3D Memory Network Nodes for High-Density Interconnects

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current memory systems face challenges in increasing packing density and performance due to shrinking device surface areas, leading to manufacturing yield reduction and increased complexity in interconnects, as well as performance losses from software overhead and bandwidth constraints in multi-processor systems.

Innovation Solution

The development of three-dimensional memory systems with network architectures that allow for interconnected network nodes and processors, enabling high-performance data sharing and access through multidimensional interconnects, which reduce the need for message passing and improve memory latency by distributing data management closer to processors.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If device surface area is shrunk to increase packing density, then memory capacity increases, but manufacturing yield decreases and interconnect complexity increases

Engineering Contradiction:
Improvememory capacityVSAvoidinterconnect complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent transitions from two-dimensional planar memory layouts to three-dimensional stacked architectures, where memory banks are arranged vertically across multiple layers. This dimensional change allows increased memory capacity without proportionally increasing interconnect complexity, as the vertical stacking enables more efficient use of the substrate area and reduces the number of interconnect layers required compared to horizontal expansion.

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

2Quantity of substance

If device surface area is shrunk to increase packing density, then memory capacity increases, but manufacturing yield decreases

Engineering Contradiction:
Improvememory capacityVSAvoidmanufacturing yield
Core Design Contradiction:
Quantity of substanceVSReliability

Solution Approach 1:

The memory system is divided into multiple independent memory banks organized in a three-dimensional stack, where each bank can be manufactured and tested separately. This segmentation allows for better yield management, as defects in one bank do not necessarily affect the entire memory array, and enables modular manufacturing approaches that can improve overall production yield while achieving high total capacity.

Inventive Principle:
Principle #1Segmentation

3Quantity of substance

If memory banks are increased to provide large data storage, then storage capacity increases, but interconnect delays increase

Engineering Contradiction:
Improvestorage capacityVSAvoidinterconnect delays
Core Design Contradiction:
Quantity of substanceVSLoss of time

Solution Approach 1:

The patent implements a hierarchical interconnect architecture where local interconnects serve nearby memory banks within the same stack layer, while global interconnects handle communication between different layers and external processors. This local quality approach ensures that frequently accessed data within the same bank experiences minimal delay, while the overall system maintains high capacity through the expanded three-dimensional structure.

Inventive Principle:
Principle #3Local quality

4Area of stationary object

If three-dimensional memory systems are implemented, then packaging density improves, but device complexity increases

Engineering Contradiction:
Improvepackaging densityVSAvoidsystem complexity
Core Design Contradiction:
Area of stationary objectVSDevice complexity

Solution Approach 1:

The memory system employs a nested hierarchical structure where memory cells are organized into banks, banks are stacked into layers, and layers are integrated into a three-dimensional array. This nesting approach achieves high packaging density by efficiently utilizing vertical space, while the regular repeating patterns at each hierarchical level provide modularity that helps manage system complexity through standardized design units.

Inventive Principle:
Principle #7Nested doll (Nesting)

Data Source

PatentUS10681136B2Memory network methods, apparatus, and systems
Publication Date: 2020.06.09 MICRON TECHNOLOGY INC
  • US10681136B2 patent drawing
  • US10681136B2 patent drawing
  • US10681136B2 patent drawing

AI summary

Apparatus and systems may include a first node group include a first network node coupled to a memory, the first network node including a first port, a second port, a processor port, and a hop port. Network node group may include a second network node coupled to a memory, the second network node including a first port, a second port, a processor port, and a hop port, the hop port of the second network node coupled to the hop port of the first network node and configured to communicate between the first network node and the second network node. Network node group may include a processor coupled to the processor port of the first network node and coupled to the processor port of the second network node, the processor configured to access the first memory through the first network node and the second memory through the second network node. Other apparatus, systems, and methods are disclosed.