中国新型芯片突破光刻机限制，北大团队研发高精度模拟矩阵计算芯片

Core Insights - A new type of chip developed by Professor Sun Zhong's team at Peking University utilizes resistive random-access memory (ReRAM) for high-precision, scalable analog matrix computing, effectively bypassing the limitations of photolithography machines [1] - This chip addresses the previous challenges of analog computing, which was deemed inaccurate, by significantly improving speed and energy efficiency while being manufacturable using mature 28-nanometer technology [1] - Currently in the experimental phase, the chip is designed for medium-scale tasks and is expected to be applied in fields such as 6G and robotics, with plans to scale up production within two years [1] - In the long term, this technology serves as a high-efficiency backup solution for AI model training and supercomputing, positioning the industry to meet future demands as they arise [1] Summary by Categories Technology Development - The chip is based on resistive random-access memory, allowing for high-precision and scalable analog matrix calculations [1] - It operates effectively without the need for advanced photolithography, utilizing a 28-nanometer process [1] Applications - Initially suitable for medium-scale tasks, the chip is targeted for use in 6G and robotics [1] - Future applications include AI model training and supercomputing, addressing the needs of high-performance computing [1] Market Positioning - The development of this chip positions the industry to leverage existing technology in response to the anticipated surge in demand for computational power [1]