08/26/2024, 12:06 AM UTC
➀ 北京大学的中国科学家开发了业界首个使用碳纳米管晶体管的张量处理单元(TPU)。➁ 该TPU具有99.9999%的半导体纯度,在MNIST图像识别中可达到高达88%的准确率,仅消耗295 µW的功率。➂ 研究人员认为,通过微调GAA FET的对齐和增加处理单元的位容量,可以进一步提高芯片的效率和计算能力。➀ Chinese scientists from Peking University have developed the industry's first tensor processing unit (TPU) using carbon nanotube transistors. ➁ The TPU features a 99.9999% semiconductor purity and can achieve up to 88% accuracy in MNIST image recognition while consuming only 295 µW of power. ➂ The researchers believe there is potential for further improvement by fine-tuning the alignment of GAA FETs and increasing the processing unit's bit capacity.
---
本文由大语言模型(LLM)生成,旨在为读者提供半导体新闻内容的知识扩展(Beta)。