中国科学家利用1纳米铁电材料制造了一种超效率芯片,使装置能够快速、低功率的AI处理。 Chinese scientists created a ultra-efficient chip using a 1-nm ferroelectric material, enabling fast, low-power AI processing on devices.

通过一种新型铁电材料, 开发出一种高性能内存计算芯片, 制造出仅1纳米厚的超薄, 均的膜, Chinese scientists at Peking University have developed a high-performance computing-in-memory chip using a novel ferroelectric material, creating ultrathin, uniform films just 1 nanometer thick that maintain strong memory properties. 该装置在0.8伏特运行,以20毫秒的速度写出数据,在1.5万亿重写周期中存活下来,超过了现行标准。 The device operates at 0.8 volts, writes data in 20 nanoseconds, and survives over 1.5 trillion rewrite cycles—exceeding current standards. 通过将内存和处理整合为一个单元,可减少能源使用并加快计算速度,与现有半导体制造的兼容性。 By integrating memory and processing in one unit, it reduces energy use and speeds up computation, with compatibility for existing semiconductor manufacturing. 这一突破发表在《科学》上,可以使智能手机和电器等日常设备能够进行强大的人工智能处理,从而减少对云计算的依赖。 The breakthrough, published in Science, could enable powerful AI processing on everyday devices like smartphones and appliances, cutting reliance on cloud computing.