Now I know why I even did not get any aswer from freescale. I've done some measurements by myself and found that the adress hold time from ALE going Low is 0 or even little bit negative. That mean that you must use one additional waitstate for ALE for multiplexed flexbus write acces which is a very bad situation and totally waste of performance. With a very fast RAM you waste about 30% of flexbus throughput. I just wonder if the same problem exists on other Cortex M3 devices from other brands.