Processor MEP Control Using Sensor Feedback and Heuristic DVFS
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques fail to efficiently adapt supply voltage to achieve optimal energy efficiency in digital integrated circuits, particularly in the near-threshold voltage regime, due to variations in workload, process, and temperature.
Innovation Solution
A processor configured with a minimum energy point (MEP) control circuit that uses sensor-driven energy computation techniques to efficiently track and adjust the MEP in real-time, leveraging pre-characterized lookup tables for fast and low-complexity computations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If conventional voltage adaptation techniques are used, then the system can operate at various supply voltages, but the energy efficiency is not optimized due to failure to track the true minimum energy point under varying conditions
Solution Approach 1:
The patent implements a feedback mechanism where the MEP controller continuously monitors operating conditions (workload, temperature, process variations) and adjusts the supply voltage to track the minimum energy point. The controller receives feedback from sensors and performance monitors to dynamically adapt the voltage frequency scaling, ensuring optimal energy efficiency under varying conditions.
Solution Approach 2:
The system transitions from static voltage operation to dynamic voltage and frequency scaling (DVFS). The MEP controller dynamically adjusts operating parameters based on real-time conditions, making the system adaptable to workload changes, temperature variations, and process differences while maintaining optimal energy efficiency.
2Use of energy by moving object
If detailed sensor-driven energy computation is performed to accurately track MEP, then energy efficiency improves, but computational complexity and overhead increase
Solution Approach 1:
The patent pre-characterizes the relationship between operating conditions and minimum energy points during manufacturing or initialization. Lookup tables (LUTs) are pre-computed and stored, containing optimal voltage-frequency pairs for various condition combinations. During operation, the controller simply queries these pre-computed tables rather than performing complex real-time calculations, reducing computational overhead while maintaining accuracy.
Solution Approach 2:
Instead of performing complex energy computations directly in real-time, the system uses pre-computed lookup tables that copy the results of detailed energy analysis. The LUTs store the essential information needed for MEP tracking, allowing the controller to make accurate decisions with minimal computational resources by referencing these pre-prepared data structures.
3Use of energy by moving object
If real-time tracking of minimum energy point is implemented, then energy consumption is reduced under changing conditions, but the system requires additional control circuitry and sensors
Solution Approach 1:
The MEP controller is designed as a multi-functional unit that integrates multiple capabilities: it monitors operating conditions, queries lookup tables, determines optimal voltage-frequency pairs, and controls the voltage regulator and frequency synthesizer. This universal controller consolidates what would otherwise require separate dedicated circuits for each function, reducing overall system complexity while enabling comprehensive MEP tracking.
Solution Approach 2:
The patent introduces an integrated voltage regulator (IVR) as an intermediary component that simplifies the voltage control path. The IVR provides precise voltage regulation and includes features like body bias control that enable fine-grained energy optimization. This intermediary component mediates between the digital control logic and the analog voltage domain, making the overall system more manageable despite the added complexity of real-time tracking.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In one embodiment, a processor includes a minimum energy point (MEP) controller to: generate a change in thermal tracking information, based at least in part on prior and current thermal information; generate a change in activity tracking information, based at least in part on prior activity information and current activity information; and determine a MEP performance state based at least in part on the change in thermal tracking information and the change in activity tracking information. Other embodiments are described and claimed.