Yes, Bulldozer core microarchitecture didn't exactly impress on the desktop, with the power-performance ratio being the biggest problem even with discounted pricing. On the server front, the dual die 8 core-pair 'Interlagos' implementation of Bulldozer fared somewhat better as there are more apps that can benefit from many-thread resources, especially integer bound Web environments where the shared FP units aren't that much of a hindrance.

Since the Trinity APU already has the updated 32nm 'Piledriver' core included, the question was when will we really see it in the desktop and server general purpose CPUs. While there may be samples of Vishera floating around, this Bulldozer desktop follow-on is still nowhere near the official release. The 'Abu Dhabi' dual die follow-on to Interlagos, implementing the 'Piledriver' core, is howewer, pretty much confirmed for the September to October timeframe.
Any reason to wait? Well, there are two expected main benefits: one is the supposedly improved instructions per-clock efficiency, which could be up to 15% based on the more optimistic figures, and the other one is the higher clock frequency, up to some 20% vs similar Bulldozer cores especially on the Opterons. Finally, yes, Haswell-compatible FMA3 fused multiply add format. And, oh yes, further improved power management. Note that there will be no new sockets (meaning AM3+ and G34/C32 still), everything has to fit within the current TDP, memory and I/O bandwidth limits. No support for PCI Express 3.0 either...

Now, how would it impact the competitive positioning? Not much as of now - 22nm Ivy Bridge-E / EP updates (more on that this week) to Core i7 and Xeon E5 lines will keep the performance chasm as it is today. It would, however, show that AMD is still in the game and continuing to develop its CPU/Heterogenous Computing stuff so that, by the time the Steamroller and Excavator core designs come, it can hopefully compete again vs Intel.

Read more: AMD Piledriver FX and Opteron refresh? After summertime.. by VR-Zone.com