3D Connectivity Die for Decoupled Compute and I/O Access

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Planar programmable device architectures face challenges in incorporating additional services like hardened interconnects, which disrupt regular structures and complicate I/O bandwidth distribution, necessitating monolithic FPGA designs that require users to be aware of device architecture for optimal mapping.

Innovation Solution

A 3D stack of semiconductor dies with a connectivity die that includes configurable interconnects, allowing regular processing blocks to communicate with irregularly arranged I/O blocks without disrupting the fabric, enabling independent configuration and fair access.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If monolithically compiled FPGA design is used to provide full I/O bandwidth to compute elements, then I/O bandwidth distribution is improved, but device architecture complexity increases and users must be aware of underlying device architecture

Engineering Contradiction:
ImproveI/O bandwidth distributionVSAvoiddevice architecture complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The device is segmented into multiple coarse-grained regions, each with its own local interconnect fabric. This segmentation allows independent configuration of each region without affecting others, reducing the complexity of monolithic compilation while maintaining full I/O bandwidth access through hierarchical interconnection.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Each coarse-grained region is equipped with local I/O blocks and local interconnect fabric, providing quality differentiation. This allows compute elements in each region to access local I/O resources directly, improving bandwidth distribution while reducing the complexity of global interconnect management.

Inventive Principle:
Principle #3Local quality

2Device complexity

If coarse-grained regions with local I/O only are used, then device complexity is reduced, but communication between blocks is limited and users must map logic to specific regions

Engineering Contradiction:
Improvedevice complexityVSAvoidcommunication between blocks
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

Multiple coarse-grained regions are merged through a hierarchical interconnect structure that combines local interconnect fabrics with global interconnect resources. This allows compute elements in different regions to communicate freely while maintaining the simplicity of local configuration within each region.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The interconnect fabric is designed to serve multiple functions: local communication within a region, global communication between regions, and I/O access. This universal interconnect structure enables versatile block communication without requiring users to map logic to specific regions, enhancing adaptability while keeping device complexity manageable.

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

3Productivity

If hardened interconnects are included in planar architectures, then I/O bandwidth is improved, but regular fabric structures are disrupted and timing characteristics worsen

Engineering Contradiction:
ImproveI/O bandwidthVSAvoidfabric structure regularity
Core Design Contradiction:
ProductivityVSStability of the object's composition

Solution Approach 1:

The architecture transitions from planar to 3D stacking, allowing hardened interconnects and I/O blocks to be placed in separate vertical layers. This dimensional separation enables high-bandwidth I/O resources to coexist with regular fabric structures without disruption, maintaining timing characteristics while improving I/O bandwidth through three-dimensional integration.

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

Data Source

PatentUS12482756B2Connectivity layer in 3D devices
Publication Date: 2025.11.25 XILINX INC
  • US12482756B2 patent drawing
  • US12482756B2 patent drawing
  • US12482756B2 patent drawing

AI summary

Embodiments herein describe a 3D stack of dies (e.g., an active-on-active (AoA) stack) with a connectivity die that enables the decoupling of processing regions in coupled dies from each other and from the physical location of I/O blocks on an I/O die. For example, the first die may have a plurality of hardware processing blocks that are arranged in a regular manner (e.g., an array with rows and columns). The connectivity die can include interconnects that couple these hardware processing blocks to I/O blocks in a second die. These I/O blocks may be arranged in an irregular manner. The interconnects in the connectivity die can provide fair access so that processing blocks on a first side of the first die can access an I/O block on the opposite side of the second die without using resources for neighboring processing blocks.