There's this article on Ars today regarding AMD: Op-ed: AMD may be facing irrelevance
It touches on many of the topics that have come up in the build a pc thread, that AMD really isn't very competitive in the desktop PC market, and intel has responded by focusing it's efforts elsewhere. As noted in the article and comments, desktop isn't the only game in town these companies care about, and must take it's share of common R&D efforts that produce a wide range of hardware.
There's lots of specifics you can go into, lots of directions that processing chips can go in, and the merits of each will suit different types of user. I guess what I'm wondering is as a group of people who generally are about gaming, how good a move away from what has traditionally been good for us, more processing power and towards different areas such as efficiency at a certain capability level could be better, and what the limits are of that.
The other side of the coin to hardware that can do lots of sums, is the software that makes those demands and runs well or poorly depending on what's available. There will always be software that consumes all resources that are available to it, but I wonder whether a extended period of limits could be beneficial to rein in inefficient code. The parallel I'd make here is to the current extended console cycle, where developers keep pulling rabbits out of hats to do most of what they want.
Having limits has a cost, as some things won't be possible (say you need to run a few thousand complex AI actors in a world) and the developer would need to produce a different game because they can't make what they want. Something I've seen though is that when a new 'generation' of hardware comes out, while there's genuinely things that couldn't be done on previous generations, there's also developers who use that power trivially and inefficiently. There's a balance to be stuck somewhere.
Developing more efficient and cheaper to produce hardware could also have the effect of making it into a commodity good. Rather than raising the ceiling of capability at a high cost, you raise the floor and bring the majority of users up to the current 'standard'. Another factor I can see is efficiency influencing form factors, moving away from big boxes to where a small quiet ITX box would be a general option worth considering for the majority of PC users.