GF showed a roadmap at the conference call in February, where they wrote 32/28SHP ... since then I wonder if there is also a PD-SOI process available at 28nm.
I doubt it because if PD-SOI was on the plate for 28nm at GF then they'd have publicized that ages (years) ago.
Foundries are in the business of touting their offerings from on high, they don't keep that stuff hidden from the public eye or buried in obscure footnotes.
Now I'm hedging my position here because I do not know for fact that GF does not have PD-SOI at 28nm, and they do seem to operate with a bit of an irrational decision-tree matrix so it can't be 100% ruled out.
So while I can at this point it is completely unexpected and not anticipated, GloFo may pull a fast one and suddenly release a PD-SOI subnode in the 11th hour similar to how they are pulling in FD-SOI at 28nm.
Given the strong relations between GF's 32 and 28nm line, I guess that wouldnt be rocket science, would it?
It isn't rocket science to build a PD-SOI node, in fact it is easier (cheaper) to develop an SOI-based node than it is to design a bulk-Si node of comparable electrical properties.
It raises the production cost though, so there is a cost/benefits equation that you compute to determine which path makes the most sense to you as a fab/IDM.
If say, for example, developing an SOI-based node saves you $200m in R&D (a fixed one-time cost benefit) but raises the production expense of the manufactured wafer by say $200/wafer then you will make money (spend less money) going with SOI provided the total production volume for the lifetime of the SOI-based node is expected to be less than $200m/$200 = 1,000,000 (1m) wafers.
If you produce 1m wafers then your cost-savings from developing the SOI node are washed out by the sum total elevated manufacturing expenses associated with the SOI process.
Produce 2m wafers and suddenly your decision to save a penny has now made you a pound foolish as it costs you an additional $200m versus what you would have expended had you developed the bulk-Si node (at higher expense) and produced those 2m wafers on bulk at lower expense per wafer.
And that is the basic fundamental reason why Intel went bulk, and stayed bulk, but AMD went SOI and stayed SOI until the foundry days began in earnest. (it is also why foundries like TSMC and Samsung are bulk).
I know all this because I was tasked with assessing the viability of converting from bulk-Si to SOI at TI, something we were interested in doing as a means of lowering our R&D costs. But our wafer volumes were simply too high (we did all the pilot R&D work anyways, due diligence and what not, it wasn't as simple as a back of envelope calc like I have portrayed here).
So the question of 28nm PD-SOI is one of "is it easy(ier) to develop if one starts with 32nm PD-SOI", that answer is "yes". The question is "has it been the POR (plan of record) long enough at GloFo for their customers to have started making plans to use it in their designs starting 2-3 yrs ago?"
If they only just recently added it as an option then it will be a few years before we see anyone producing chips with it.