I found this comment from Sean Walsh, Founder of BlockC in BitcoinMagazine interesting.
“What I can tell you is that Avalon was offered the 14nm Samsung chip many months ago, and they declined,” explained Sean Walsh, Founder of BlockC, in an interview with Bitcoin Magazine . “The 14nm process node is very new, very slow to design/produce, very difficult, very expensive and doesn't currently yield efficiency gains that even come close to compensating for all this. Avalon will produce at smaller process nodes, but only once it actually makes financial sense to do so.”
Which just means Avalon is going to be VERY late to the party at 14/16nm given lead times on development and that some folks like Bitfury, KnC, and probably BitMain and Innosilicon are probably already starting design work on their full-custom 14/16nm generation of chips (and in the case of BitFury and KnC have already announced tape-out or production on said generation).
S8 efficiency will be very close to, or the SAME, as S7 efficiency, as it will be using the same chips - IF BitMain bothers releasing an S8 model.
I'm still betting on an S7+ instead.
I'm guessing that the next major upgrade will be an S9 model, ballpark end of 2016 timeframe, using 14nm full custom at ballpark 0.05w/GH efficiency.
I DON'T think Bitmain is likely to get their next generation out before the halfing, or in significant numbers if they DO manage to do so.
Yield issues on the 14/16nm process are pretty significant to date.
Bitmain announced a while back that there will be NO S6 Model.
The S7 should still be profitable after the halfing - but only marginally so and dependent on VERY cheap electric to manage a profit.
Ditto the Avalon 6, but it's going to depend a lot on if either of them handle undervolting well if at all.
I suspect the B-Eleven will be in the same boat once it's released, for the same reasons.
Even the SP50 isn't going to be very profitable at that point, despite it's slightly better efficiency than anything announced to date.