Asynchronous Wavefront Clock Distribution for Large Multicore Chips

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Distributing a zero-skew clock signal to all nodes in multicore processors becomes increasingly difficult as chips grow larger, leading to issues with process, voltage, and temperature variations, and requiring higher design margins.

Innovation Solution

Propagate the clock signal in a wave pattern rather than a structured zero-skew manner, adjusting the relative timing of data signals based on the direction of propagation to maintain data coherency and synchronization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a zero-skew clock signal is distributed to all nodes in a structured manner (e.g., H-tree), then clock signal distribution is reliable, but the design margin increases and scalability deteriorates as chips grow larger

Engineering Contradiction:
Improveclock signal distribution reliabilityVSAvoiddesign margin requirements
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The clock distribution network is segmented into multiple wavefronts that propagate sequentially through the computational node network. Instead of distributing clock signals to all nodes simultaneously from a central source, the clock signal is divided into multiple wavefronts that travel through different paths, with each wavefront serving a specific region or subset of nodes. This segmentation reduces the complexity of maintaining zero-skew across the entire network while improving scalability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces dynamic timing adjustment mechanisms that allow the clock signal propagation delays to be adapted based on the specific path and distance. By making the clock distribution system dynamic rather than static, the design can accommodate variations in process, voltage, and temperature without requiring excessive design margins, thus resolving the contradiction between reliability and device complexity.

Inventive Principle:
Principle #15Dynamics

2Productivity

If the chip size increases to accommodate more nodes and cores, then computational capability improves, but clock signal distribution becomes increasingly difficult and less reliable

Engineering Contradiction:
Improvecomputational capabilityVSAvoidclock signal distribution reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent transitions from a traditional two-dimensional clock distribution approach to a multi-dimensional wavefront propagation model. By introducing the concept of sequential wavefronts that propagate through time and space in a structured manner, the system can scale to larger chip sizes while maintaining clock signal reliability. The wavefront approach adds a temporal dimension to the spatial distribution, allowing efficient clocking across larger distances.

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

Solution Approach 2:

The clock signal paths are pre-configured and pre-timed to account for process, voltage, and temperature variations before the actual operation begins. By performing preliminary timing adjustments and path optimizations, the system ensures reliable clock distribution across large chip areas without requiring real-time adjustments, thus maintaining reliability as computational capability scales.

Inventive Principle:
Principle #10Preliminary action

3Stability of the object's composition

If process, voltage, and temperature variations are accounted for with higher design margins, then clock signal stability improves, but the device complexity and design difficulty increase

Engineering Contradiction:
Improveclock signal stabilityVSAvoiddesign margin requirements
Core Design Contradiction:
Stability of the object's compositionVSDevice complexity

Solution Approach 1:

The patent employs parameter changes in the form of adjustable propagation delays and timing offsets that can be modified based on process, voltage, and temperature conditions. By making the clock distribution parameters adjustable rather than fixed, the system can adapt to environmental variations without requiring excessive design margins, thus improving stability while reducing design complexity.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentUS20250306624A1Multicore Processor Clock Distribution Using Asynchronous Wavefront
Publication Date: 2025.10.02 TENSTORRENT USA INC
  • US20250306624A1 patent drawing
  • US20250306624A1 patent drawing
  • US20250306624A1 patent drawing

AI summary

Systems and methods related to multicore processor clock distribution using an asynchronous wavefront are disclosed herein. A clock signal may be propagated to each node in a predictable and repeatable manner. The clock signal may be provided from a clock source to a subset of nodes of a network of nodes and may be distributed from each node of the subset of nodes to a respective adjacent node. The clock signal may be propagated via the adjacent nodes to any additional nodes of the network that are not among the subset of nodes or the adjacent nodes. Propagating the clock signal in this way may avoid issues related to distributing a zero-skew clock signal directly to all nodes, such as the common point in a clock distribution growing farther in time between two leaf points and higher design margins to account for larger processes, voltage, and temperature variations.