Memory Control Circuit Error Thresholding for MRAM Cache Reliability

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

MRAMs used as cache memory are unreliable in writing and data retention, requiring an ECC error correction mechanism that incurs significant area and latency overhead, and may terminate as a fatal error due to uncorrectable errors, leading to system crashes.

Innovation Solution

A memory control circuit with an error detector, error correction circuitry, and access controller that determines the number of error bits and controls access to a secondary memory based on a threshold value, allowing for write-through or write-back operations to maintain data integrity and reduce system crashes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If ECC error correction mechanism is incorporated to correct errors in MRAM cache memory, then data reliability is improved, but area overhead and latency overhead increase significantly

Engineering Contradiction:
Improvedata reliabilityVSAvoidarea overhead and latency overhead
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies partial error correction by determining the number of error bits and comparing it with a threshold value. When the error bit count is within the threshold, only partial correction is performed; when exceeding the threshold, the system switches to accessing higher-level memory. This selective approach reduces the area and latency overhead compared to full ECC correction while maintaining acceptable data reliability for cases with fewer errors.

Inventive Principle:
Principle #16Partial or excessive action

2Use of energy by moving object

If MRAMs are used as cache memory to replace SRAMs, then power consumption is reduced, but writing reliability and data retention reliability deteriorate

Engineering Contradiction:
Improvepower consumptionVSAvoidwriting reliability and data retention reliability
Core Design Contradiction:
Use of energy by moving objectVSReliability

Solution Approach 1:

The patent introduces an intermediary error detection and management system between the MRAM cache and the higher-level memory. This intermediary monitors error bits, determines their count, and decides whether to correct errors locally or access higher-level memory, thereby bridging the reliability gap between power-efficient MRAM and reliable traditional memory systems.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system dynamically changes operational parameters by adjusting the threshold value for error bits based on system requirements and MRAM characteristics. This allows flexible adaptation of the error tolerance level, enabling the system to optimize between power consumption and data reliability by changing the decision parameter for error handling.

Inventive Principle:
Principle #35Parameter changes

3Reliability

If uncorrectable error occurs in ECC process, then data integrity is maintained, but system crashes occur due to fatal error termination

Engineering Contradiction:
Improvedata integrityVSAvoidsystem continuity
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent performs preliminary error detection and counting before attempting full correction. By determining the number of error bits in advance and comparing with the threshold, the system proactively identifies cases that would lead to uncorrectable errors and handles them by accessing higher-level memory, preventing fatal crashes before they occur.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system prepares backup error handling strategies in advance by having higher-level memory accessible as a fallback. When error bits exceed the correctable threshold, the system smoothly transitions to accessing higher-level memory, cushioning against the impact of uncorrectable errors and maintaining system continuity without crashes.

Inventive Principle:
Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentUS10042725B2Memory control circuit, cache memory and memory control method
Publication Date: 2018.08.07 KIOXIA CORP
  • US10042725B2 patent drawing
  • US10042725B2 patent drawing
  • US10042725B2 patent drawing

AI summary

A memory control circuit has an error determination circuitry to determine whether an error-bit number is larger than a predetermined threshold value set based on a maximum number of error bits correctable by the error correction circuitry, when it is detected by the error detector that an error is contained in data read for verification of data written to the first memory or in data read from the first memory, and an access controller to control access to a second memory having an access priority lower than the first memory when it is determined that the error-bit number is larger than the threshold value, and to control access to the first memory without accessing the second memory when it is determined that the error-bit number is equal to or less than the threshold value.