There have been comments (by Dogie I think, or someone from Bitmain itself) that they are going to skip the S6 and go straight to an S7. In any case, if they came out with another water-cooled rig it would be probably be called the "C3" in keeping with their current naming progression.
Buying an S5 and needed power supply right now is VERY VERY IIFY on ROI unless you have 5 cent/KWH or cheaper power. The size of difficulty increase is going up since all of the "next gen" announcements to date, and looks likely to go up faster for a few months or a bit longer than it did the first half of this year, with the widespread adoption of "next gen" mining gear in the big mining farms as the gear becomes available.
I'm hoping for an S7 that does over 2TH for under 600 watts and $600 or less, otherwise ROI is going to be very iffy even WITH cheap electric due to the bitcoin halfing next summer (unless bitcoin climbs a LOT somewhere between now and then, or the current rate of difficulty increase drops by at least half of the last couple of incriments for the next year). I'm also hoping that Bitmain goes back to making the voltage adjustable, rather than the S5's "chained power" setup - that would be worth an extra $100 easy, as you could undervolt later on and keep the mining profitable for a lot longer or at higher power costs (an SP20E can be profitable at 18 cents/kwh if you undervolt it enough, an S5 right now loses money at that power cost with current difficulty and bitcoin pricing).
2TH at 400 watts or 3Th at 600, might be able to sell as high as $800 and still manage to ROI at some point, but I suspect Bitmain is going to release the S7 with a chained-power design at or very close to the highest rated voltage the chip can handle (just like the S5) at the cost of efficiency and somewhat of a loss of long-term ROI ability.
What I'd LIKE to see is an "S5 Lite" model, still using the BM1384 but chaining 18 chips - should run about 800Mh/s at less than 350 watts - in the $400 range. THAT would ROI no problem, and still be profitable after the halving for quite a while even if Bitcoin price stays flat and difficulty increases stay in the RECENT 2-3% on average range.
That sound like wishful thinking to me =/
There's little incentive for them to develop(or sell) such effective miners. Truth be told, a jump from .5 to .29 or less stock is pretty big and probably expensive to develop versus the alternative.
There's sure to be a gain in efficiency over the "last generation" but the big ASIC market demand is to miners with cheap electricity. Big datacenter located in China and other few places in the world where electricity is 4cents~/kwh. Then a smaller demand from smaller miners with also cheap electricity. And then a even smaller demand from smaller miners with no so cheap electricity (8+ cents).
So all in all, maybe we'll get a expensive unit with high efficiency, but i wouldn't bet on it, since its not that lucrative for Bitmain anymore.