If developers can include already written HSA optimized code in their PC ports targeted for AMD"s upcoming HSA APUs and GPUs for little additional cost, why wouldn't they? Notice nearly every AAA PC game now carries AMD's Gaming Evolved logo. AMD is clearly sitting at the head of the table with developers across the boards. Developers have every incentive to optimize for AMD's HSA APUs, the better their games run on cheap entry level and low end computers, the more games they sell. Including code already developed for Sony and Microsoft console APUs in their PC ports to maximize the gaming potential of AMD's Kaveri and beyond is a no brainer.
Nvidia has no piece of the consoles whatsoever. Over time their share of developers time and attention will steadily shrink as developers increasingly focus their attention on optimizing for AMD HSA APUs, which will provide by far the best gaming experience for the money -> AMD APUs will become the gamer's go to chip of choice.
You act as though PC's won far outstrip console performance by the end of next year, if not before. The scenario you describe might work out for AMD with their APU's IF they can manage to create an APU that has the performance advantage to make up for the system overhead and DX overhead that is included in a Windows PC combined with the performance advantages of the optimizations that may not even be carried over to PC due to sheer lack of systems that can take advantage of them.
Some things will favor AMD GPU's, but how is that any different than last year or years before when nearly every multiplatform console game was developed first and foremost for the AMD GPU in the Xbox 360? That didn't lead to superior performance for AMD GPU's. It led to nVidia ramping up its marketing machine and building up a relationship with publishers where they basically did a lot of the QA/QC/optimization work for free.
Right now, AMD has money to spend because they stalled their next gen update until the end of this year, having their driver team refocus on making their drivers tip-top for frame latency and Crossfire stutters (instead of making drivers for a new line of cards), and so with all the money they saved for the time being, they're shoving it at publishers to win them over to make fantastic bundles. And they really are fantastic bundles and great wins for "Gaming Evolved."
But don't make the mistake of thinking just because AMD is doing a great job of marketing and paying these companies lots of money to win their support. Publishers will support the GPU maker that pays them the most money. For the longest time, AMD didn't even try in this realm. Now they are and so they're competing heavily for something nVidia's just not used to having to work at. No doubt, this has caught nVidia off guard in the marketing department. And no doubt, they'll redouble their efforts, too.
It's all come together like a perfect storm. AMD needs big releases, big news, they need to look like they have a future given all the things they've done recently (ie., fired huge amounts of R&D, stalled CPU and GPU deliveries, changing product maps willy-nilly, losing the high end CPU and GPU markets definitively). They are likely selling Sony and MS their console's APU's for below cost just to get the word of mouth. They time that to go with a huge push on games and make the case they're the company focused on games.
Meanwhile, nVidia is trying to make inroads in mobile and keeps hitting the fabrication problems that prevent any high end CPU maker (no matter what type) from getting to where they want to be, which leaves them out of position while they try to coast on what they got in the GPU market since AMD is not launching anything new. Even the Titan is just the scraps from the Compute market repackaged.
Yet none of this really means AMD is going to actually have an advantage because unfortunately for AMD, Intel is out there, raining on their parade. No PC game is going to leverage the added capabilities of AMD APU's since no other system could take advantage, so there'd be a very limited audience for such an optimization. Even if they did, would they outweigh the sheer brute force of high performance Intel and nVidia/AMD discrete?
No. By the time any company would even remotely consider using such a technology since the market would have grown, you'll see computers have become so much faster that it won't even matter. They'll so far outstrip the APU that it'll be like arguing about fixed function T&L versions programmable shaders. By the time those APU advantages catch up, the middle and mid-high and high end will all be well outperforming it.
Leaving APU's to take up the bottom rungs where Intel seems rather strongly positioned to catch up in no time with its rather large, consistent leaps in iGPU performance every year. Meanwhile, AMD is having execution problems and has been delaying releases for the last few months.
If I had to bet, I'd bet money that Intel will take the low end and discrete will continue to dominate the high end for some time to come. Where in all that will AMD position the APU?