University Faculty of Electronics achieved this great success.
The world’s first carbon nanotube tensor processor chip was developed at Peking University
This chip contains 3000 carbon nanotube field effect transistors. So, imagine, small but incredibly powerful. These transistors can perform convolution operations and matrix multiplication with high efficiency. This could be a revolutionary development in fields such as artificial intelligence and image processing.
Using new device technology and pulsation array architecture, parallel 2-bit integer multiplication and addition operations can be performed. Experiments showed that this chip achieved 88% MNIST image recognition accuracy with a low power consumption of 295μW. So how was it developed?
By optimizing the carbon nanotube production process, the research team was able to obtain semiconductor materials with 99.9999% purity and ultra-clean surfaces. In this way, they were able to produce transistors with high current density and homogeneity.
Simulation results show that 8-bit carbon nanotube TPU produced at 180 nanometer process node can reach 850 MHz main frequency and trillion operations per watt (TOPS) level. These data are very important in terms of performance.
This technology, which can meet the demands for high-performance and energy-efficient chips in the age of artificial intelligence, can truly revolutionize the chip world. These carbon nanotube-based chips will make a big difference, especially in applications requiring artificial intelligence and high performance.
This study was also published in the journal Nature Electronics on July 22. For more detailed information, you can read the article here. What are you thinking? You can share your opinions in the comments section below.
{{user}} {{datetime}}
{{text}}