Taipei, Monday, Apr 23, 2018, 00:05


Big Data Helps TSMC in Improving Defect-free Rate

By Vincent Wang
Published: Nov 20,2014

TAIPEI, Taiwan — The Ministry of Science and Technology (MOST) illustrated how big data is going to play an important role in improving the defect-free rate of wafer production at the early test stage. The industry-academic cooperation findings could lower the defect rate 14 percent of TSMC's wafer production at the early test stage.

More on This

Taiwan Establishes National Quantum Computer Research Project

TAIPEI, Taiwan - Taiwan’s Ministry of Science and Technology (MOST) yesterday held an “Exploring Secrets of Quantum ...

Groundbreaking for TSMC 5nm Factory Estimated for 2020

TAIPEI, Taiwan - TSMC on last Friday in Southern Taiwan Science Park held a groundbreaking ceremony for the first phase ...

The Deputy Minister of MOST Chien Chung-liang emphasized that this findings showing TSMC's defect rate lowered 11 percent to 14 percent which means wafer production at the early stage, not that of at the mass production stage. “I hope that this findings won't confuse you all, becasue the defect-free rate of TSMC is very high.”

Chien Chung-liang said that in the 2016 MOST R&D program has highlighted big data, echoing Intel and TSMC tried to find solutions out of big data lately.

This big data research team is led by the research fellow of Institute of Statistical Science Academia Sinica Ing Ching-kang, he stated that the manufacturing process of wafer is complicated for in which needed to be painted layer by layer. “It's extremely important for those wafer developers to find out the key machines at the early test stage to improve the defect-free rate. Big data is the way to find out that key machines.”

“If one wafer is needed to pass through 300 machines, generally speaking, to find out the problem machine is 2 to the power of 300, which is definitely a big data.” Ing added.

Ing Ching-kang gave a clearly example on how big data is working in this field. Firstly, ranking the machines in terms of performance based on statistics. Secondly, setting the truncation point by probability. Thirdly, examining to find out the problem machines.

CTIMES loves to interact with the global technology related companies and individuals, you can deliver your products information or share industrial intelligence. Please email us to

2153 viewed

comments powered by Disqus