I see no one has yet mentioned to stay away from early socket M2 cores and specifically Revision F chips because of the IMC...I know I have always been one to support the IMC and feel it's one of the better things to happen to CPU design in a while but it's not perfect
You look at AMDs previous trends, newcastle revision was full of bugs with the memory controller and winchester also suffered similar issues with certain capacities and such...wasn't until RevE cores which was nearly a well over a year since existing AMD64 cores were released that most of the problems with the IMC were worked out and obviously in this case redesigning the controller for DDRII support is going to encounter similar bugs in development that may lead to some problems
If anyone followed the K8 development you'd see that prototype results with early AMD64 cores about the same performance as comparable XP cores, same trend as RevF cores are having with existing AMD64 cores...and I expect the IMC development trend to continue with RevF and later revisions
Furthermore, smaller fabrication processes are also not really a valid reason to upgrade...the initial winchester die shrink proved rather unsuccessful and while it was true the winchester overclocked well for its era, the farbrication had an issue with higher frequencies which is why RevE cores replaced the Winchester rather quickly