Processor Power-Setting Control with Self-Tuning NNPID Thermal Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing device-specific thermal control algorithms require significant tuning efforts by manufacturers, leading to sub-optimal configurations and performance degradation.
Innovation Solution
A power-setting mechanism utilizing a Neural Network Proportional Integral Derivative (NNPID) controller that learns platform limitations, reducing tuning requirements and achieving improved thermal management and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If device-specific thermal control algorithms are used, then thermal management capability is provided, but significant tuning efforts are required and performance is sub-optimal
Solution Approach 1:
The NNPID controller performs self-tuning by automatically learning platform-specific thermal characteristics and limitations during operation. The controller adapts to different platforms without requiring manual manufacturer tuning, enabling the system to serve itself in configuring optimal thermal parameters for each platform.
Solution Approach 2:
The controller dynamically adjusts thermal control parameters based on learned platform characteristics. By changing parameters automatically through machine learning rather than fixed manual configuration, the system achieves optimal thermal management across diverse platforms without requiring complex pre-tuning for each device.
2Reliability
If traditional PID controllers are used, then thermal control is provided, but performance is degraded due to platform-specific limitations
Solution Approach 1:
The NNPID controller implements enhanced feedback mechanisms where temperature sensor inputs continuously inform the controller about actual thermal conditions. This feedback loop enables the controller to learn from real-world thermal responses and adjust power settings dynamically, improving both thermal control accuracy and overall system performance compared to traditional fixed PID controllers.
Solution Approach 2:
The patent replaces traditional mechanical/manual tuning processes with an intelligent software-based NNPID controller that uses machine learning algorithms. This substitution eliminates the need for manual parameter adjustment and enables automatic adaptation to different platforms, significantly improving performance while maintaining thermal control.
3Reliability
If manual tuning is performed for each platform, then thermal characteristics are addressed, but manufacturing complexity and time increase
Solution Approach 1:
The NNPID controller is designed as a universal solution that can operate across multiple different platforms without requiring platform-specific customization. The controller automatically adapts to various thermal characteristics through self-learning, enabling a single design to serve multiple platforms and thereby simplifying the manufacturing process while maintaining effective thermal control.
Solution Approach 2:
The controller performs preliminary learning and adaptation automatically during initial operation on each platform. By conducting the tuning process automatically and preliminarily without requiring manual manufacturer intervention, the system prepares itself for optimal thermal management before deployment, reducing manufacturing complexity and time.
Data Source
AI summary
For example, a power-setting controller may be configured to provide a power setting for a processor based on one or more sensor-based inputs corresponding to one or more temperature sensors. For example, a sensor-based input corresponding to a temperature sensor may include a temperature input and a target temperature. For example, the power-setting controller may include one or more NNPID-based power controllers configured to provide one or more sensor-based power settings corresponding to the one or more temperature sensors. For example, an NNPID-based power controller of the NNPID-based power controllers may be configured to provide a sensor-based power setting corresponding to the temperature sensor based on the sensor-based input corresponding to the temperature sensor. For example, the NNPID-based power controller may include an NNPID controller configured to determine a temperature setting based on the temperature input and the target temperature corresponding to the temperature sensor.