A research team hailing from Peking University’s School of Integrated Circuits and the Institute for Artificial Intelligence has achieved a remarkable feat by successfully developing a high-precision, scalable analog matrix computing chip. This innovative chip is built upon resistive random-access memory (RRAM) technology. It stands as the first of its kind to elevate the precision of analog computing to a 24-bit fixed-point accuracy level. This level of precision is on par with that of digital computing and represents a staggering five-order-of-magnitude improvement over traditional analog computing precision.

When tackling crucial scientific challenges, such as large-scale Multiple Input Multiple Output (MIMO) signal detection, this chip demonstrates exceptional performance. It attains a computing throughput and energy efficiency that is 100 to 1000 times greater than that of current state-of-the-art digital processors, including Graphics Processing Units (GPUs). This significant achievement paves the way for a transformation in computing paradigms in the post-Moore era. Moreover, it offers promising solutions to address the computational demands in cutting-edge fields like artificial intelligence and 6G communications.