Processor Core Thermal Management Using Temperature Evolution Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional processor devices face challenges in thermal management due to localized hot spots and inefficiencies in temperature monitoring, as thermal sensors are not optimally placed and may introduce latency, leading to ineffective temperature control.
Innovation Solution
A processor device employs a cluster thermal management circuit that uses a temperature evolution model to predict temperatures at hot spots based on power consumption measurements and external thermal sensor readings, enabling proactive thermal management through clock throttling and dynamic voltage and frequency scaling without requiring sensors at every point of interest.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If thermal sensors are placed at every point of interest (POI) within the processor core, then temperature measurement precision is improved, but device complexity and area occupation increase
Solution Approach 1:
The patent creates a virtual copy of the thermal sensing capability through software modeling. Instead of placing physical sensors at every POI, the system uses a temperature evolution model that copies thermal behavior patterns from limited physical sensor readings to predict temperatures at unmonitored locations. This virtual sensing approach achieves comprehensive temperature coverage without the hardware overhead of dense sensor deployment.
Solution Approach 2:
The temperature evolution model acts as an intermediary between the limited physical thermal sensors and the multiple POIs requiring temperature monitoring. The model receives inputs from physical sensors and power consumption data, then translates these into predicted temperature values at all POIs, effectively mediating the information gap between sparse measurements and comprehensive monitoring requirements.
2Measurement precision
If thermal sensors are placed closer to hot spots, then temperature measurement precision is improved, but area occupation and device complexity increase
Solution Approach 1:
The system copies thermal behavior characteristics by modeling temperature evolution patterns at hot spots based on power consumption data and limited sensor readings. The temperature evolution model captures the thermal dynamics of hot spots without requiring physical sensors to be embedded at these critical locations, achieving precise hot spot monitoring while preserving processor core area for functional elements.
3Adaptability or versatility
If more thermal sensors are added to monitor multiple POIs, then temperature monitoring coverage is improved, but manufacturing complexity and cost increase
Solution Approach 1:
The temperature evolution model serves as a universal solution that handles multiple POIs simultaneously using a single modeling framework. Rather than designing separate sensing mechanisms for each POI, the system uses one multi-functional model that adapts to predict temperatures at any number of POIs based on shared inputs from power consumption measurements and limited sensor data, simplifying the manufacturing process.
Solution Approach 2:
The system uses software-based temperature evolution models that can be copied and deployed across multiple POIs without additional hardware. This approach allows comprehensive monitoring coverage to be achieved through replicable computational models rather than complex hardware arrangements, reducing manufacturing complexity.
4Speed
If thermal sensors are used to monitor temperature in real-time, then temperature control responsiveness is improved, but latency in obtaining measurements increases
Solution Approach 1:
The system performs preliminary thermal analysis by continuously running the temperature evolution model that predicts future temperature states based on current power consumption patterns and historical thermal data. This proactive approach allows the thermal management system to anticipate temperature violations before they occur and take preventive action, reducing the effective latency compared to waiting for physical sensors to detect and report actual temperature violations.
Solution Approach 2:
The patent replaces the mechanical/physical thermal sensor measurement system with a computational temperature evolution model. Instead of relying on physical sensors to detect temperature changes, the system uses software-based thermal simulations that compute temperature states, eliminating the inherent latency of physical sensing and enabling faster thermal management responses.
5Manufacturing precision
If physical thermal sensors are deployed at all critical locations, then temperature control accuracy is improved, but the number of components and device complexity increase
Solution Approach 1:
The system creates virtual replicas of thermal sensing functionality through the temperature evolution model, which copies and extrapolates thermal behavior from limited physical measurements to predict temperatures at all critical locations. This approach achieves comprehensive temperature control accuracy without the hardware complexity of deploying physical sensors throughout the processor.
Solution Approach 2:
The temperature evolution model serves as an intermediary layer between the simplified physical sensor network and the complex thermal management requirements. It translates limited sensor inputs into comprehensive temperature information for all critical locations, reducing the complexity of the overall thermal management architecture while maintaining control accuracy.
Data Source
AI summary
Performing thermal management based on temperature evolution models in processor devices is disclosed herein. In some aspects, a processor device provides a cluster thermal management circuit that is configured to determine power consumption measurements for corresponding functional units of a processor core of a plurality of processor cores of a core cluster. The cluster thermal management circuit also determines temperature measurements by corresponding digital thermal sensors (thermal sensor) external to a point of interest (POI) within the processor core. The cluster thermal management circuit generates a predicted temperature at the POI based on a temperature evolution model that correlates power consumption measurements and temperature measurements with the predicted temperature at the POI. If the cluster thermal management circuit determines the predicted temperature at the POI exceeds a thermal mitigation threshold, the cluster thermal management circuit performs a thermal management operation.