Search Memory Architecture Reducing Chip Area and Power
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing number of CAM arrays in search memory devices leads to a rise in chip area and current consumption, particularly when executing parallel search processes with multiple search keys, as all search keys need to be transmitted across a wide data bus.
Innovation Solution
A search memory configuration that includes an input control section and N search units, each with multiple selectors, allowing for the allocation of two types of entry data in search blocks, reducing the need for all search keys to be inputted across the data bus, thereby decreasing chip area and current consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple search keys are transmitted across the data bus to execute parallel search processes, then search productivity is improved, but the data bus width increases leading to larger chip area and higher current consumption
Solution Approach 1:
The patent divides the search memory into multiple search units, each handling a subset of search keys. Instead of transmitting all N search keys across the entire data bus, each search unit receives only 2 search keys through a narrower local data bus. This segmentation of the search memory system allows parallel search operations to proceed while significantly reducing the width of the data bus required, thereby reducing chip area without compromising search productivity.
2Productivity
If multiple search keys are transmitted across the data bus to execute parallel search processes, then search productivity is improved, but current consumption increases
Solution Approach 1:
By segmenting the search memory into multiple independent search units, each unit processes a limited number of search keys (2 keys per unit) through dedicated narrower data buses. This reduces the total current consumption compared to a single large data bus transmitting all N search keys, while still enabling parallel search operations across multiple units to maintain high productivity.
3Area of stationary object
If the data bus width is reduced to decrease chip area, then chip area and current consumption are reduced, but the ability to transmit multiple search keys for parallel search is limited
Solution Approach 1:
The patent resolves this contradiction by dividing the search memory into multiple search units (N units), where each unit has its own narrow data bus for receiving 2 search keys. This segmentation allows the system to use narrow data buses (reducing chip area and current consumption) while maintaining parallel search capability across multiple units, thus preserving search productivity.
Solution Approach 2:
Multiple search units are combined to work in parallel, each handling 2 search keys through narrow data buses. The collective capability of N search units processing 2 keys each equals the capability of a single wide data bus transmitting N keys, thereby achieving the same productivity with reduced data bus width and lower power consumption.
Data Source
AI summary
Provided is a search memory capable of suppressing an increase in the area of a chip and an increase in the amount of current consumption. The search memory includes an input control section, N search units, and N first selectors. The input control section receives N (N: two or more) search data that are parallelly inputted. The N first selectors are respectively disposed in association with the N search units to select two search data from the N search data. The search units each include multiple search blocks and M second selectors. The M second selectors select one of two search data selected by the first selectors. For at least one of the N search units, allocation is performed so that two types out of N types of entry data respectively associated with the N search data are respectively stored in the search blocks.


