AI-Based Operating Point Estimator for Shared Power Thermal Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current algorithms for configuring hardware components under shared power/thermal constraints are inefficient in optimizing power consumption while achieving optimal performance, often resulting in increased power usage despite improved performance.
Innovation Solution
An AI network is trained to correlate workload descriptions, performance metrics, and power consumption metrics to determine an optimal operating point for chip hardware, balancing performance and power consumption by selecting appropriate clock frequencies and voltage states for each hardware component.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If current algorithms configure hardware components to achieve optimal performance in particular applications, then performance is improved, but power consumption increases
Solution Approach 1:
The patent applies parameter changes by using an AI network to dynamically adjust operating parameters (clock frequencies, voltage states) of hardware components based on workload characteristics. The AI network learns optimal parameter combinations that balance performance and power consumption, transforming the static configuration approach into a dynamic one that adapts to different workload conditions.
Solution Approach 2:
The patent implements feedback mechanisms by training the AI network with correlated data including workload descriptions, performance metrics, and power consumption metrics. The system continuously monitors actual performance and power consumption, using this feedback to refine AI network predictions and adjust hardware operating points accordingly, creating a closed-loop control system.
2Productivity
If hardware components are configured to operate at high frequency to increase computational capability, then frames per second and graphics experience are improved, but power consumption increases
Solution Approach 1:
The patent applies dynamics by transitioning from static hardware configuration to dynamic adjustment of operating points. The AI network enables real-time modification of clock frequencies and voltage states based on actual workload demands, allowing the system to adapt its performance characteristics dynamically rather than being locked into fixed high-frequency operation.
Solution Approach 2:
The patent changes operating parameters (frequency, voltage) based on AI network predictions that correlate workload characteristics with optimal performance-power tradeoffs. Instead of maintaining constantly high frequencies, the system adjusts parameters to match actual computational needs, reducing power consumption during less demanding workloads while maintaining performance when needed.
Data Source
AI summary
Integrated circuits, or computer chips, typically include multiple hardware components (e.g. memory, processors, etc.) operating under a shared power (e.g. thermal) constraint that is sourced by one or more power sources for the chip. Typically, the hardware components can be individually configured to operate at certain states (e.g. to operate at a certain frequency by setting a clock speed for a clock dedicated to the hardware component). Thus, each hardware component can be configured to operate at an operating point that is determined to be optimal, usually in terms of achieving some desired goal for a specific application (e.g. frame rates for gaming, etc.). In the context of chip hardware that operates under a shared power/thermal constraint, a method, computer readable medium, and system are provided for determining the optimal operating point for the chip that takes into consideration both performance of the chip and power consumption by the chip.