DRAM Row Access Counters Mitigate Rowhammer Data Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Volatile memory devices, such as DRAM, face issues with data loss due to the leakage of charges stored in capacitors, particularly exacerbated by the 'rowhammer' effect where intensive access to one row can corrupt adjacent rows, necessitating effective monitoring and refresh mechanisms.
Innovation Solution
A memory device incorporating counters, first flags, a queue, and a refresh control circuit to monitor and manage access counts for each row, determine whether to queue incoming row addresses based on access counts and flag values, and perform targeted refresh operations to mitigate rowhammer effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If intensive access to a certain row is performed, then access speed is improved, but data stored in memory cells of adjacent rows may be lost due to rowhammer effect
Solution Approach 1:
The patent performs preliminary actions by monitoring access patterns and proactively identifying potential victim rows before actual data loss occurs. The refresh control circuit continuously tracks access counts and flag values, and preemptively refreshes adjacent rows that are at risk of being affected by rowhammer attacks, thus preventing data loss before it happens.
Solution Approach 2:
The patent implements feedback mechanisms through counters that track access frequencies and flags that indicate refresh status. The refresh control circuit uses this feedback information to dynamically adjust refresh operations, identifying rows that require refresh based on their access patterns and the status of adjacent rows, thereby responding to changing conditions in real-time.
2Reliability
If all rows are monitored for rowhammer attacks, then data integrity is improved, but device complexity increases due to additional counters, flags, and control circuits
Solution Approach 1:
The patent divides the monitoring task into manageable segments by associating individual counters and flags with specific rows or groups of rows. Rather than implementing a monolithic monitoring system, each row has its own counter and flag, allowing the refresh control circuit to process and manage refresh operations in a distributed, modular manner, thus reducing overall system complexity.
Solution Approach 2:
The patent enables self-service by allowing the monitoring system to automatically identify and refresh victim rows without external intervention. The counters and flags automatically track access patterns and refresh status, and the refresh control circuit autonomously determines which rows need refreshing based on this information, eliminating the need for complex external control mechanisms.
3Reliability
If frequent refresh operations are performed on all rows, then data integrity is maintained, but productivity decreases due to increased refresh overhead
Solution Approach 1:
The patent applies local quality by performing refresh operations selectively on specific rows that are identified as potential victims of rowhammer attacks, rather than uniformly refreshing all rows. The refresh control circuit uses counter and flag information to determine which adjacent rows need refreshing based on their proximity to aggressively accessed rows, thus concentrating refresh efforts where they are most needed and minimizing unnecessary refresh operations.
Solution Approach 2:
The patent employs partial action by performing refresh operations only on the subset of rows that are at risk, rather than executing full refresh cycles on all rows. This selective approach refreshes only the necessary adjacent rows identified through monitoring, reducing the total number of refresh operations and minimizing the impact on productivity while still maintaining data integrity for vulnerable rows.
Data Source
AI summary
A memory device may include counters respectively corresponding to rows and each configured to count a number of accesses to a corresponding row, a refresh control circuit, a queue, and first flags respectively corresponding to the rows. The refresh control circuit may change a second flag set in a refresh period every refresh period, and determine whether to put an incoming row address into the queue based on a count value of a counter corresponding to a target row indicated by the incoming row address among the counters, a first flag value of a first flag corresponding to the target row among the first flags, and a second flag value of the second flag set in a current refresh period.