^^ Thanks to me
This is the thing I was trying to convey for a very long time but most of us didn't try to understand and some tagged me as an AMD fanboy. It was the Game Engine architecture which isn't optimized for Multi-threading environment due to the different RISC based In Order Vector architecture of the consoles like PS3 or Xbox 260, which is the reason behind poor performance of games in Multi-Core AMD CPUs, not the AMD Processor architecture.
Now Heterogeneous Computing, unified memory design for CPU and GPU, dynamic code path selection, parallel processing and task sharing between CPU cores and GPU etc are pretty well known and have been implemented in the PS4 and Xbox 720. So obviously these are going to come to PC game design very soon and believe me, multi-threaded work distributionis going to play the major role here rather than strong single threaded performance.
You might think I am exaggerating things a little but tell me one thing: in Video Encoder application like Handbrake which has OpenCl acceleration for using CPU + GPU as an heterogeneous compute platform where it can divide the work among CPU and GPU as per requirement, a Phenom II 965 + HD 6670 combination can beat a Core i7 3960X. So, when HSA in action, a X86 single threaded performance is not a very good thing to compete with it.
The future is going to be HSA and from now on Software developers will be targeting that platform instead of optimizing performance for a single core. AMD is promoting HSA for very long time and their APUs are product of the same thought. In Maxwell Design, nVidia is doing the same thing for their GPU design where it can work as an unified unit with CPU rather than a separate unit. Current generation GCN has the ability to use X86 based memory management and that is why it has far better compute or HSA performance as it can work more closely with the CPU.