使用累積學習預測生產測試中晶片的最小工作電壓

Minimum Operating Voltage Prediction in Production Test Using Accumulative Learning

郭彥庭(Yen-Ting Kuo)

指導教授 ： 李建模

國立臺灣大學/電機資訊學院/電子工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202004287

摘要

隨著技術的縮減，工藝變化對晶片效能的影響越來越大。 因此，我們需要透過昂貴的功能測試來測試所有的生產晶片。 為了降低測試成本，在這篇論文中，我們提出了一種新的方法來預測生產晶片的最小工作電壓（Vmin）。 此外我們提出了兩個新的關鍵特徵以提高預測準確性。 我們提出的累積學習法可以減少批貨之間差異的影響。 兩個7奈米工業製程設計(大約有120萬顆量產晶片，這些晶片來自142批貨)的實驗結果表明，我們的預測準確度分別為93.4％和89.8％。 我們可以達到95％以上良好的預測(良好的預測意味著在誤差範圍內，預測的最小工作電壓不低於實際最小工作電壓)。 與傳統測試相比，我們的方法可以節省75％的測試時間。 要實施此方法，我們需要針對初始訓練和累積訓練分別設置測試流程。

關鍵字

製程變異 ； 機器學習 ； 晶片效能預測

並列摘要

As technology scales down, process variations have an increasing impact on chip performance. Therefore, we need to test all production chips by expensive functional test. For test cost reduction, in this thesis, we propose a new methodology to predict minimum operating voltage (Vmin) for production chips. In addition, we propose two new key features to improve the prediction accuracy. Our proposed accumulative learning can reduce the impact of lot-to-lot variations. Experimental results on two 7nm industry designs (about 1.2M production chips from 142 lots) show that our prediction accuracies are 93.4% and 89.8%, respectively. We can achieve above 95% good prediction (good prediction means that the predicted Vmin is no lower than the actual Vmin within an error bound). Our methodology can save 75% test time compared with traditional testing. To implement this method, we will need to have separate test flow for the initial training and accumulative training.

並列關鍵字

Process variation ； Machine Learning ； Chip Performance Prediction

參考文獻

Yongchan Ban, Yongseok Kang, and Woohyun Paik, "Model-based CMP (Chemical-Mechanical Polishing) proximity correction for mitigating systematic process variations," 2015 International SoC Design Conference (ISOCC). IEEE, 2015.

Shekhar Borkar, et al., "Parameter variations and impact on circuits and microarchitecture," Proceedings of the 40th annual Design Automation Conference. 2003.

Shekhar Borkar,"Designing reliable systems from unreliable components: the challenges of transistor variability and degradation," IEEE Micro 25.6 (2005): 10-16.

Keng-Wei Chang, et al., "DVFS Binning Using Machine-Learning Techniques," IEEE International Test Conference in Asia (ITC-Asia), 2018.

Janine Chen, et al. "Data learning techniques and methodology for Fmax prediction," 2009 IEEE International Test Conference, 2009.