I agree with the feedback so far in that the best answer is "it depends".
Personally, I think the biggest reason to separate broad/phrase/exact into different adgroups is to take advantage of embedded match. (And the only reason to do that is if you have short-tail broad match terms that convert well, but not for the phrase itself. For example, if the broad match term "widgets" is converting at a $5 CPA, but "widgets" as an exact match term is converting at $15, you'd want them in separate adgroups so you can make sure broad match doesn't "contaminate" exact match when your broad match bids exceed your exact match bids. (Separate adgroups, however isn't enough. You have to separate adgroups AND use embedded match to prevent this from happening.)
The biggest downside to separating into 3 separate adgroups is that it takes that much longer to pick a winning ad (assuming you're split-testing multiple ads.)
In a nutshell, I'd say to choose one or the other depending on whether you want to
1) have a extremely precise control at the keyword level, and make sure your winning broad match terms don't "contaminate" your exact match terms,
or 2) run split tests between different ads and be able to determine a winner as quickly as possible.
PS - to have the best of both worlds, you'd do embedded match, and then compile the ad stats from the 3 different ad groups by using a pivot table, but that's taking "geekiness" to a whole new level. It's probably only worth going to this level of trouble for ultra high-traffic accounts.
You can read more about embedded match at:
[
adwords.google.com ]