On-Chip Dynamic Power Estimator With Real-Time Weight Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power consumption estimates for processing units are inaccurate due to offline assumptions, failing to account for real-time fluctuations caused by varying workloads and environmental factors.
Innovation Solution
A dynamic power estimation unit adjusts weights in real-time by multiplying counter values with corresponding coefficients, using an on-chip learning algorithm to minimize errors between estimated and actual power consumption, ensuring more accurate predictions and maintaining the processor within a thermal envelope.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If offline assumptions and pre-silicon estimates are used to generate power consumption estimates, then the estimation process is simple and fast, but the accuracy of power consumption estimates deteriorates due to real-time fluctuations
Solution Approach 1:
The patent applies dynamics by transitioning from static offline assumptions to dynamic real-time weight adjustment. The learning algorithm continuously updates weights based on current system state, enabling the estimator to adapt to real-time fluctuations in power consumption while maintaining manageable system complexity through on-chip implementation.
Solution Approach 2:
The patent implements feedback through the learning algorithm that compares estimated power consumption with actual measurements and uses this error signal to adjust weights. This closed-loop feedback mechanism continuously improves estimation accuracy by learning from actual system behavior rather than relying on fixed pre-silicon models.
2Measurement precision
If real-time weight adjustment is implemented to improve power estimation accuracy, then prediction accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by performing weight adjustment in advance or during low-priority periods, rather than continuously interrupting critical operations. The learning algorithm can be executed during idle cycles or background processes, preparing updated weights before they are needed for accurate power estimation without causing delays.
Solution Approach 2:
The system performs self-service by automatically adjusting its own weights using an on-chip learning algorithm without requiring external intervention. This self-tuning capability allows the system to improve its estimation accuracy autonomously, reducing the need for manual calibration and minimizing the impact on overall system performance.
3Adaptability or versatility
If an on-chip learning algorithm is used to dynamically adjust weights, then adaptability to real-time conditions improves, but device complexity and manufacturing difficulty increase
Solution Approach 1:
The patent applies universality by designing the learning algorithm to serve multiple functions: it adapts to different workloads, compensates for environmental variations, and continuously improves estimation accuracy. This multi-functional approach consolidates what would otherwise require separate systems into a single integrated on-chip solution, managing complexity through functional consolidation.
Data Source
AI summary
Systems, apparatuses, and methods for implementing a dynamic power estimation (DPE) unit that adapts weights in real-time are described. A system includes a processor, a DPE unit, and a power management unit (PMU). The DPE unit generates a power consumption estimate for the processor by multiplying a plurality of weights by a plurality of counter values, with each weight multiplied by a corresponding counter. The DPE unit calculates the sum of the products of the plurality of weights and plurality of counters. The accumulated sum is used as an estimate of the processor's power consumption. On a periodic basis, the estimate is compared to a current sense value to measure the error. If the error is greater than a threshold, then an on-chip learning algorithm dynamically adjust the weights. The PMU uses the power consumption estimates to keep the processor within a thermal envelope.