運用線性迴歸預測處理器效能

Abstract

本研究以桌上型主機板之處理器為主要探討標的，以研發人員的產業領域知識(Domain Knowledge)來找到影響處理器效能的關鍵參數，進行實驗室大量測試數據(Data)的整理，運用簡單的線性迴歸(Linear Regression)方法，讓主機板讀取到使用的處理器及散熱器特性後，來預測處理器在搭配的散熱器下最佳操作頻率及需求電壓，使主機板的處理器在作業系統中能以優化後的效能穩定運作，提供效能需求及內容創作的使用者能增加其使用的應用程式效能、提升使用者的工作效率。此為本研究探討的議題，除了提供效能需求的使用者更佳的體驗，也為產業數據趨動、線性迴歸的應用和發揮提供一個學術上的案例參考。

This research investigates the performance prediction model of desktop processors and identifies the key performance indicators affecting desktop processors performance. We apply the linear regression to the prediction of desktop processor performance so that a motherboard is able to know the capability of desktop processors and coolers. The proposed method allows us to predict the optimized operating frequency and required voltage while the operating system achieves the optimal performance at the stable status of a motherboard. The results help content creation and performance-hungry users to have a better experience over daily applications and to increase their working efficiency.