On Tue, 21 Oct 2014 09:30:30 +1100, you wrote: >On 21/10/14 08:11, Douglas Eadline wrote: > >> Which means to me, $1.5bn is far less than they would spend over >> the next 5-10 years running a FAB. And, I'm sure they >> get a special parking spot out front for when they need >> to make wafers. (i.e. there is more to the deal than the $1.5bn dowry)
My guesswork: The $1.5 billion (which is payable over 3 years) keeps the current fab running producing the processors IBM needs for the next 3 years while they redesign their chips to work with a newer fab process. Running fabs has become a big $ cost business, and more specifically running a fab capable of producing server (or more generally non-ARM) class chips. This gets IBM out of it chicken-egg situation of designing for an obsolete fab, and running a fab for its designs, without having to put a lot more money into updating its fab. In 3 years IBM can then have their designs being manufactured on modern processes in someone elses fab. Another guess, someone will specialize in making non-ARM chips and get the business from Oracle (Sparc), IBM (Power), and AMD (x64) and perhaps the AMD and NVIDIA GPU business, with the hope that the production from the 3(5) of them will be viable given their needs are the most similar. _______________________________________________ Beowulf mailing list, [email protected] sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
