MRAM Integration for Memory Architecture Simplification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current computing systems rely on separate memory technologies like SRAM, DRAM, and flash memory, which require additional processing steps, consume more power, and have complex architectures due to the need for high-resistance blocks and frequent refreshing, limiting speed, area efficiency, and power consumption.

Innovation Solution

Integration of magnetoresistive random access memory (MRAM) cells into a single chip, combining level-1, level-2, and level-3 caches, main memory, and secondary storage, using a unified MRAM processing procedure that eliminates the need for separate resistance blocks and reduces interconnections, offering high resistance, low power consumption, and non-volatile performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If separate memory technologies (SRAM, DRAM, flash) are used in computing systems, then memory functionality is provided, but device complexity increases due to multiple memory types and interconnections

Engineering Contradiction:
Improvememory architecture complexityVSAvoidmemory functionality
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The patent combines multiple separate memory technologies (SRAM, DRAM, flash memory) into a single unified memory device using MRAM technology. This integration eliminates the need for multiple distinct memory components and their complex interconnections, while maintaining all required memory functionalities including high-speed caching and non-volatile storage capabilities.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The MRAM-based memory device performs multiple functions that previously required separate memory technologies. It provides fast access speeds comparable to SRAM, non-volatile storage like flash memory, and can be configured for different memory hierarchy levels (L1, L2, L3 caches, main memory, secondary storage), making it a universal memory solution.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Use of energy by stationary object

If separate memory technologies are used, then different memory functions are achieved, but power consumption increases due to frequent refreshing and additional processing

Engineering Contradiction:
Improvepower consumptionVSAvoidmemory function
Core Design Contradiction:
Use of energy by stationary objectVSAdaptability or versatility

Solution Approach 1:

The MRAM-based memory device is non-volatile and does not require frequent refreshing operations that consume significant power in DRAM systems. The memory cells retain data without continuous power input, eliminating the need for refresh circuits and reducing overall power consumption while maintaining all memory functions.

Inventive Principle:
Principle #25Self-service

3Device complexity

If high-resistance blocks are added to memory systems, then resistance requirements are met, but device complexity and area increase

Engineering Contradiction:
Improveresistance block requirementsVSAvoidresistance performance
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent removes the need for separate high-resistance blocks by using MRAM technology that inherently provides high resistance performance at the memory cell level. The magnetic tunnel junction structure naturally exhibits high resistance states, eliminating the requirement for additional resistance blocks and simplifying the overall device architecture.

Inventive Principle:
Principle #2Taking out (Extraction)

4Area of stationary object

If multiple separate memory chips are used, then memory capacity and functionality are achieved, but interconnect losses and area increase

Engineering Contradiction:
Improvechip areaVSAvoidmemory access speed
Core Design Contradiction:
Area of stationary objectVSProductivity

Solution Approach 1:

The patent integrates all memory functions into a single MRAM-based chip, eliminating multiple separate memory chips and their interconnections. This consolidation reduces interconnect losses, decreases the total chip area required, and improves memory access speed by removing bottlenecks associated with inter-chip communication.

Inventive Principle:
Principle #5Merging (Combining)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach simplifies system architecture, reduces power consumption, increases speed, and enhances security by integrating MRAM into the processor chip, providing faster, denser, and more flexible memory solutions with reduced chip area and interconnect losses.

Implementation Method 1

The MRAM cell 100 comprises a fixed or pinned magnetic layer 102 and a free magnetic layer 104 separated by a tunnel barrier 106. The magnetic state of the pinned magnetic layer 102 is fixed. The magnetic state of the free magnetic layer 104 is changed to store a data bit.

Methodology Applied
Scientific EffectMagnetoresistance: Magnetoresistance

Data Source

PatentUS20240251566A1Processing and memory device and system
Publication Date: 2024.07.25 TAIWAN SEMICONDUCTOR MANUFACTURING CO LTD
  • US20240251566A1 patent drawing
  • US20240251566A1 patent drawing
  • US20240251566A1 patent drawing

AI summary

An embodiment of an integrated circuit chip includes a combination processing core and magnetoresistive random access memory (MRAM) circuitry integrated into the chip. The MRAM circuitry includes a plurality of MRAM cells. The MRAM cells are organized into a number of memories, including a cache memory, a main or working memory and an optional secondary storage memory. The cache memory includes multiple cache levels.