1.) ARM foundries are struggling with the introduction of 32nm/28nm high-K metal gate. 2.) 20nm production is even further behind. 3.) ARM's fabrication road map ends at 22nm. 4.) ARM's need to implement EUV and 450mm Wafers at the same time they implement 14nm will add years to the process. 5.) ARM doesn't have the technology to move to 14nm. 6.) The number of foundries able to produce state-of-the-art fabrication is shrinking. 7.) The number of foundries able to afford new fabs is shrinking. 8.) The fabless model is unable to assure necessary capacity when needed and is unable to provide the necessary capital funding. 9.) ARM foundries are unable to match Intel's "every eight quarter" fabrication cycle. 10.) ARM foundries struggle with production yields and ramps as die sizes decrease.
This message was approved by:
Executives of chip manufacturing equipment vendors:
Meanwhile Intel's 22nm has been put back six months from Q42011 to Q22012, yield problems ?
2) 20nm ARM designs have taped out already.
3) TSMC recognizes the need to introduce FinFETS at 14nm and below which shouldn't fundamentally be a problem considering they produced the first working one a decade ago.
4) TSMC would like to implement 450mm and EUV at 14nm but it will depend on economics and is not on the critical path so will not hold back implementation of the process.
5) TSMC has all the experience it needs to implement 14nm. It developed gate-last HKMG totally independent of Intel as well as building the world's first FinFET transistor.
6) The number of foundries able to produce state-of-the-art fabrication is increasing with Glo-Fo joining TSMC and Samsung.
7) The number of foundries able to afford new fabs is increasing with both Samsung and Glo-Fo opening new US fabs.
8) The fabless model is able to assure necessary capacity when needed and is able to provide the necessary capital funding. Samsung and TSMC are very successful and cash-rich and Glo-Fo is well supported by the rich Abu Dhabi sovereign fund.
9) ARM foundries are able to match Intel's new "every ten quarter" fabrication cycle.
10) As die sizes decrease yields go up which is basic fab101 knowledge to everyone except for the Wally fantasist ;-). ARM cores are much smaller than x86 ones and so naturally yield better.
Good explanation, but to think that TSMC and Samsung are going to just come up to speed and not face any of the problems that INTC faced is pure fantasy. I appreciate your loyalty to ARMH, but you really need to get a grip and a good dose of reality.
When the facility opens in 2013, about 1,000 people will be needed to run an operation that Intel executives suggest will be the most advanced high-volume semiconductor-manufacturing facility in the world.