CAM Memory Strings for In-Memory Approximate Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing non-volatile ternary content addressable memory (TCAM) devices struggle to distinguish between all-match and 1-bit-mismatch states, limiting their ability to perform in-memory approximate searching.
Innovation Solution
A memory apparatus and method that utilize a plurality of memory strings with CAM cells and a sensing amplifier circuit to generate string currents based on the number of mismatched bits between search and storage data, allowing for the determination of all-matched, partially matched, or all-mismatched degrees, enabling in-memory approximate searching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by stationary object
If non-volatile RRAM-based TCAM is used to reduce cell area and standby power consumption, then memory density and power efficiency are improved, but the ability to distinguish all-match state and 1-bit-mismatch state deteriorates
Solution Approach 1:
The patent segments the match evaluation into multiple levels by introducing intermediate sensing stages. The sensing amplifier circuit performs preliminary detection to identify clearly matched or mismatched strings, then only ambiguous cases proceed to more precise but power-intensive verification stages, enabling gradient-based power consumption management.
Solution Approach 2:
The patent dynamically adjusts sensing parameters including voltage levels and sensing thresholds based on the matching degree. By changing the sensing voltage amplitude and threshold according to whether a string is likely matched or mismatched, the system optimizes the balance between detection precision and power consumption for different match states.
2Productivity
If conventional TCAM based on SRAM is used, then search capability and parallelism are improved, but memory density and access power consumption deteriorate
Solution Approach 1:
The patent replaces the traditional SRAM-based TCAM mechanical/electrical structure with a resistive memory (RRAM) based system. This substitution leverages the high-density characteristics of resistive memory while maintaining TCAM search functionality, achieving both high memory density and acceptable power consumption through the inherent properties of resistive switching devices.
Solution Approach 2:
The patent employs periodic sensing operations with varying voltage amplitudes. The sensing process is divided into multiple phases with different voltage levels, where initial low-voltage sensing quickly identifies obvious mismatches, followed by higher-voltage sensing only for potentially matched strings, creating a periodic action pattern that reduces overall power consumption while maintaining search parallelism.
3Use of energy by stationary object
If existing non-volatile TCAM is used, then power consumption is reduced through dense memory density, but the ability to implement in-memory approximate searching deteriorates
Solution Approach 1:
The patent introduces dynamic sensing thresholds and voltage levels that adapt based on the search requirements. The sensing circuit can adjust its operating parameters in real-time to distinguish between exact matches, approximate matches, and mismatches, enabling the system to perform in-memory approximate searching while maintaining low power consumption through adaptive parameter adjustment.
Solution Approach 2:
The patent implements a feedback mechanism where the sensing results from preliminary detection phases inform subsequent verification phases. The output from initial low-power sensing feeds back to control the voltage amplitude and sensing duration of follow-up operations, creating a feedback loop that optimizes power consumption based on actual match probability while enabling approximate search functionality.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables efficient in-memory approximate searching by accurately differentiating match degrees through string currents, improving search operations and reducing power consumption while maintaining high memory density.
Implementation Method 1
a string voltage is applied to the plurality of memory strings and a search data is input for comparing with a storage data stored in the plurality of CAM cells of the plurality of memory strings. On each of the plurality of memory strings, the string voltage and the string resistance corresponding to a number of mismatched bit(s) between the search data and the storage data enables a string current
Data Source
AI summary
The application provides a content addressable memory (CAM) device and a method for searching and comparing data thereof. The CAM device comprises: a plurality of memory strings; and a sensing amplifier circuit coupled to the memory strings; wherein in data searching, a search data is compared with a storage data stored in the memory strings, the memory strings generate a plurality of string currents, the sensing amplifier circuit senses the string currents to generate a plurality of sensing results; based on the sensing results, a match degree between the search data and the storage data is determined as one of the follows: all-matched, partially-matched and all-mismatched.