3D Cross-Point Switch Stacking for Scalable Port Routing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Traditional cross-point switches are inflexible and costly due to their hardwired design, which struggles to accommodate varying user requirements for different applications, such as different port numbers and widths, leading to increased complexity and cost when trying to satisfy multiple user needs.

Innovation Solution

A scalable three-dimensional cross-point switch architecture using stacked switching dies with through-silicon vias (TSVs) and multiplexers, allowing for configurable port numbers and widths by programmable logic, enabling flexible and efficient routing between components.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a hardened cross point switch is designed with maximum connection capability to satisfy all possible user requirements, then all user applications can be supported, but the cost and complexity of the device increases significantly

Engineering Contradiction:
Improvecross point switch connection capabilityVSAvoiddevice complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The cross point switch is divided into multiple functional modules: ingress stage switches, middle stage switches, egress stage switches, and multiplexers. Each module is independently configurable, allowing the system to be segmented into different operational configurations based on user requirements without requiring maximum capability in all paths simultaneously.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The switch fabric employs dynamic multiplexing where multiplexers can be programmatically configured to route signals through different paths (direct or indirect) based on real-time requirements. This dynamic reconfiguration allows the same physical hardware to adapt to varying port width and connection needs without being permanently committed to a maximum configuration.

Inventive Principle:
Principle #15Dynamics

2Adaptability or versatility

If a hardened cross point switch is designed with maximum connection capability to satisfy all possible user requirements, then all user applications can be supported, but the cost of the device increases significantly

Engineering Contradiction:
Improvecross point switch connection capabilityVSAvoiddevice cost
Core Design Contradiction:
Adaptability or versatilityVSEase of manufacture

Solution Approach 1:

The switch fabric is designed as a universal platform where the same ingress switches, middle stage switches, and multiplexers can serve multiple different application scenarios. By programmatically configuring the multiplexers and routing paths, a single hardware implementation can universally support various port widths (e.g., 100 bits or 400 bits) and connection topologies, eliminating the need to manufacture different hardware versions for different users.

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

Solution Approach 2:

The system allows dynamic changing of operational parameters such as port width and routing paths through programmable logic configuration rather than requiring physical hardware changes. This enables the same device to be manufactured once and then configured with different parameters to meet various user requirements, significantly reducing manufacturing costs.

Inventive Principle:
Principle #35Parameter changes

3Device complexity

If a cross point switch is designed for specific user requirements with fixed parameters, then the device complexity and cost are reduced, but the adaptability to different user applications is limited

Engineering Contradiction:
Improvedevice complexityVSAvoidcross point switch connection capability
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The multiplexers in the switch fabric are designed to be dynamically reconfigurable through programmable logic, allowing the same physical switch to change its routing behavior and port configurations based on different user applications. This dynamic capability maintains low hardware complexity while achieving high adaptability.

Inventive Principle:
Principle #15Dynamics

4Device complexity

If traditional two-dimensional crossbar switch architecture is used, then the structure is simple, but the scalability to accommodate varying port numbers and widths is limited

Engineering Contradiction:
Improveswitch architecture complexityVSAvoidswitch scalability
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The patent transitions from a traditional two-dimensional crossbar architecture to a three-dimensional switch fabric with multiple stages (ingress, middle, egress) arranged in vertical layers. This dimensional expansion allows signals to route through multiple levels and paths, enabling scalable accommodation of varying port numbers and widths while maintaining manageable complexity through structured organization.

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

Solution Approach 2:

The switch fabric is segmented into multiple functional stages (ingress stage, middle stage, egress stage) with each stage containing multiple switches and multiplexers. This segmentation allows independent configuration and scaling of each stage to meet specific port requirements without redesigning the entire switch architecture.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP2947893B1Method and circuit for scalable cross point switching using 3-d die stacking
Publication Date: 2018.10.10 ALTERA CORP
  • EP2947893B1 patent drawingFigure 1A~1B
  • EP2947893B1 patent drawingFigure 2A~2B
  • EP2947893B1 patent drawingFigure 3~4

AI summary

A cross-point switch having stacked switching dies on a component die is disclosed. The cross point switch allows scalability by adding switching dies. The switching dies include ingress switches that are coupled to multiplexers to a middle stage switches. The inputs and outputs of the ingress switches are connected to the switching interface region via through silicon vias (TSVs). The outputs of the ingress switches are also coupled by TSVs to multiplexers for routing to middle stage switches on a switching die above. If the switching die is stacked on another switching die, the outputs of the ingress switches are coupled by TSVs to the multiplexers for routing to the middle stage switches of the switching die below. By adding switching dies, the switch is configurable to increase the number of ports as well as the width of the ports.