Distributed Targeted Refresh for Memory Power and Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In memory devices, repeated access to certain memory cells (aggressor rows) causes increased data degradation in nearby cells (victim rows) due to electromagnetic coupling, leading to potential data loss if not refreshed promptly, and existing refresh methods do not efficiently manage power consumption during targeted refresh operations.
Innovation Solution
Implementing a distributed targeted refresh operation system that identifies aggressor rows and schedules targeted refreshes among auto-refresh operations to evenly distribute power load, using a refresh control circuit to determine the type and timing of refresh operations, including both auto-refresh and multiple types of targeted refreshes, to prevent data loss and optimize power usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If targeted refresh operations are performed frequently to address row hammer effects, then data reliability is improved, but power consumption increases
Solution Approach 1:
The system performs preliminary identification of aggressor rows through monitoring mechanisms that detect repeated access patterns before significant data degradation occurs. By proactively identifying which rows require targeted refresh operations, the system can schedule refreshes in advance, preventing data loss while avoiding unnecessary refresh operations that would consume excessive power.
Solution Approach 2:
The refresh strategy dynamically adapts between auto-refresh and targeted refresh operations based on real-time monitoring of row access patterns. The system adjusts the refresh approach according to actual usage conditions, switching from conventional periodic auto-refresh to targeted refresh only when row hammer effects are detected, thereby optimizing the balance between data reliability and power consumption.
2Reliability
If multiple targeted refresh operations are scheduled within a pump cycle, then data degradation is reduced, but power load distribution becomes uneven
Solution Approach 1:
The pump cycle is segmented into multiple discrete refresh operations that can be independently scheduled. Each targeted refresh operation is separated by at least one auto-refresh operation, creating an interleaved pattern that distributes power load more evenly across the cycle. This segmentation prevents concentration of power consumption in a single time window while maintaining comprehensive coverage of all victim rows.
3Reliability
If targeted refresh operations are performed on all victim rows, then data loss is prevented, but refresh operation complexity increases
Solution Approach 1:
The refresh operation applies local quality by treating different rows differently based on their specific characteristics and risk levels. Instead of uniformly applying the same refresh strategy to all rows, the system identifies aggressor rows that generate row hammer effects and applies targeted refresh only to their associated victim rows. This localized approach reduces overall complexity by focusing refresh operations only where needed rather than treating all rows equally.
Data Source
AI summary
Embodiments of the disclosure are drawn to apparatuses and methods for distributed timing of targeted refresh operations. Information stored in volatile memory cells may decay unless refresh operations are performed. A memory device may perform auto-refresh operations, as well as one or more types of targeted refresh operations, where particular rows are targeted for a refresh. Targeted refresh operations may draw less power than an auto-refresh operation. It may be desirable to distribute targeted refresh operations throughout a sequence of refresh operations, to average out a power draw in the memory device. Responsive to an activation of a refresh signal, the memory device may perform a group of refresh operations. At least one refresh operation in each group may be a targeted refresh operation.