DRAM Refresh Rate Adjustment for Row Hammer Mitigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing DRAM systems lack versatility in refresh operations, as they typically refresh all memory rows at a single rate when accessed a certain number of times, leading to inefficiencies and potential data flip issues due to the row hammer effect, which can cause abnormal functioning.
Innovation Solution
The proposed DRAM system includes a refresh unit with an accessing device and multiple refresh devices that adjust refresh rates based on access circumstances, such as identifying and refreshing neighbor memory rows at different rates when accessed, to mitigate the row hammer effect and improve efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If all memory rows are refreshed at a single rate when accessed a certain number of times, then the refresh operation is simple to implement, but it leads to inefficiencies and potential data flip issues due to the row hammer effect
Solution Approach 1:
The patent segments the memory rows into different categories (accessed rows and unaccessed rows) and applies different refresh rates to each category. This segmentation allows the system to treat rows differently based on their access patterns, thereby preventing the row hammer effect on unaccessed rows while maintaining efficiency on accessed rows.
Solution Approach 2:
The patent applies local quality by assigning different refresh rates to different memory rows based on their specific access characteristics. Unaccessed rows that are susceptible to row hammer effects receive higher refresh rates, while accessed rows receive standard refresh rates, optimizing both reliability and efficiency for each local region.
2Reliability
If refresh rates are adjusted based on access circumstances, then data flip errors are reduced, but the system complexity increases
Solution Approach 1:
The patent implements dynamic refresh rate adjustment based on real-time access circumstances. The refresh rate for each memory row is not fixed but dynamically changed according to whether the row has been accessed or not, allowing the system to adapt to varying access patterns and mitigate row hammer effects effectively.
Solution Approach 2:
The patent employs feedback mechanisms where the refresh control logic monitors access patterns of memory rows and adjusts refresh rates accordingly. This feedback loop enables the system to detect when unaccessed rows are being repeatedly accessed and increase their refresh rates to prevent data flip errors, thereby maintaining reliability without excessive complexity.
3Reliability
If neighbor memory rows are refreshed at higher rates, then row hammer effects are mitigated, but power consumption increases
Solution Approach 1:
The patent applies local quality by selectively increasing refresh rates only for neighbor memory rows that are adjacent to accessed rows and susceptible to row hammer effects, rather than increasing refresh rates for all memory rows. This localized approach mitigates row hammer effects while minimizing the overall power consumption impact.
Solution Approach 2:
The patent changes the refresh rate parameter dynamically based on the access patterns of neighboring rows. When a row is accessed, the refresh rate of its neighbor rows is increased temporarily to prevent row hammer effects, and this parameter change is applied selectively rather than globally, thereby balancing reliability improvement with power consumption management.
Data Source
AI summary
The present disclosure provides a dynamic random access memory (DRAM). The DRAM includes a refresh unit, an accessing device and a refresh device. The refresh unit has a plurality of memory rows. The accessing device is configured to access the memory rows. The refresh device is configured to refresh the refresh unit in a first manner in response to a first event, in which a quantity of accessed memory rows of the refresh unit is not greater than a threshold quantity. The refresh device is configured to refresh the refresh unit in a second manner in response to a second event, in which the quantity of accessed memory rows of the refresh unit is greater than the threshold quantity.