Oscillator-Controlled Random Sampling for Row Hammer Repair
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current memory systems face challenges in effectively addressing 'row hammering' effects, where data in adjacent memory rows degrade due to increased cell-to-cell leakage from frequent access, as existing techniques struggle to maintain accurate refresh signals at high operating speeds.
Innovation Solution
Implementing an oscillator-controlled random sampling method to select refresh addresses independently of the memory system's timing, ensuring equal probability for any row to be chosen for row hammer repair, thereby reducing data degradation in adjacent rows by frequent refreshes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If sampling of activated rows is employed to identify rows for refresh, then the refresh process can be automated, but the sampled activation row is always tied to the refresh interval which increases the probability of missed activations
Solution Approach 1:
The patent introduces dynamic randomization by using an oscillator to generate unpredictable time intervals between sampling operations. Instead of fixed periodic sampling tied to refresh intervals, the sampling timing becomes dynamic and adaptive, allowing the system to capture row activations that occur at unpredictable times, thereby reducing missed activations while maintaining automation.
2Measurement precision
If tracking and recording the number of times every row is accessed is implemented to identify hammered rows, then accurate identification of rows needing refresh can be achieved, but the circuitry footprint becomes relatively large and maintaining counts is cumbersome
Solution Approach 1:
The patent extracts only the essential information needed for row hammer detection by sampling activation rows at random intervals rather than tracking complete access counts for all rows. This selective sampling approach removes the need for complex counting circuits and large storage structures, significantly reducing circuitry footprint while maintaining the ability to identify rows that require refresh.
Solution Approach 2:
Instead of maintaining complete access count records for every row, the patent uses random sampling to create a representative subset of activation events. This sampling copy provides sufficient information to identify hammered rows without requiring full tracking of all row accesses, thereby reducing the complexity and footprint of the monitoring circuitry.
3Reliability
If rows adjacent the accessed rows are sent a refresh signal based on tracked access counts, then data degradation in adjacent rows can be prevented, but the process becomes cumbersome and difficult to implement at high operating speeds
Solution Approach 1:
The patent implements periodic random sampling of row activations using an oscillator that generates sampling pulses at irregular intervals. This periodic yet randomized approach allows the system to continuously monitor for row hammer conditions without requiring complex real-time analysis, making it easier to implement at high operating speeds while still effectively preventing data degradation in adjacent rows through timely refresh signals.
Data Source
AI summary
Various embodiments comprise methods and apparatuses for selecting a randomly-chosen seed row from among a stream of available data in a memory system. A refresh operation is then performed on at least one selected row of memory in the memory system based on the randomly-chosen seed row. Additional apparatuses and methods are described.