The MCM package is installed on something resembling a mini-ITX motherboard, due to the vertically-aligned DIMM slots (two of which exist here; it could also still be a NUC-class motherboard, though.) The differentiation between Intel and AMD’s part of the design is clear here, with Intel’s own press release mock-up of the configuration; the Intel CPU (a quad-core, eight thread part) is located to the left of the package, and is connected by Intel’s EMIB (Embedded Multi-Die Interconnect Bridge) to the AMD VEGA GPU to its right. The last piece of silicon to the right of AMD’s GPU is the packaged HBM2 memory, which should be stacked at up to 4 GB in these solutions: two of which we’ve already seen some benchmark numbers of.
It’s likely that Intel will keep using its already existing CPU designs (which include an integrated GPU) with AMD’s GPU solution. There are many good reasons for this, but two of those are particularly relevant: first, dual-graphics solutions for low-power, office and Internet scenarios powered by Intel are commonplace in the industry, and pretty much used everywhere there is an Intel CPU and a discrete graphics adapter from either AMD or NVIDIA. There’s really no reason for this to be any different here – power savings from this particular design combo have been proven time and again.
Second, ditching its integrated GPU would mean that Intel would have to manufacture these CPUs (which it will likely appropriate from its current mobile CPU designs) in a totally new rendition, absent of any graphics execution units; this would be counter-productive in some ways, not the least of which is the need for a new manufacturing line, which just doesn’t make much sense in its cost-to-return ratio. Why would Intel go to that effort, when they can simply make use of their mobile-bound CPUs, is unclear. That said, only Intel know for sure right now, of course – we’re all just making some educated guesses.Sources: Bits and Chips’ Twitter, ChipHell, via NotebookCheck