Hierarchical STT-RAM Architecture for Scalable Memory Access

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional Spin Transfer Torque Magnetic Random Access Memory (STT-RAM) faces challenges in scalability and access times, limiting its development as a next-generation nonvolatile memory.

Innovation Solution

A magnetic memory system employing a hierarchical architecture with memory array tiles, global bit lines, global word lines, and intermediate circuitry, which allows for faster access times and increased write and read margins by controlling operations through global circuitry and reducing parasitic resistance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If conventional STT-RAM architecture is used, then non-volatility and low cost are maintained, but scalability and access times are limited

Engineering Contradiction:
Improveaccess timesVSAvoidarchitecture complexity
Core Design Contradiction:
SpeedVSDevice complexity

Solution Approach 1:

The memory system is divided into multiple memory array tiles, each with its own local bit lines and word lines. These tiles are organized in a hierarchical structure with intermediate circuitry that manages groups of tiles. This segmentation allows parallel access to multiple tiles simultaneously, improving overall access speed while distributing the complexity across modular units rather than requiring a monolithic high-speed architecture.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a hierarchical architecture that adds a vertical dimension to the traditional 2D memory array. By organizing memory tiles in multiple levels with intermediate circuitry between them, the system achieves three-dimensional scalability. This allows data access to occur across different hierarchical levels, effectively increasing access bandwidth and speed without proportionally increasing the complexity of any single level.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Quantity of substance

If memory density is increased, then capacity benefits are achieved, but parasitic resistance increases affecting read/write margins

Engineering Contradiction:
Improvememory densityVSAvoidparasitic resistance
Core Design Contradiction:
Quantity of substanceVSObject-affected harmful factors

Solution Approach 1:

The memory system divides the high-density storage into multiple smaller memory array tiles, each with its own dedicated bit lines and word lines. This segmentation ensures that the parasitic resistance of interconnect lines does not scale with total memory capacity, as each tile operates independently with localized wiring. The intermediate circuitry manages these tiles in groups, further containing the impact of parasitic resistance to manageable levels.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Intermediate circuitry is introduced between the memory array tiles and the global bit lines/word lines. This intermediate layer acts as a buffer that drives signals to groups of tiles, reducing the burden on global lines and minimizing the impact of parasitic resistance. The intermediate circuitry can selectively activate groups of tiles, allowing high-density storage while maintaining adequate read/write margins through controlled signal distribution.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Productivity

If hierarchical architecture is implemented, then access times are reduced and read/write margins are increased, but device complexity increases

Engineering Contradiction:
Improveaccess speedVSAvoidcircuitry complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The control logic is segmented and distributed across different hierarchical levels rather than centralized. Each memory array tile has local control, intermediate circuitry manages groups of tiles, and global circuitry coordinates overall operations. This distributed control reduces the complexity burden on any single component while enabling fast access through localized decision-making and parallel operation of multiple tiles.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The hierarchical architecture uses universal control structures that can manage multiple memory tiles through standardized interfaces. The intermediate circuitry and global circuitry are designed to handle groups of tiles uniformly, allowing the same circuitry design to be replicated and scaled. This multi-functionality reduces overall complexity by avoiding the need for unique control logic for each individual tile, while still achieving fast access times through parallelism.

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

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

The hierarchical architecture enables scalable, high-density memory with short write times, low power consumption, and reduced read/write penalties, maintaining the benefits of STT-RAM such as non-volatility and low cost while improving access times and density.

Implementation Method 1

The magnetic element 12 is configured to be changeable between high and low resistance states by a current flowing through the conventional magnetic element 12... When the current density is sufficiently large, the current carriers driven through the magnetic element 12 may impart sufficient torque to change the state of the magnetic element 12

Methodology Applied
Scientific EffectSpin transfer effect:

Data Source

PatentEP2476121B1Method and system for providing a hierarchical data path for spin transfer torque random access memory
Publication Date: 2014.05.21 GRANDIS INC
  • EP2476121B1 patent drawingFigure 1~2
  • EP2476121B1 patent drawingFigure 3~4
  • EP2476121B1 patent drawingFigure 5~6

AI summary

A method and system for providing a magnetic memory are described. The method and system include providing memory array tiles (MATs), intermediate circuitry, global bit lines, global word lines, and global circuitry. Each MAT includes magnetic storage cells, bit lines, and word lines. Each of the magnetic storage cells includes at least one magnetic element and at least one selection device. The magnetic element(s) are programmable using write current(s) driven through the magnetic element(s). The bit lines and the word lines correspond to the magnetic storage cells. The intermediate circuitry controls read and write operations within the MATs. Each global bit line corresponds to a first portion of the plurality of MATs. Each global word line corresponds to a second portion of the MATs. The global circuitry selects and drives part of the global bit lines and part of the global word lines for the read and write operations.