@Bold That doesn't exactly make any sense ... Unless chip designers want to avoid the cost of redesigning the low end segment or avoid cluttering the foundries newly offered process node with order's of low margin parts so that the volume of high end parts aren't compromised. 28nm is always going to be cheaper per transistor than 40nm and that's the most important quality. 

I'm not even sure if Intel's 14nm process node will provide a cost reduction. That aspect remains to be seen since we haven't got our first die size measurements ...

40nm will only be cheaper per transister up to a point when compared to a newer lithography.

http://www.eetimes.com/author.asp?section_id=36&doc_id=1321536

Actually that belongs to 28nm ...

20/14nm is when cost per transistor will rise cause of the fact that double patterning is needed to achieve smaller feature sizes ...

No, it belongs to every fabrication process ever used and will ever be used.

There was a reason why Intel was slow to move Atom to a cutting edge lithography, why they used to release it's motherboard chipsets a lithography or two behind it's processors, it was costs and profit margins, it was simply cheaper to use the older lithography.

fatslob-:O said:

The reason why Intel was slow to move Atom for cutting edge lithography is cause they didn't want to compromise on higher margin parts but if they had some extra fab capacity then they would've without doubt moved Atom on to more cutting edge lithography instantly ...

The older technology, the cheaper it is mentality doesn't apply to chip manufacturing since newer process nodes have ALWAYS provided cost reduction aside from 28/22nm and below ...

It provide reduction in costs after the mature of the new tech... at start a 28nm process will cost more than a mature 40nm for a series of reasons: production line not optimized to the new process, issues with the process (here you can be a lot of things that make the % of good yelds be really low) and low utilizations of the process.

Most companies start making only a chip in the new process until it gets mature enough to move all chips.

But yes... a solid and matured 28nm process will be always cheaper than a solid and matured 40nm process.

The reason why Intel was slow to move Atom for cutting edge lithography is cause they didn't want to compromise on higher margin parts but if they had some extra fab capacity then they would've without doubt moved Atom on to more cutting edge lithography instantly ...

The older technology, the cheaper it is mentality doesn't apply to chip manufacturing since newer process nodes have ALWAYS provided cost reduction aside from 28/22nm and below ...

It does apply, again, up to a point.
Maturity, Die Size, Fab Capacity all play a role, it's not as black and white as you think it is.
Not to mention additional R&D is required to shift anything to a new node, which also costs.

They all have one thing in common, they are all, small and cheap dies even on older process nodes.
Atom only recently started using the latest and greatest lithography due to Intel wanting to balloon it's transister counts to be competitive against ARM.

You also only have finite fab capacity, it doesn't make sense to cut into that capacity with low-margin parts, unless there is a damn good reason.