Thank you Shenghui Yang for your help.
but my point is not to find a workaround but to show what is for me a bug... or a sort of...
I made a new comparison between Mathematica 9 and 10 (see code in attached file) :
I reduced the plot range to be sure that the plot results are comparable (same quality result).
I fixed also the number of points to 500 and the ploting method to "speed".
I still have a big difference in term of computation time :
Mathematica 10 : 155s
Mathematica 9 : 6s !!!
Attachments: