Memory Device Victim Row Determination for Row-Hammer Mitigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As memory devices, such as DRAM, experience increased access frequencies, memory cells adjacent to frequently accessed cells can degrade, leading to reduced data reliability due to the row-hammer phenomenon.
Innovation Solution
A memory device with a control logic system that determines a victim row address based on the health level and access frequency of each row, and performs a targeted refresh operation on the identified victim row to mitigate degradation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the access frequency of memory cells is increased, then the productivity of the memory device is improved, but the memory cells adjacent to frequently accessed cells degrade due to the row-hammer phenomenon
Solution Approach 1:
The patent performs preliminary actions by counting error occurrences in advance through ECC operations and calculating victim degrees before actual data damage occurs. The system proactively identifies rows that will become victims of row-hammer attacks by monitoring error patterns and access frequencies, then refreshes these rows before they suffer degradation from intensive adjacent row accesses.
Solution Approach 2:
The patent implements feedback mechanisms by continuously monitoring error occurrence signals from ECC operations and using this information to dynamically adjust refresh operations. The system feeds back error count information to the control logic, which then determines victim rows based on accumulated error data and access patterns, creating a closed-loop system that adapts to actual memory degradation trends.
2Reliability
If a targeted refresh operation is performed on identified victim rows, then the data reliability is improved, but the device complexity increases due to the need for health level determination and victim row identification systems
Solution Approach 1:
The patent applies multi-functionality by having the ECC engine serve dual purposes: its primary function of error correction and an additional function of health level monitoring. The same ECC decoding circuit that corrects errors also generates error occurrence signals used for victim row identification. This eliminates the need for separate monitoring hardware, reducing overall system complexity while maintaining enhanced reliability capabilities.
Solution Approach 2:
The system performs self-service by using its own internal error correction mechanisms to generate the information needed for victim row identification. The ECC operations that are already necessary for data integrity automatically provide error occurrence data that the control logic uses to determine which rows need refreshing, eliminating the need for external or separate monitoring systems.
3Measurement precision
If the number of error corrections for each row is monitored to determine health levels, then the measurement precision of row condition is improved, but the loss of time increases due to additional monitoring operations
Solution Approach 1:
The patent merges the health monitoring function with the existing ECC operation timeline. Error counting and health level determination are performed during the same time window as regular ECC operations, rather than as separate additional operations. The control logic accumulates error occurrence signals that are generated during normal read/write operations, combining multiple functions into a unified operational flow that minimizes time overhead.
Data Source
AI summary
A memory device includes a memory cell array including a plurality of rows, an ECC engine configured to determine a health level for each of the plurality of rows based on the number of corrections of errors of data read from each of the plurality of rows, a control logic configured to determine a victim row address based on the health level and the number of accesses for each of the plurality of rows, and a refresh control circuit configured to perform a refresh on a row corresponding to the determined victim row address.