By: Ricardo B (ricardo.b.delete@this.xxxxxx.xx), February 2, 2013 7:25 pm
Room: Moderated Discussions
Richard Cownie (tich.delete@this.pobox.com) on February 2, 2013 6:55 pm wrote:
> Ricardo B (ricardo.b.delete@this.xxxxxx.xx) on February 2, 2013 6:31 pm wrote:
>
> > Intel's margins in the x86 business because they manage to deliver a win-win: better products
> > (performance, power) with lower production costs (area, yield) than the competition.
>
> Intel's competition in x86 has been constrained by the patent and licensing
> issues around the various versions of the x86 ISA. In practice, AMD has been
> the only plausible high-volume competitor, and when they briefly had better
> designs - even with inferior manufacturing - Intel lost share. The more
> open - and cheap - licensing of ARM IP allows many more competitors.
But when Intel straightened out it's stuff, their product was just head and shoulders above AMDs.
To be clear, my argument is not that Intel doesn't do mistakes or that "process" always trumps "design".
My argument is that as a fully integrated operation, Intel has an ability to combine both aspects ("process" and "design") in a way their foundry based competitors do not.
And when they get that combination right, their products are damn better than the competition.
> > If Intel can deliver the same double-whammy for smartphone SoCs, then
> > they'll be able to get higher margins than current SoC manufacturers.
>
> They haven't done so yet. Medfield is too big: I've seen its manufacturing
> cost estimated at $18, which would not allow it to compete with existing chips
> at a healthy margin. Maybe a shrink and some tweaks can fix that - but it
> seems just as likely that the ARM-based chips will also get better and cheaper
> (especially with the arrival of much better ARM cores).
Intel has a lot of R&D to do, a lot of blocks do design to get this right.
Not only they need a suitable CPU core, but also a suitable GPU core, memory controller, I/O, baseband, etc.
These blocks need to have the right balance between size, power, features and performance for the target market.
Intel had nothing suitable in the stable and, unlike the foundry based eco-system, they can't just go out and buy some IP and quickly port it to their process.
> Ricardo B (ricardo.b.delete@this.xxxxxx.xx) on February 2, 2013 6:31 pm wrote:
>
> > Intel's margins in the x86 business because they manage to deliver a win-win: better products
> > (performance, power) with lower production costs (area, yield) than the competition.
>
> Intel's competition in x86 has been constrained by the patent and licensing
> issues around the various versions of the x86 ISA. In practice, AMD has been
> the only plausible high-volume competitor, and when they briefly had better
> designs - even with inferior manufacturing - Intel lost share. The more
> open - and cheap - licensing of ARM IP allows many more competitors.
But when Intel straightened out it's stuff, their product was just head and shoulders above AMDs.
To be clear, my argument is not that Intel doesn't do mistakes or that "process" always trumps "design".
My argument is that as a fully integrated operation, Intel has an ability to combine both aspects ("process" and "design") in a way their foundry based competitors do not.
And when they get that combination right, their products are damn better than the competition.
> > If Intel can deliver the same double-whammy for smartphone SoCs, then
> > they'll be able to get higher margins than current SoC manufacturers.
>
> They haven't done so yet. Medfield is too big: I've seen its manufacturing
> cost estimated at $18, which would not allow it to compete with existing chips
> at a healthy margin. Maybe a shrink and some tweaks can fix that - but it
> seems just as likely that the ARM-based chips will also get better and cheaper
> (especially with the arrival of much better ARM cores).
Intel has a lot of R&D to do, a lot of blocks do design to get this right.
Not only they need a suitable CPU core, but also a suitable GPU core, memory controller, I/O, baseband, etc.
These blocks need to have the right balance between size, power, features and performance for the target market.
Intel had nothing suitable in the stable and, unlike the foundry based eco-system, they can't just go out and buy some IP and quickly port it to their process.