SoC Partition Profiling for Adaptive Clock Frequency Tuning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing hardware accelerators are limited in functionality due to the requirement of prior knowledge of tasks, which is often not available due to confidentiality, privacy, and intellectual property concerns, and they struggle to efficiently handle evolving workloads without pre-designed performance adjustments.

Innovation Solution

A system on a chip (SOC) with runtime profiling and performance controls that dynamically adjusts clock frequency and settings based on real-time metrics and constraints, allowing it to optimize performance without prior knowledge of tasks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If hardware accelerators are specialized for specific tasks with pre-fixed functionality, then performance efficiency is improved, but adaptability to evolving workloads deteriorates

Engineering Contradiction:
Improveperformance efficiencyVSAvoidadaptability to evolving workloads
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent implements dynamic performance management by continuously monitoring runtime metrics and adjusting clock frequencies of hardware partitions in real-time. The system transitions from static pre-configured performance settings to dynamic adaptation, allowing hardware accelerators to optimize their performance characteristics based on actual workload conditions without requiring prior knowledge of task types.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system employs feedback mechanisms through runtime metric collection and analysis. Performance managers monitor usage patterns of hardware partitions and use this feedback to dynamically adjust clock frequencies and performance settings. This closed-loop control enables the system to adapt to evolving workloads while maintaining optimal performance efficiency.

Inventive Principle:
Principle #23Feedback

2Productivity

If prior knowledge of tasks is used to program hardware acceleration features, then performance is improved, but confidentiality and intellectual property are compromised

Engineering Contradiction:
ImproveperformanceVSAvoidconfidentiality and intellectual property
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The system enables hardware accelerators to self-optimize their performance through automated runtime profiling and dynamic clock frequency adjustment. Instead of requiring external programming with prior task knowledge, the system autonomously monitors its own performance metrics and adjusts settings to maintain optimal performance, thereby eliminating the need to expose confidential task information.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent changes the approach from static task-specific programming to dynamic parameter adjustment. By modifying clock frequencies and performance parameters based on runtime observations rather than pre-programming, the system achieves high performance without requiring prior knowledge of task characteristics, thus protecting confidentiality and intellectual property.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If clock frequency is increased to improve processing speed, then throughput is improved, but power consumption increases

Engineering Contradiction:
ImprovethroughputVSAvoidpower consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The system dynamically adjusts clock frequencies based on actual workload demands and performance requirements. Rather than maintaining high clock frequencies continuously, the system optimizes frequency settings in real-time, increasing throughput when necessary while reducing power consumption during lower-demand periods, achieving an optimal balance between performance and energy efficiency.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent implements parameter changes by adjusting clock frequencies dynamically based on runtime performance metrics. The system modifies operational parameters (clock frequency) to match actual workload requirements, improving throughput when high performance is needed while minimizing power consumption when full performance is not required.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4650913A1Systems and method for performance management of system on a chip
Publication Date: 2025.11.19 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP4650913A1 patent drawingFigure 1
  • EP4650913A1 patent drawingFigure 2
  • EP4650913A1 patent drawingFigure 3

AI summary

Examples of the present disclosure describe devices, systems, and methods for runtime profiling workload on a system on a chip (SOC). In examples. A SOC records runtime metrics while running a workload using counters to calculate the usage of logical partitions of the SOC. The SOC uses the runtime metrics to determine the logical partitions' performance characteristics and the processors' optimal clock frequency in each logical partition based on the performance characteristics. The SOC sets the clock speeds of a processor in a logical partition while the workload is still running on the SOC to its optimal clock frequency.