Ok here's my take on the announcement and some of the misconceptions shown by a few posters in this thread.
People tend to look at everything from a PC component standpoint and look at unified devices in the same terms. They are completely different & shouldn't be looked at in the same way.
PC components have been designed in an evolutionary method, to provide backwards compatibility for older hardware & software.
So when you build a component based PC, the architecture of the design is 30+ years old and has bottlenecks between the relevant components.
That is your typical PC architecture design and you can see that in order for the CPU to access the GPU, it needs to go through the northbridge which handles the internal IO processes.
The new PS4 will be based around a newer version of this design.
AMD have integrated the Northbridge, GPU & CPU into 1 APU but with this design it has a channel to a DDR3 memory bus.
With the PS4 they aren't using conventional DDR3 RAM but using GDDR5.
Essentially DDR3 - Low latency, low bandwidth & GDDR5 - High latency, high bandwidth.
See this post for the technical details of the differences between the 2.
http://www.techspot.com/community/t...-memory-and-gddr5-memory.186408/#post-1243450
And if it's also based around AMD's new design which is a fully integrated APU with a Heterogeneous Systems Architecture which I expect it will be, then there'll be parallel workloads instead of the current architecture which has the CPU & GPU using different memory spaces.
http://www.anandtech.com/show/5493/...ed-memory-for-cpugpu-in-2013-hsa-gpus-in-2014
So Sony are only doing what they have done in the past and focus on getting the best graphical performance of the current gen. But this time they've learnt from their mistakes and made the platform accessible to more developers by using a conventional architecture (x86-64).
There maybe a few mistakes in my post but it's close enough to what I think Sony are doing.