DRAM Refresh via External Multi-Bank Timing Control

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current DRAM refresh mechanisms, both internal auto-refresh and external row-granular refresh, are incompatible, leading to inefficiencies in energy and performance due to the linear increase in refresh overhead with memory capacity, with external refresh schemes unable to exploit internal optimizations.

Innovation Solution

A method for an external multi-row, multi-bank refresh function using optimized command sequences, which determines inter-operation timing parameters and applies optimized row-level activation and pre-charge commands to reduce refresh cycle time and energy consumption, replacing the internal auto-refresh mechanism without modifying the DRAM or its protocol.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If internal auto-refresh mechanism is used, then refresh operations are handled internally with optimized architecture, but external controllers cannot exploit these optimizations and must follow strict timing rules

Engineering Contradiction:
Improveinternal auto-refresh handlingVSAvoidexternal controller optimization
Core Design Contradiction:
Extent of automationVSAdaptability or versatility

Solution Approach 1:

The patent introduces an intermediary mechanism where the external controller issues refresh commands that are translated and executed by the internal auto-refresh logic. The controller sends row address commands, and the internal mechanism handles the actual refresh operations, combining external control flexibility with internal optimization efficiency.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of operation

If external row-granular refresh is used, then specific rows can be targeted for refresh, but the refresh overhead grows linearly with capacity and cannot skip rows

Engineering Contradiction:
Improverow-level refresh controlVSAvoidrefresh cycle time
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The patent implements preliminary action by predicting which rows will need refresh based on access patterns and proactively refreshing them before they are accessed. The system monitors row access history and pre-refreshes rows that are likely to be accessed soon, eliminating the need to refresh all rows systematically and reducing overall refresh time.

Inventive Principle:
Principle #10Preliminary action

3Adaptability or versatility

If manual external refresh with ACT and PRE commands is used, then row-by-row refresh is possible, but performance and energy efficiency are reduced compared to internal optimizations

Engineering Contradiction:
Improveexternal refresh flexibilityVSAvoidrefresh performance
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent merges external row-granular control with internal auto-refresh optimizations by combining manual ACT/PRE command flexibility with the internal mechanism's ability to handle multiple rows simultaneously. The external controller specifies which rows need refresh, and the internal logic executes them using optimized parallel operations, achieving both flexibility and high performance.

Inventive Principle:
Principle #5Merging (Combining)

4Reliability

If refresh operations are performed frequently to maintain data integrity, then data availability is ensured, but energy consumption increases

Engineering Contradiction:
Improvedata retentionVSAvoidrefresh energy consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent implements feedback mechanisms where the system continuously monitors row access patterns, hit rates, and data retention status. Based on this feedback, the refresh policy is dynamically adjusted - rows with high access frequency or poor retention are refreshed more frequently, while rows with low access frequency are refreshed less often, optimizing the balance between reliability and energy consumption.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS10622054B2Using runtime reverse engineering to optimize DRAM refresh
Publication Date: 2020.04.14 TU KAISERSLAUTERN
  • US10622054B2 patent drawing
  • US10622054B2 patent drawing
  • US10622054B2 patent drawing

AI summary

A method and a related DRAM controller for refreshing a DRAM with an external multi-row, multi-bank refresh function based on optimized command sequences involves determining, at initialization time of the DRAM, inter-operation timing parameters for the external multi-row, multi-bank refresh function, determining optimized timing parameters for row-level activation (ACT) and pre-charge (PRE) commands, and applying the optimized timing parameters for the row-level ACT and PRE commands for refreshing the DRAM with the external multi-row multi-bank refresh function. The auto-refresh function of an SDRAM is replaced.