Processor Domain Power Allocation via Efficiency Ratings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques fail to optimally distribute power budgets among multi-frequency domains in processors, leading to inefficiencies in energy consumption and performance, as they lack effective metrics for cross-domain evaluation and dynamic workload sensitivity considerations.
Innovation Solution
A method is introduced to determine efficiency ratings for each domain by calculating a scalability factor and cost factor, with an adjustment factor to ensure fairness, and using these ratings to allocate power budgets dynamically, prioritizing domains with higher efficiency ratings to maximize performance per unit of power.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If power budget is distributed among multi-frequency domains based on current techniques, then power allocation is performed, but optimal distribution is not achieved leading to energy inefficiency
Solution Approach 1:
The patent changes the parameter of power allocation by introducing efficiency ratings that dynamically adjust power distribution based on domain performance characteristics. The system calculates efficiency ratings for each domain and uses these ratings to determine optimal power allocation, transforming the static power distribution into a dynamic parameter-adjustment mechanism that resolves the contradiction between energy efficiency and performance.
Solution Approach 2:
The patent implements a feedback mechanism where domain efficiency is continuously evaluated and used to adjust power allocation. The system monitors domain performance, calculates efficiency ratings, and feeds this information back into the power management decision-making process, enabling adaptive power distribution that maintains both energy efficiency and performance requirements.
2Ease of operation
If power budget is allocated without cross-domain evaluation metrics, then allocation decisions can be made, but effective power budget balancing cannot be achieved
Solution Approach 1:
The patent introduces efficiency ratings as an intermediary metric that mediates between power allocation decisions and energy efficiency outcomes. This intermediary measurement system enables informed decision-making by providing a quantitative basis for evaluating domain performance and guiding power distribution, thus resolving the contradiction between ease of operation and energy wastage.
Solution Approach 2:
The patent replaces ad-hoc or rule-based power allocation mechanisms with a systematic efficiency rating-based approach. By substituting the previous mechanical decision-making process with a data-driven efficiency evaluation system, the patent achieves both ease of operation through automated rating calculation and reduced energy wastage through optimized allocation.
3Device complexity
If utilization is used as the primary metric for power allocation, then allocation can be performed, but true power demand cannot be differentiated between domains
Solution Approach 1:
The patent applies local quality by introducing domain-specific efficiency ratings that capture the unique performance characteristics of each domain. Instead of using a single global metric like utilization, the system evaluates each domain's local efficiency properties, enabling precise differentiation of power demand while maintaining manageable complexity through standardized rating calculations.
Solution Approach 2:
The patent transforms the static utilization metric into a dynamic efficiency rating that adapts to changing domain performance conditions. The efficiency rating system dynamically adjusts to reflect true power demand by incorporating domain-specific performance factors, thereby improving measurement precision without significantly increasing device complexity.
Data Source
Figure 1
Figure 2
Figure 3~6
AI summary
The efficiency rating (ER) of each domain, in a processor, may be compared and then the power budget may be allocated, effectively, among the domains based on the ERs of the domains. The ER may indicate relative advantage among domains in terms of performance return for a given power budget, i.e., a higher effectiveness may be expected in power utilization if the ER is higher for a domain. The ER of a domain may be defined as (scalability factor/cost factor * alpha). The scalability factor may be defined as a performance increase (in %) brought about by an increase in the clock frequency (in %) provided to the domain. The cost factor may be defined as a power budget value required in bringing about an increase in the clock frequency provided to the domain and alpha is an adjustment factor.