I've just checked the manual (page 2-15) and it's pretty clear, you have three options
1: one slot at x16 from the CPU
2: two slots at x8 from the CPU
3: one slot at x8 from the CPU and two slots at x16 from a NF200
The don't explicitly say the width of the link linking the NF200 to the CPU but by a process of elimination it has to be x8 rendering the x16 links from the NF200 to the GPUs kinda pointless.
IMO this is a rather stupid way to design a board. If you are going to use a NF200 you want to put all the CPUs lanes into it.
It's faster to run two video cards off the CPU at 8x/8x then the NF200 chip at 16x/16x. I can understand why the board is designed that way. The added latency of the NF200 could cause your video cards to perform less than cheaper motherboards. When using three video cards the NF200 chip in this manner actually becomes an advantage since cheaper boards can only run 8x/4x/4x and NF200 boards will run 16x/16x/8x.
If your only running two cards, a cheaper ASUS board will perform just fine without the added cost of the M4E.
If you have Tri-SLU/CF setup then the M4E is the board for you.