I'm curious about the APUs handling of heavy floating point work. I do a lot of scientific/statistical computing work, and the FX series does have new extensions for just this type of work.

Of course, the best of that series also comes with a 220W TDP.