Memory Manager for SoC Reliability via Sensor-Based Content Relocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In System on Chips (SoCs), unreliable memory regions can impact functionality and performance by providing incorrect values, which existing technologies fail to address effectively.
Innovation Solution
A memory manager system that includes sensors and a reliability module to determine memory reliability using sensor information and reliability information, allowing for the relocation of content from unreliable memory regions to more reliable ones and adjusting memory operating parameters based on determined reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If memory regions are used without reliability assessment, then device complexity is reduced, but reliability deteriorates due to incorrect values from unreliable memory
Solution Approach 1:
The system performs preliminary characterization of memory regions during manufacturing to determine their reliability properties. This pre-assessment creates reliability information that is stored and used during runtime, allowing the memory manager to make informed decisions about which memory regions to use without adding complex real-time assessment mechanisms.
Solution Approach 2:
The system implements feedback mechanisms where sensor information (temperature, voltage, process parameters) is continuously monitored and fed back to the memory manager. This feedback loop allows dynamic adjustment of memory region selection and content relocation based on current operating conditions, improving reliability while maintaining manageable complexity through automated decision-making.
2Reliability
If content is moved from unreliable memory regions, then reliability is improved, but productivity decreases due to additional memory operations
Solution Approach 1:
The memory management system operates autonomously to monitor memory reliability conditions and automatically relocate content from unreliable to reliable memory regions without requiring CPU intervention. This self-service approach minimizes the impact on productivity by handling reliability management independently of the main processing operations.
Solution Approach 2:
The system dynamically changes memory allocation parameters based on reliability assessments. When a memory region's reliability deteriorates below a threshold, the system automatically adjusts by relocating content to alternative regions. This parameter adaptation allows the system to maintain high productivity by optimizing memory usage patterns based on actual reliability conditions rather than using static allocation.
3Measurement precision
If multiple sensor types are used for reliability determination, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system merges multiple sensor types (temperature sensors, voltage sensors, process sensors) into a unified reliability assessment framework. By combining the information from these different sensors through a centralized memory manager, the system achieves comprehensive reliability measurement precision without requiring each sensor to operate independently, thus managing complexity through integration rather than multiplication of separate systems.
Data Source
AI summary
An embodiment of a manager includes at least one input node configured to receive information regarding a region of an integrated circuit, and a determiner configured to determine, in response to the information, a likelihood that the region will cause an error. For example, the region may include a memory, and contents of the memory may be transferred to another, more reliable memory, if the likelihood that the memory will cause an error in the data that it stores equals or exceeds a likelihood threshold.