The efficiency of CIN and scanf

At the bottom of the poj 2406 question, a message is displayed: "This problem has huge input, use scanf instead of CIN to avoid time limit exceed ."

Think of the previous issues with tree arrays that often times out with CIN. After switching to scanf, we can use AC. It seems that the efficiency of CIN and scanf is quite different.

Reading others said Using STD: IOS: sync_with_stdio can eliminate iostream Input and Output cache, can save a lot of time (http://www.cnblogs.com/cszero/archive/2012/02/11/Zero0ne.html)