16-14nm is going to be current technology for quite a while until we shift to 12nm or whatever the next step will be, power and performance gains will be very minor in the future, and ASICs should have a marginally longer life than they currently have. That said, it's your choice, and I actually don't recommend mining right now.
10nm pure silicon or 7nm silicon/germanium alloy/matrix, depending on if the Intel vision or the IBM vision makes it into mass production first.
BOTH have said "10nm is end of the road for pure silicon" - which implies a rather long gap to get the infrastructure built up to handle making wafers that are NOT pure silicon in sufficient quantity past 10nm.
The other issue with ETH mining the last month is that the profitability has gone back into "major drop" mode due to the huge number of folks moving into ETH mining + dag file size growth dropping hashrates a little across the board for existing rigs + price seems to be stagnating at best.
I suspect there will be a short "bump" in a few months, when the DAG file gets so large that 2GB cards can't mine ETH any more, then back to dropping profitability 'till the big POS switch finally happens.
X11 is NOT a GPU option any more, can't even break even at 2c/KWH with ANY gpu due to the widespread deployment of multiple ASIC that handle X11. Baikal also handles X13/X15 and a couple other algorythms that don't have any "large scale" coins at all, write those off too.
XMR might be an option - but watch it tank on profitability hard if a bunch of ex-ETH mining cards suddenly shift when they can't do the DAG any more.
It does look like the ASIC wars are about to reopen, after too many months of BitMain having the only small-miner available new option. 8-)