Memory Controller Row Hammer Targeted Refresh
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-density memory devices experience intermittent failures due to 'row hammer' issues, where repeated access to a single row within a memory cell's refresh window causes data corruption in adjacent rows, particularly in DRAM systems, leading to increased risk of data loss and reduced storage reliability.
Innovation Solution
A memory controller is designed to detect row hammer events by monitoring access counts within a specified time frame and sends targeted refresh commands to adjacent rows to mitigate data corruption, using either address map information or default offset schemes to identify and refresh potential victim rows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If higher density memory devices are used to achieve smaller size and increased storage capacity, then device integration is improved, but row hammer failures occur due to repeated access to single rows within refresh window
Solution Approach 1:
The system performs preliminary detection of row hammer conditions by monitoring access counts to specific rows within refresh windows. When a row is accessed excessively (e.g., 550K times or more in 64ms), the system proactively identifies the physically adjacent wordlines as victims and performs preventive refresh operations before data corruption occurs, rather than waiting for failures to manifest
Solution Approach 2:
The patent introduces an intermediary detection and control mechanism between the memory controller and memory device. This intermediary system monitors access patterns, identifies row hammer conditions, and triggers targeted refresh operations on victim rows, acting as a mediator that prevents the harmful interaction between repeated accesses and adjacent row corruption
2Reliability
If targeted refresh operations are performed on victim rows to prevent data corruption, then reliability is improved, but additional refresh commands and processing overhead are introduced
Solution Approach 1:
Instead of performing blanket refresh operations across entire memory banks, the system applies local quality by targeting only the specific victim rows adjacent to the hammered row. This localized approach refreshes only the rows that are at risk of corruption while leaving other rows unaffected, reducing unnecessary refresh overhead and simplifying control logic compared to global refresh strategies
Data Source
AI summary
A memory controller to implement targeted refreshes of potential victim rows of a row hammer event. In an embodiment, the memory controller receives an indication that a specific row of a memory device is experiencing repeated accesses which threaten the integrity of data in one or more victim rows physically adjacent to the specific row. The memory controller accesses default offset information in the absence of address map information which specifies an offset between physically adjacent rows of the memory device. In another embodiment, the memory controller determines addresses for potential victim rows based on the default offset information. In response to the received indication of the row hammer event, the memory controller sends for each of the determined plurality of addresses a respective command to the memory device, where the commands are for the memory device to perform targeted refreshes of potential victim rows.