Improving Energy Consumption in Content-Addressable Memory Through Precomputation

Content-addressable memory (CAM) is a storage medium used for high-speed searching of data due to its parallel search capability. Instead of retrieving data based on designated address as in RAM, the user inputs a search item and the data's address is retrieved as a result if it exists. However, the high power consumption associated with parallel search is a significant challenge for CAM memories. In this paper, a new technique is proposed to reduce power consumption in CAM memory while maintaining an acceptable search speed. The technique involves precomputing the number of contiguous ones in an input and comparing it to the same calculations of the stored data, thereby reducing the number of necessary comparisons and resulting in power reduction. The experiment results show that the proposed design outperforms other approaches with an energy consumption of 0.119 fj/bit/search and the cycle delay time of 0.67 ns.