Hi, Ken,
Thank you for you reply.
I could agree with your comment that “the NAND writes is calculated based on total writes”. If there are some data generated in the SSD when cache is working, there must be an improvement of the Total Host writes. In fact this condition haven’t appeared, only 32Mb improved, but about 2Gb for Total NAND writes, here the Write Amplification is at least 50! It is abnormal. You could assume that it is an OS or monitoring software error, there will be some temporary data produced, then deleted quickly, but the Total Host writes nearly never increased during this process. How could we explain this phenomenon? Possibly the OS or the controller have omitted such extensive quantity of data 
Another possibility for this case is that there is the same phenomenon occurred in other company’s product. I haven’t another SSD, so never test this matter. Maybe “Putting IE/Chrome or system cache into other medium/drive saving your SSD's wear” is suitable only for other company’s product, not for Intel’s SSD, or even it is suitable for nothing.
Even though I mention this NAND writes problem here, the SSD could be used up for a long time in the normal operation condition. There is nobody besides me caring this matter, I had expected maybe we could find a “bug” and improve this product. But now I want to close this stuffy topic. It is not because we have got a solution, but there is no further proof for it, and I am just an user.
Thanks for your attention, Joe and Kan.