Interconnect Physical Layer Re-Initialization Across Parallel Lanes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current interconnect architectures in high-performance computing systems face challenges in meeting the increasing demand for bandwidth and power efficiency, particularly in servers and mobile devices, where they need to balance performance with power consumption.
Innovation Solution
The development of a High Performance Interconnect (HPI) architecture that includes a layered protocol stack with a transaction layer, link layer, and physical layer, along with features like credit-based flow control, virtual channels, and embedded clock signaling, to enable efficient data transfer and power management across multiple devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional multi-drop buses are used for electrical communications, then device complexity is reduced, but communication performance and bandwidth are insufficient
Solution Approach 1:
The interconnect architecture is segmented into multiple independent point-to-point links instead of a single shared bus. Each link has dedicated bandwidth and can operate independently, eliminating the contention and bottlenecks of shared buses while maintaining manageable complexity through modular design
Solution Approach 2:
The architecture transitions from a one-dimensional shared bus to a multi-dimensional hierarchical structure with multiple levels (physical layer, link layer, transaction layer) and multiple parallel links. This dimensional expansion provides both higher bandwidth through parallelism and better performance through layered protocol management
2Productivity
If processing power and number of devices are increased, then computing capability is improved, but power consumption increases
Solution Approach 1:
The system implements periodic link training and state transitions where links can enter low-power states when not actively transmitting. The layered protocol enables power management by allowing different layers to operate at different times, reducing overall power consumption while maintaining high computing capability when needed
Solution Approach 2:
The interconnect supports dynamic parameter changes including variable link widths, adjustable speeds, and configurable topology adaptations. These parameter changes allow the system to optimize the balance between bandwidth requirements and power consumption based on actual workload conditions
3Productivity
If interconnect bandwidth is increased to meet demand, then data transfer performance is improved, but power consumption increases
Solution Approach 1:
The interconnect architecture provides dynamic resource allocation where link width and speed can be adjusted based on actual data transfer requirements. The system can dynamically activate only the necessary number of lanes and regulate speeds to achieve high performance when needed while minimizing power consumption during low-demand periods
Solution Approach 2:
The bandwidth resource is segmented into multiple independent lanes that can be selectively activated. This segmentation allows the system to provide high aggregate bandwidth when needed while consuming power proportional to the actual number of active lanes, rather than requiring all lanes to be active continuously
Data Source
AI summary
Re-initialization of a link can take place without termination of the link, where the link includes, a transmitter and a receiver are to be coupled to each lane in the number of lanes, and re-initialization of the link is to include transmission of a pre-defined sequence on each of the lanes.