It seems a bit confusing that the big push for Ivy is the improvement in GPU with hardly any advancement in CPU from Sandy, according to AnandTech.
Why? This is Intel's 'tick/tock' strategy they have been following for several years now. Ivy Bridge is a "tick" ( a shrink of substantively the same micro-architecture ). A substantial fraction of the introduction of Anadtech's article covers this.
Is Apple planning new MBP/A with only integrated Intel graphics and not a dedicated card?
The MBA is only integrated graphics now. Why would 2012 be any different?
The MBP 13" is only integrated graphics now. Unless they dump the ODD drive for more cooling (fan ) and GPU+VRAM , they would drop the same way in 2012.
The MBP's in general have a battery saving mode where the discrete GPU is turned off and battery life extended.
So why wouldn't Apple want a IGP solution that was faster but consumed the same (or less ) amount of power? The HD4000 is better than the HD3000 solution. Why not?
My 2007 MBP has an Nvidia chip that Apple replaced free just last year because it was bad, so reconciling with Nvidia seems backwards. But how can there be both an Ivy GPU and an Nvidia or AMD card as well?
The Ivy GPU comes whether Apple wants it or not. The only question is whether the Ivy GPU is "good enough" for the box and is there room, power, and cooling available to insert another GPU into the box.
This whole notion of "bring back" Nvidia integrated graphics is grossly flawed. That 'war' is over. The memory controllers have moved onto the CPU die. There is no "integrated graphics" if you loose control over where the memory controllers are placed.
AMD and Intel have placed those onto the CPU die at this point in all but the upper end server designs. Those too may get limited GPGPU additions in the future as transistor budgets for those designs get bigger also. With a 3 billion transistor budget it isn't too hard to add some hetergenous (non x86) performance to the die.
As FCPX will likely require more GPU in future updates, it seems there should be more VRAM from the onset, but is that even possible with an integrated GPU such as Ivy?
Not necessary if FCPX using OpenCL. The Ivy GPU could run the computations (up to a certain level) and the discrete GPU simply render them. Integrated graphics tap into the GB's of RAM already present in system. Or the roles could be switched if the OpenCL data isn't quite as large.
I'd hope that the new MBP would have more than 1gb VRAM so I won't have to play catch-up when FCPX and third party plug-ins require or recommend 2gb VRAM,
The issue is space and VRAM memory density. There is limited room for VRAM chips. Denser ones cost more.
And CUDA is available on PC laptops at a very hefty price. I'm trying to avoid switching to Premiere.
At some point Premiere will leverage OpenCL. It is a nice divisive point to bring up into these threads that tend to split into AMD/ATI vs Nvidia debates ( kick-started by SemiAccurate reports that seems inevitable. )
