Forget broadwell, I'm more interested in Next Gen DGPU Abilities

[dusk007]

But architecturally it isn't any different.
The GPU is in the chipset also leeched of the main memory and didn't have their own. The dedicated word doesn't mean it is a seperate chip but that it has its own dedicated video memory.
Aside from that architectural difference chip size is the only determinante of speed. How much transistors did they spend on the GPU. In that respect those old two chip designs on 40nm process don't really compare to a 22nm chip.
There used to be a northbridge and a southbridge. The northbridge is now always on the same chip. With the Haswell dual cores even the southbridge is on chip.
On how many chips something is spread out means nothing in most cases it is a waste of power and performance. It is only done if the resulting chip would be too big and expensive with everything included.

Assuming the same process having two chips half the size gets you nothing over one chip at full size, other than a lot of wasted logicboard space you could spend on getting a bigger fan in.

 

[leman]

What I meant is, that the pre early 2011 had the GPU integrated into the chipset, while early 2011 and after have it integrated into the cpu (as well as some parts of the chipset). So while you previously had 2 chips (chipset/gpu and cpu), which both needed cooling, you now have one

You still have two chips, its just that modern chipsets use so little power that they are usually passively cooled. And I don't really see how this is relevant here - an integrated GPU is an integrated GPU, no matter what chip its integrated with. In the past, it was a common thing to integrate the GPU into the chipset anyway - the GMA series of Intel iGPUs were also parts of the chipset - still doesn't make it a dGPU. Integration of a GPU and GPU on the same die is a quite new development. AFAIK, for Intel it only happened in 2010 and for AMD in 2011. And yet the concept of an iGPU existed long before that.

 

[maflynn]

The next-gen AMD Volcanic Islands have already been released. There are no mobile cards yet, AFAIK.


The next-gen Nvidia architecture, Maxwell is supposed to come in 2014.

I'm guessing Apple will stick with the nvidia dGPU. I wonder if AMD's bleak financial woes are impacting Apple's GPU selection. They I think may a slight profit last quarter but they're not really rolling in the dough at this point