熱暫態模式於電子裝置系統效能最佳化之應用

Thermal Transient Model Applied to Predict the Performance Optimization of Electronic Device Systems

林昇照(Sheng-Chao Lin)

指導教授 ： 蔡瑞益

中原大學/工學院/機械工程研究所/博士(2021年)

https://doi.org/10.6840/cycu202100073

摘要

動態熱管理機制被廣泛應用在追求外觀時尚、輕薄短小的現代電子裝置上，動態管理系統效能與散熱問題。在業界，有效率地來判定動態熱管理控制機制的最佳參數是必須的。利用藉由集總熱容法及能量法所建立的熱暫態模擬方法，來預測中央處理器及機殼溫度的熱暫態變化，可用來取代系統效能最佳化過程中所必須經歷的散熱測試，以縮短效能最佳化所需的作業時間。受測機台的熱暫態模擬參數，可由定功率熱測試所產生的熱暫態實驗數據，透過最小平方方法，或是經由執行一般應用程式所產生的熱暫態實驗數據，再利用粒子群演算法而求得。在所有樣本機台測試數據中，熱暫態模擬的預測誤差在9.1%之內。系統效能最佳化後的最佳控制參數，能藉由最大化額外增加效能法求得，以最大化系統效能並平衡系統散熱需求。實驗數據顯示，在一台惠普公司裝置有英特爾第十一代酷睿i7 中央處理器　(i7-1185G7)　的Corvette機台上，中央處理器及機殼溫度與藉由熱暫態模擬法預測的結果完全吻合。並且，系統效能經過最佳化之後，效能也比最佳化前增加了27.49%。在這個實際的應用例子裡，經由熱暫態模擬進行系統效能最佳化所花的時間，也能由十個小時縮短至五個小時。

關鍵字

熱暫態模擬 ； 系統效能最佳化 ； CPU溫度暫態預測 ； 機殼溫度暫態預測 ； 粒子群演算法 ； 動態熱管理 ； 系統效能預測 ； 散熱模組 ； 筆記型電腦散熱

並列摘要

Dynamic Thermal Management is widely used as a solution to manage system performance and thermal issues in modern thin and stylish electronics devices. Efficiently optimizing the control policy parameters of the Dynamic Thermal Management is necessary in the industry. In this study, the thermal transient prediction method was established using lumped thermal capacity method to predict CPU and chassis skin temperatures for replacing the thermal tests in optimization processes. The thermal transient parameters of test system were extracted from step power test or application test data by Least Square Method or Particle Swarm Optimization. Tests among sample systems showed the error was within 9.1%. Control policy’s parameters were optimized by maximizing extra performance gain (∆W) method to maximize system performance and balance thermal constraints. Validation results showed that CPU and chassis skin temperatures responded right on the edge as predicted and the system performance was improved by 27.49% on a HP Corvette, 11th Generation Intel® Core™ i7 Processors (i7-1185G7) system. In practical applications, system optimization time reduced from 10 hours to 5 hours with thermal transient prediction methods.

並列關鍵字

Thermal transient prediction ； CPU and chassis skin temperatures ； particle swarm optimization ； Lumped Thermal Capacity Method ； system performance

參考文獻

[1] D. Shin, S. Chung, E. Chung, and N. Chang, Energy-Optimal Dynamic Thermal Management: Computation and Cooling Power Co-Optimization, IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, 6 (3) (2010).

[2] E. Rotem, A. Naveh, A. Ananthakrishnan, E. Weissmann, and D. Rajwan, "Power-Management Architecture of the Intel Microarchitecture Code-Named Sandy

Bridge", IEEE Micro, 32 (2) (2012) 20-27.

[3] Enhanced Intel® SpeedStep® Technology for the Intel® Pentium® M Processor, White Paper, Mar. 2004

[4] C. Abomailek, F. Capelli, J.-R. Riba, P. Casals-Torrens, Transient thermal modelling of substation connectors by means of dimensionality reduction, Applied Thermal Engineering, 111 (2017) 562–572