Loading-Aware PDN Control for Processing Unit Power Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power distribution networks (PDNs) for processing units face challenges in managing dynamic load changes and high-power demands, leading to increased costs and performance loss due to inadequate load awareness, necessitating innovative strategies to optimize power control.
Innovation Solution
A sub-microsecond adaptive performance limitation strategy using a loading aware engine and controller to manage power and frequency based on real-time characteristic signals from the processing unit, employing a trained model to predict dynamic power and current, and adjust operating frequency and voltage accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If more voltage regulators are added to PMIC to meet high-power demands, then the power delivery capability is improved, but the PDN cost increases
Solution Approach 1:
The patent implements dynamic frequency adjustment based on real-time load monitoring. The system continuously monitors power consumption and dynamically adjusts the operating frequency of the processing unit to match actual demand, replacing static over-provisioning with adaptive control. This allows the PDN to operate efficiently at lower power levels while maintaining the capability to scale up when needed.
Solution Approach 2:
The patent employs a feedback mechanism where the system monitors actual power consumption and load conditions, then adjusts frequency accordingly. This closed-loop control ensures the PDN delivers only the necessary power, avoiding the need for excessive voltage regulators while maintaining stability and performance when required.
2Power
If frequency is limited to meet PDN requirements, then power demand is controlled, but computing power and performance are lost
Solution Approach 1:
The system dynamically adjusts frequency based on real-time load monitoring rather than applying static limitations. When load increases and power demand approaches PDN capacity, the system responds by adjusting frequency to maintain power within limits while maximizing performance within available power budget. This prevents unnecessary performance loss during low-power scenarios.
Solution Approach 2:
The patent changes the operating frequency parameter adaptively based on monitored power consumption and load conditions. By continuously adjusting this key parameter, the system optimizes the trade-off between power consumption and computing performance, ensuring maximum productivity within PDN constraints rather than applying fixed frequency caps.
3Device complexity
If static frequency assignment is used, then system simplicity is maintained, but performance optimization under varying load is lost
Solution Approach 1:
The patent transitions from static to dynamic frequency management by implementing real-time load monitoring and adaptive adjustment mechanisms. The system maintains relative simplicity through integrated monitoring and control logic while achieving significant performance optimization by continuously adapting frequency to actual load conditions rather than relying on predetermined static assignments.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
In an aspect of the disclosure, a PDN for power control of a processing unit includes a loading aware engine configured to receive multiple characteristic signals from the processing unit, and determine a loading information of the processing unit according to the multiple characteristic signals using a trained model. The loading information is related to a dynamic power and/or a dynamic current. The PDN also includes a clock generator configured to provide a clock signal with an operating frequency to the processing unit. The PDN also includes a controller coupled to the loading aware engine and the clock generator, and configured to control the clock generator based on the loading information.