HKU chip performs search task 100 million times faster than standard CPU

(ECNS) -- Researchers at The University of Hong Kong (HKU) have developed a minimalist analogue content-addressable memory (CAM) chip that performs complex searches directly where data are stored, the university said Tuesday.

When tasked with calculating the analogue Hamming distance for machine learning classification, the new analogue CAM performed the task approximately 108 (100 million) times faster than a standard CPU, scoring exceptionally high accuracy across multiple datasets, according to the university.

The chip is designed to reduce the slow and energy-intensive transfer of data between memory and processors, known as the “von Neumann bottleneck.” Its architecture can compare incoming search data with all stored data simultaneously in a single step.

Led by professor Li Can from the Department of Electrical and Computer Engineering of the Faculty of Engineering and the Centre for Advanced Semiconductors and Integrated Circuits (CASIC), the team built each analogue CAM cell with just two transistors, compared with up to 16 in a conventional digital CAM cell. The design reduces chip area and allows continuous analogue signals to be processed directly.

Li said the team is exploring how the technology could support search mechanisms in large AI models. The findings were published in Nature Nanotechnology.

(By Intern Yang Hongran, Zhang Dongfang)