Thank you Leon
I installed every drivers already.
I face problem still.
First assumption is that we have 4 GPU and I assigned manually 3 GPU. When we use GPU 1 which seems like internally used in the windows (I checked by task manager and it goes 100% however others just use only 20 % during the training) we face the freeze. So I assigned GPU ID only use GPU 0,2,3. Now mult- gpu training is working.
In this afternoon I will check again. In the windows, I only can use the Cuda 9 for tensorflow. I agree your opinion however our organization can not allowed dual-booting. So I only can use the windows computer in the office.