Richard Cownie (tich.delete@this.pobox.com) on February 2, 2013 6:07 pm wrote:
> Richard Cownie (tich.delete@this.pobox.com) on February 2, 2013 5:05 pm wrote:
>
> > Other than Apple and Samsung, those other companies are barely making a profit.
> > So yes, they'll buy their SoC's from a variety of suppliers, but I think not
> > at margins that would look attractive to Intel if the products have to come from
> > near-leading-edge fabs. They can't afford it.
>
> A few estimated numbers:
> Tegra 3 is about 80mm2, I've seen ASP estimated at $21
> Estimated Intel laptop CPU ASP - $94
> dual-core Ivy Bridge + GT1 die size - 94mm2
> Medfield die size estimated 62mm2
>
> The laptop ASP include dual and quad chips, so maybe the duals are only $60.
> That would imply ASP of $60/94mm2 = 0.64 $/mm2
>
> To match that, Medfield would need ASP of 62*0.64 = $39.68
>
> I don't see how that works when a high-end competitive ARM-based chip
> is about half that price. And as far as I can see, it isn't working.
>

Ivy Bridge is 22 nm, Medfield is 32 nm.
Squeeze Medfield to 22nm and it'll be some ~45 mm², which would bring the target ASP to 45*0.64 = $28.8, using your methodology.

And there's a bunch of factors.
On one hand, Intel may be able to command a higher price than NVIDIA, if they deliver a better product.

On the other hand, a 45 mm² Medfield will have better yields than a 94 mm² Ivy Bridge and thus lower cost/mm².

Finally, what Intel really needs to stay relevant is enough revenue to sustain their massive R&D costs. In they have to get it more from volume and less from margins, they still have a working business model.
The need for margins like their current x86 margins is not an absolute.