AMD swings to a loss in Q4

They did make a BIG mistake

IMHO they did make one SERIOUS error in 2011 and that was killing the AM3 socket instead of consolidating it. Lets face it most users aren't even stressing dual cores much less triples, quads, and sextacores, so AMD could have kept AM3 while cutting it down to one or two Thuban chips which then could have had cores disabled to cover the lower markets and get close to 100% yields on wafers. There is a pretty big market in those that need more than the E series and Bulldozer frankly scores worse than Thuban pretty much across the board. there are also many that don't want an APU because they plan on adding their own discrete so that GPU would be just wasting cycles. Finally because AM3 was a mature socket boards for AM3 are dirt cheap while having plenty of features that appeal to the DIY crowd.

So while cranking up the E series was a smart move giving up the DIY and low midrange market by only having BD is a dumb move IMHO. At the current price difference when I can no longer get AM3 chips I'll be seriously looking at Intel because the higher price of BD and worse benchmarks make it not very competitive which is a shame because Thuban was a really nice chip and with Zosma they could have just sold the ones with bad cores as Phenom II Quads. dumb move AMD, dumb move.

No it does not

The CPU division taken separately grew 7.2% on revenue and 81% on profit. Where exactly do you see the Windows drop in these numbers.

It was the discrete GPU, their share in the foundry portion, etc which tanked the results.

That is not surprising. Nvidia has a BIG lead now and it is growing. Both Intel and AMD thought that they can catch up by going hybrid. Instead of catching up that strategy opened a new front between them while allowing Nvidia to escape into the safety of the open (and uncontested) discrete GPU and CGPU market.