近年來，晶片產業越來越重要，運用範圍也越來越廣。除了是各種智慧型產品、遊戲顯示卡、數據中心…等的重要核心外，更是未來AI、雲端物聯網、自駕車等不可或缺的一部分，就連中國也正舉全國之力要來發展半導體晶片產業。
對於一個IC設計公司來說，業務銷售推廣及IC設計部門固然重要，但是產品品質控管及售後服務也不能忽視，而提升產品及服務品質往往是公司能更上一層走向永續經營的關鍵之一。
本文除了介紹何謂退貨授權(Return Merchandise Authorization, RMA)及GPU晶片外，主要專注討論於我們如何在一般退貨授權的流程中能夠更有效率地找出晶片不良原因並將結果快速回報給客戶和公司，以增加客戶退貨的效益及減少不必要的資源浪費。
本文提出在基於退貨授權的流程中執行現場驗證(On Site Verification, OSV)來驗證不良的晶片，並透過物聯網無線通訊技術，即時地將OSV數據呈現在客戶及公司端上。有別於以往RMA，晶片不需要先寄送到國外相關部門而是直接在客戶端先行驗證，能達到在最短的時間初步判斷產品狀況，除此之外也比較不會讓客戶有沒材料更換的情況發生。

In recent years, the chip industry has become more and more important and its application scope has become wider and wider. In addition to being an important core of various smart products, graphics cards, data centers, etc., it is also an indispensable part of the future AI industry, cloud Internet of things, and self-driving cars. Even China is doing its best to develop the semiconductor chip industry.
For an IC design company, the business sales promotion and IC design department are important, but product quality control and after-sales service also cannot be ignored. Improving the quality of products and services has always been one of the keys for the company to move to a higher level and move towards sustainable operation.
In addition to introducing what is meant by Return Merchandise Authorization (RMA) and GPU chips, this article focuses on how we can more efficiently find out the causes of chip defects in the general return authorization process and quickly return the results from customers and companies. Increase the efficiency of customer returns and reduce unnecessary waste of resources.
This article proposes to perform On Site Verification to verify defective chips in a return material authorization-based process. Unlike previous RMA, the chips do not need to be sent to relevant foreign departments first, but directly on the client side doing verification. Achieve a preliminary judgment on the product condition in the shortest time. In addition, it is less likely to let customers have no material replacement.

[1]林佳樺(2003)，「退貨授權(RMA)資訊在品質保證系統上之探討─以某通信公司為例」，元智大學，碩士論文，第1頁。
[2]大和有話說， dahetalk.com/2018/01/28/萬物聯網，淺談iot低功耗廣域網路趨勢：lora、sigfox、nb-iot｜
[3]LoRa Alliance_What is LoRalan.pdf,p.18.
[4]洪俊輝(2020)，「LoRa通信技術於戶外活動資訊收集之應用」，中正大學，碩士論文，第15-20頁。
[5]林裕哲(2019)，「NB-IoT於智慧農業之應用」，崑山科技大學，碩士論文，第17-18頁。
[6]https://zh.wikipedia.org/wiki/圖形處理器 ，WIKI 圖形處理器。
[7] NVIDIA CUDA Compute Unied Device Architecture Programming Guide. 2007.PDF, pp.2-16.
[8] John D. Owens, David Luebke, Naga Govindaraju, Mark Harris, Jens Krüger, Aaron E. Lefohn, and Timothy J. Purcell(2005)，”A Survey of General-Purpose Computation on Graphics Hardware” ，EUROGRAPHICS 2005,pp.21-22.
[9] 王培學(2017)，「利用GPU加速SIFT特徵之擷取與比對」，中央大學，碩士論文，第8-11頁。
[10]M.H. Attia; S.A. Elshehaby; A.S.Elmaghraby(2015),”Implementation of edge-enhancement nonlinear anisotropic diffusion filtering using different CUDA memory models”,2015 IEEE ISSPIT,pp.501-502.
[11]謝逸凡(2012)，「使用GPU提升事件相關電位之動態因果模型的運算效能」，中央大學，碩士論文，第15-21頁。
[12]Yeh-cheng Chen; C.N.Chu; Hung Min Sun; Ruey-Shun Chen; Mengru Tu; S.C Lin(2007)，”Using case-based reasoning method to design a return merchandise authorization system for supply chain management in Internet of Thing”，2007 IEEE 17th ICCT,pp568-569.
[13]圖解品質管理，五南出版，陳耀茂編著，第18-23頁。
[14]NVIDIA OSV FOR RMA GUIDE 2019Ver.PDF。
[15]NVIDIA MODS GUIDE.PDF。