I remember discussions on SemiWiki a year or two ago on 28nm yields at TSMC and there was even a guessing contest! I may be wrong on this but I think TSMC did not officially disclose their yields for 28nm nodes (or did so at a much later date).
In the absence of published data, I have seen horrible yield numbers speculated for 20nm nodes (10% to 30%) so one would think this will not get any better at 14nm. So it seems to me that the buzz on 14nm nodes is more a headline-garnering topic than a viable business one!
Ummmm, Chang said something about orders on 16nm in 2015. Still he does not say that Apple or Qualcomm have cancelled ALL 16nm orders on TSMC, moreover he say that TSMC will be a lot better in 2016 and 2017 on this node; in short words in 2015 Chang will give to customers all the capacity TSMC have decided to dedicate to 16nm for economic reasons.
Now, i think nobody knows the real story. I have the suspect about a mid 2015 nearly risk production at Samsung just to satisfy Apple and my bet is in an Apple multi-foundry approach in 2016. About Qualcomm, it will be on TSMC 20nm in 2015 so i think that TSMC will not lose much on 16nm node with this firm. Another variable is that Samsung has not enough wafer capacity so again i can't see many problems for TSMC, it is growing in revenue so well doing the devices on the more yielding process availabe at the moment, without doing strange push in hardcore spaces.
But the story has another downside.
I don't see winners here. The real winner is the foundry capable to deliver devices on a node at the higher gross margin. Samsung want to rush?? ok! still i can't see much net profit here but only an useless red flag to make expensive competition to the MUCH bigger TSMC foundry asset. I fully agreed this is an attempt to have a stronger relevance in foundry world but without a multiyears experience in FinFet lithographic design, all this could end in a sad story.
This is the reason Intel has delayed 14nm. Broadwell was ready in Q1/2014, Intel was capable to manufacture it at the beginning of this year. Still a 60/70% yield level does not fit with Intel business model......so better wait another year, the very low power 22nm FinFet process is good enough in a world in which ALL foundries have huge delay and 28/20nm will are the bulk of the volume production for others two years from now.
My Mom the Radio Star Max MaxfieldPost a comment I've said it before and I'll say it again -- it's a funny old world when you come to think about it. Last Friday lunchtime, for example, I received an email from Tim Levell, the editor for ...
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...