Think of it this way. Multi GPU tech now has one GPU render one page while another is rendering the next one instead of them working together on the same page. This means that the optimal patern looks like this:
a--b--a--b--a--b--a--b
a is the first GPU, b is the second.
However, sometimes we see it look like this:
a-b---a-b---a-b---a-b---
We have two frames shot out right very quickly, then a larger space between the next two. Dual GPU platforms can do this, this causes micro-stutter. Having three GPUs should look like this:
a--b--c--a--b--c--a--b--c--a--b--c
It is a lot more unlikely for three GPUs to have a frame ready at the same time, so micro-stuttering is about as good as a single card solution, sometimes even better. However, FPS doesn't scale with three GPUs as well as it does with two. I think it's a fair trade off because if you have a micro-stuttering problem, it doesn't matter how high your FPS is, it still looks like crap. The scaling also helps because it means that instead of one GPU working to get 20 FPS or two woring to get 40 FPS, you can have three working to get 50FPS. Think of this: each GPU has more time to render each frame, so if it runs into a particualrly heavy frame, it has a lower impact on performance and a much lower chance of throwing the frame order out of whack.
three GPUs have a different problem too, but it's not nearly as bad, especially at high FPS. You need to wait an amount of frames equal to the number of GPUs before your mouse can register a change in position. Dual GPUs have almost double the FPS so it's not a big deal. Three GPUs can have a bigger deal if you have such high settings that the three GPUs can't get out a high enough FPS. If you have high FPS, waiting three frames is no big deal. If you have low FPS, well then it increases latency, that can be almost as bad as micro-stuttering.