I suspect we won't see a big-core FX Steamroller, but we might for Excavator. Steamroller, core-for-core, thread-for-thread, will probably be about on par with Nehalem, which is considerably better than before, but not enough to substantially close the gap with Intel on the server and high-end desktop. However, Excavator is supposed to widen the execution path considerably, and might well have IPC about on par with Sandy Bridge. That would be good enough to be competitive in high-end markets. For all the talk about the "death of the desktop" (which is really just about slower replacement cycles), servers aren't going anywhere - that's where all the vaunted "cloud" work is actually done, after all. And the margins on server products are higher than for those aimed at consumers. AMD has to focus on their biggest strengths (APUs and GPUs) as a priority, but I think they want to recapture some of that sweet server revenue in the long run. And that's not going to happen with APUs or Vishera. And once you've got a large multi-core server CPU, selling the lower-binned products to enthusiasts becomes a cheap and easy sideline. That said, I don't think any FX products beyond Piledriver will use the existing sockets. More likely they will create a new unified socket (basically AMD's answer to LGA 2011) to replace AM3+, C32, and G34.
Of course, all this is speculation. We could see Vishera stagnate forever if AMD runs into serious financial problems. Or if everything goes right (console and GPU revenues stay strong, Steamroller is a big hit with reviewers, and they score some important design wins) we might see a few extra bucks thrown towards R&D so that Warsaw can come out with Steamroller cores instead of Piledriver. I don't know, and I don't think that even AMD's management knows for sure yet. A lot of this is contingent.
What AMD really needs is an angel investor to throw $10 billion at the company. There's tons of future potential there, but they're held back by their financial limitations and the ongoing impact of the poor decisions made by Hector Ruiz.