Multi-Level STT-MRAM Encoding for Energy Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-level STT-MRAM devices face high write energy consumption and reduced cell endurance due to the requirement for two-step and hard state transitions, which are energy-intensive and affect device reliability.
Innovation Solution
A method is introduced to minimize two-step and hard state transitions by dividing data into sub-groups, identifying 'hot bits', and using an encoding scheme that complements the most significant bits of these hot bits, eliminating the need for high-energy transitions through XOR operations and encoding flags.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multi-level STT-MRAM devices use two-step and hard state transitions for data writing, then data storage capacity is improved, but write energy consumption increases and cell endurance decreases
Solution Approach 1:
The patent segments data into sub-groups of bits and identifies hot bits within each sub-group. By applying encoding only to sub-groups containing hot bits rather than all data bits, the system reduces the overall encoding overhead and energy consumption while maintaining the ability to handle multi-level data storage requirements.
Solution Approach 2:
The patent applies different treatment to different portions of data based on their characteristics. Specifically, encoding is applied selectively only to sub-groups containing hot bits (bits that require two-step or hard transitions), while other sub-groups are written without encoding. This local differentiation reduces total energy consumption while maintaining reliability where needed.
2Quantity of substance
If multi-level STT-MRAM devices use two-step and hard state transitions for data writing, then data storage capacity is improved, but device reliability decreases
Solution Approach 1:
The patent segments data into sub-groups and identifies hot bits within each sub-group. By applying encoding only to sub-groups containing hot bits rather than all data bits, the system reduces the overall encoding overhead and energy consumption while maintaining the ability to handle multi-level data storage requirements.
Solution Approach 2:
The patent applies different treatment to different portions of data based on their characteristics. Specifically, encoding is applied selectively only to sub-groups containing hot bits (bits that require two-step or hard transitions), while other sub-groups are written without encoding. This local differentiation reduces total energy consumption while maintaining reliability where needed.
3Reliability
If encoding is applied to all data bits, then transition states are minimized, but write operations become more complex and time-consuming
Solution Approach 1:
The patent segments data into sub-groups and identifies hot bits within each sub-group. By applying encoding only to sub-groups containing hot bits rather than all data bits, the system reduces the overall encoding overhead and energy consumption while maintaining the ability to handle multi-level data storage requirements.
Solution Approach 2:
The patent applies encoding selectively only to the extent necessary - specifically to sub-groups containing hot bits - rather than applying it universally to all data. This partial action approach reduces complexity and energy consumption while still achieving the goal of minimizing problematic transitions where they would occur.
Data Source
AI summary
Data is stored in a multi-level MRAM (MLC MRAM) cell in a manner that reduces transition states that require high energy. A new data block is received, and the new data block is divided into one or more sub-groups of bits, with each sub-group comprising at least two bits. Each sub-group is assigned data bit locations in a memory store. The received bits are compared with sub-groups present at the data bit locations to determine subgroups of hot bits. For each subgroup of hot bits, an encoding flag value is determined by XORing their most significant bits. The most significant bits of each subgroup of hot bits are complemented and the encoding flag is SET. A data block is generated to establish a data group for each subgroup of hot bits including the subgroup of hot bits and the encoding flag for that subgroup.