Multi-Root Semiconductor Fabric for IP Integration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The integration of diverse intellectual property (IP) blocks into a single system-on-chip (SoC) is hindered by design complexity and customization requirements, necessitating re-design to accommodate varying interface and signaling needs, which slows the development of advanced semiconductor devices.

Innovation Solution

The implementation of a standardized on-die interconnect protocol, such as the Integrated On-Chip System Fabric (IOSF) specification, enables efficient integration of IP blocks by defining interfaces, protocols, and arbitration mechanisms, allowing for broad use across different types of chips and facilitating design reuse.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If IP blocks are customized to accommodate varying interface and signaling requirements of different SoC segments, then adaptability to different markets is improved, but device complexity and design time increase significantly

Engineering Contradiction:
Improveadaptability to different marketsVSAvoiddesign complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a universal root complex design that can operate in multiple root space configurations (single-root and multi-root modes) without requiring separate customized designs for each market segment. The root complex includes configurable components such as the address decoder and arbitration logic that can be programmed to support different SoC architectures, thereby achieving adaptability across server, desktop, mobile, and embedded markets while maintaining a single standardized design

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

2Manufacturing precision

If IP blocks are re-designed for each SoC to meet specific interface requirements, then integration accuracy is improved, but productivity and development speed decrease

Engineering Contradiction:
Improveintegration accuracyVSAvoiddevelopment speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent performs preliminary configuration of the root complex components during system initialization rather than requiring redesign for each integration. The address decoder and arbitration mechanisms are pre-designed with configurable parameters that are programmed according to the specific SoC requirements, allowing rapid adaptation while maintaining integration accuracy through predefined configuration sequences

Inventive Principle:
Principle #10Preliminary action

3Adaptability or versatility

If multiple root spaces are implemented in a single semiconductor device, then adaptability to different market segments is improved, but device complexity increases

Engineering Contradiction:
Improvemulti-root space capabilityVSAvoidinternal structure complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements dynamic root space configuration where the root complex can be programmed to operate in either single-root or multi-root mode based on the specific market segment requirements. The arbitration logic and address decoding mechanisms are designed to be dynamically reconfigurable through programming, allowing the same hardware structure to adapt to different root space configurations without requiring physical redesign or increasing inherent structural complexity

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3304328B1Providing multiple roots in a semiconductor device
Publication Date: 2021.09.29 INTEL CORP
  • EP3304328B1 patent drawingFigure 1
  • EP3304328B1 patent drawingFigure 2
  • EP3304328B1 patent drawingFigure 3

AI summary

In one embodiment, a system includes: a first root space associated with a first root space identifier and including at least one first host processor and a first agent, the at least one first host processor and the first agent associated with the first root space identifier; a second root space associated with a second root space identifier and including at least one second host processor and a second agent, the at least one second host processor and the second agent associated with the second root space identifier; and a shared fabric to couple the first root space and the second root space, the shared fabric to route a transaction to the first root space or the second root space based at least in part on a root space field of the transaction. Other embodiments are described and claimed.