I doubt we'll see 10 nm (TSMC) or 7nm (GF) available to miner makers before 2019 or VERY VERY LATE 2018.

 The first experimental chips might show up in early 2018 or very late 2017, but there won't be capacity to spare from the BIG long-term contract customers for months after that.


 With that said - I'd predict the S11 will be around 35 TH at around 1300 watts (at the wall with the AP3++) and available in Febuary 2019 at the soonest.

 They have a lot more direct competition on the D3 coupled with the D3 being AT BEST the "3'd best miner" available from reputable sources (might be 4'th, I forget the specs on the new Pinidea model offhand).

 I suspect their pricing is mostly a result of the new IBeLink DM22 model with some input from the new Pinidea, the Innosilicon A5 isn't priced to compete at all.

 This is the first time Bitmain has been the INFERIOR product in a specific market since the S5 vs SP20 price war.

 Probability pretty close to ZERO of a new Bitmain scrypt miner in 2018 - and it would be end of the year if it DID happen.
 They'd have to move to a new process node to gain major efficiency, and THERE IS NO SUCH NEW PROCESS NODE that is likely to be available in 2018 (and for a smaller company like Bitmain even 2019 is iffy).


 People need to get through their heads that the cycle of "miner manufacturing moving to new process node every 6-8 months for major performance/efficiency gains" is OVER, miner manufacturing has caught up to semiconductor state-of-the-art and are now locked into the 3-5 year "new node" cycle of the high-end semiconductor industry - and even THAT cycle is in the middle of a shakeup as pure silicon is reaching it's end of ABILITY to go smaller.


 I doubt they care about Christmas - they care about selling the miners they CAN make when they CAN get some foundry capacity to make the chips out of TSMC.


 They MIGHT have an option for "a new model" to replace the D3 - that one appears to have been made on 28nm and there IS better process to move to on it - but the best they're going to be able to do on the L3+ (or the S9/R4/T9) is small stuff like the RX 470/480 to RX 570/580 from "minor process improvement on the same node" type gains.

 German post office owns PART of DHL, but I think they've been divesting of that over time - but very very slowly.

 Also, didn't DHL buy out FedEx a couple years or so back?

 I'm pretty sure there are only 3 basic "cards" (and some variations on those) that are 3-slot wide on the 1080ti.
 Gigabyte Aorus - which from personal experience is a MONSTER and cools very very well if you can give it space for the fans to access air.
 MSI .... spacing the name, it's their highest-end air-cooled card
 Zotac's highest-end card.

 Might be a couple I missed though.

 It's NOT needed to have water cooling or blowers in a 3-card build - they can be helpfull if you install your rigs in cases but there are SOME fan-cooled cards that also direct a significant amount of the airflow out the back of the card outside of the case - the Zotac Mini and Sapphire's "Nitro" series are good examples there.
 It IS very very helpfull for the "3'd card" to be a short card, like the Gigabyte ITX or MSI's Aero shortie, or a Zotac mini if your middle card is LONG, to let the middle card get enough air for cooling.
 If motherboard makers would SPACE THE BLOODY SLOTS OUT RIGHT that last wouldn't be an issue, but every single "3 PCI-E 16x slots" motherboard I've ever seen uses a 2-5-7 spacing which crams the last 2 cards too close together, instead of a SANE 1-4-7 spacing that would leave space for COOLING the middle card.

 I've got only 4 blower-type cards on hand at this time - a pair of Sapphire reference-type 470s (which are so-so on cooling), an Asus "Turbo" 1070 (which is BAD on cooling if you can't keep the air intake area COMPLETELY CLEAR due to the excessive blockage on the output metal bracket - the card turns into a "POOR" cooling card if you remove the bracket entirely), and an old HIS IceQ HD 7870 that's a "2.5 slot" card with a HUGE blower on it that stays very cool indeed.

 I've only got 2 1080 ti cards at this time:
Gigabyte 3-fan "Windforce" model, decent but not great cooling but 2 slots wide and CAN survive in the middle card slot if you have a VERY SHORT outside card - the Zotac does not work, the Gigabyte ITX does and the MSI super-shortie should also work.
 Gigabyte Aorus - VERY nice cooling but doesn't play well with other cards due to it's huge size, except probably in a riser rig setup with WIDE spacing.

 
 I just pulled the trigger on my next 2 "new" system builds - the Aorus along with a Ryzen and a big M.2 Ultra and probably one of the HDs from the current gaming system become the core of that, then the MB/RAM/CPU (AM3+ based) get used in my rationalization project grabbing the 3 GPUs from one of my existing FM2-based mining systems, then the AMD GPUs in the gaming system get used with that FM2 MB/APU/RAM to build a new dedicated MooWrapper system with.
 I'm not sure yet what I'm going to do with the MB that the 1080ti is currently in - it's the "failed experiment" Intel system I built a few months back, and I'm not sure if it would work with my pre-built "mining LINUX" setup (ignoring the iGPU on the G4600).



 EVGA has some very nice pricing going on right now on their B-list page - but check them against Newegg, sometimes Newegg will BETTER that pricing by a few dollars on "refurbished" card offers for the same parts, and EVGA doesn't accept BitCoin for payment.

 The previous Cryptocoin bubbles have only lasted MONTHS.

 This one is already pushing "longevity" records - but I'm not sure if it's really a bubble.

 Nothing close except for one small fire about 30 miles upwind a couple weeks back, but pretty much ALL of Washington state (and Oregon and lots of California and Idaho and part of Montana) has been getting smoked out for large chunks of this summer due to massive wildfires all across the Western States and in British Columbia.

 Air quality was bad enough today I didn't set foot outside for more than about 1 minute total all day.

 I don't think plat is worth paying more than a VERY SMALL premium for, unless your electric rate per KWH is well above 10 cents.

 Best answer though - crunch the numbers and see HOW MANY years it would take for a 2% savings (typical Plat vs Gold efficiency gain) to pay the difference in the price of the Plat over the Gold.


 The only reasons I recently bought some Seasonic SS-860 plat supplies were (1) the GOLD supplies I normally buy are out of stock and no longer made (2) I haven't found ANY decent replacement at the gold level in current production since everyone seems to be moving to that garbage "Fluid Dynamic Bearing" sleeve-type JUNK fan design (3) the plats were on sale at the time.


 I'm to the point of refiguring my "normal build" with the aim to drop power draw enough to be able to run it comfortably on a 750 instead of an 850 - I can still get Seasonic G12 series units, at least for NOW....

 https://www.newegg.com/Product/Product.aspx?Item=N82E16814125955 I believe is the Gaming version OP was talking about.

 That one is on sale sometimes around $700 (which is one of the best deals on a low-cost 1080ti when it IS on sale), clocks lower than the Aorus but it's 2-slot-width and not a LOT lower performance, while having a LOT better cooling than the "turbo" blower-type card you posted.

 Mining works FINE on a PCI-E 1x slot - it's NOT high level data intensive.

 Not much probability of them achieving ROI in a year, but they're the most EFFICIENT miner available which means they have plenty of time to achieve ROI.

 To be fair though, anyone mining with a USB stick has NEVER had a high probability of rapid ROI achievement - they are ALWAYS going to be quite a bit more expensive per hash than a "big" miner.

 I don't see Litecoin difficulty going 5x from current anytime soon. Sometime next year perhaps, and I won't be totally shocked if it DOUBLES by the end of the year.

 DASH though I see going 10x current AT LEAST by the end of October, and probably more like 15-20x by the end of the year.

 Breakeven in 6 months AT THIS POINT is unlikely barring a resumed major price rise - you needed to get into the first batch of the D3, or an "ordered months ago" batch of the L3+, to have a good chance at that fast of ROI achievement.
 Keep in mind that anything you order TODAY would not arrive 'till November, by which point diffs will be up a LOT.

 Breakeven 6 months after you GET the miners is a lot higher probability, even if prices stagnate that whole time.

 At this point, though, it's more a matter of what you CAN get an order placed for given the state of Bitmain's horribly overloaded order system.

 I'd also be cautious about HOW MUCH A/C you have available - keep in mind that a 1500 watt miner generates a bit over 5000 BTU of heat - and how much power you have available if you have many many tons of A/C capacity.


 Where the heck are you at that your electric is 3 TENTHS of a cent per KWH?

 There are perhaps a half dozen folks IN THE WORLD that have written pool software for Scrypt coins, why do you ASSUME they hang out here?

 I'm starting to think the current wallet version may have BROKEN staking - I normally get a payout every day or 2 at my current "GRC on hand" level, but it's been almost a WEEK since my last payout.

 I'm also STILL wondering what they did with the "unlock wallet" button, having to go through the settings menu to start up staking is a bit irritating and NOT USER FRIENDLY.

To be more specific, fglrx is no longer supported at all - 15.12 was the last version of it, which only supports cards prior to the Fury and *SOME* of the AMD APU iGPUs.

 Just don't get me started on AMDGPU-PRO - almost 2 years they've been working on it and it's STILL highly incomplete, missing anything resembling a usable "control" tool, and what control it DOES offer isn't documented worth beans.
 Fan control on multi-card setups often just flat out does not WORK for no apparent reason.

 I'm to the point of not even LOOKING at AMD cards any more because of the state of the LINUX drivers for them - AMD has a very long history of GARBAGE support for LINUX, and the current state is pretty bad even compared to THAT history.

1) Probably. ETH has almost NEVER been the best option for a 1070 over the last month and change.
3) Don't skimp on the power supply, a 1000 is marginal AT BEST for 6 of those 1070 models (I have a few of those specific cards, good cards but their TDP is 180 watts).
    I'd STRONGLY recommend going to AT LEAST a Seasonic X-1250 or EVGA G2 1300 and even THOSE are going to limit how hard you can push the GPUs.
    You might want to STRONGLY consider going to a 2 power supply option.
4) Unless you mine something with a VERY small total network hashrate, go pool for reliable income.
5) On ETH with those cards, probably around 180-185 IF you have enough power to feed them.
    If you have to turn them down far enough to SURVIVE on a 1000, more like 160.
6) all calculators are snapshots of CURRENT conditions, I prefer whattomine because they'll look at the "last 24 hour average" which is a lot more accurate than a one-shot snapshot.
     Keep in mind though that conditions CHANGE, so even an ACCURATE estimate will be off tomorrow as price changes, difficulty changes, etc.


 You do have a low electric rate, which should help your long-term profitability a LOT.

If wire up multiple outlets, you don't NEED to use a PDU.

 8-)

Perhaps a bit off topic, but one of the reasons WHY I insist on positive pressure FILTERED air intake.

https://photos.google.com/photo/AF1QipMYI-7BPIY5hzhHjCLureej6CYygjzr42Gk3fI4

 The antenna towers you can BARELY SEE on top of the ridge in the picture are about 1 mile from my door - and the bloody sky is GREY today.

 FULL Moon last night was a rather ominous DARK orange (it should have been bright near-white), but the smoke got WORSE today.

 I'd go with at LEAST a 1000 or perhaps an 1100 (3 cards + MB + HD + etc) and at least an 850 but preferably another 1000 (other 3 cards) - TDP on the 1080ti starts at 250, and I suspect there is SOMEONE out there that has even higher from the factory.
 Keep in mind that TDP is a TARGET, the card can AND WILL exceed it momentarily at times when the load level shifts.

 I'd EXPECT occasional issues from running JUST 3 cards on an 850, that's VERY little room for spike handling, unless they are going to be significantly reduced TDP all the time.

 An 850 is NOT going to be enough to run 3 cards AND the CPU/RAM/Motherboard/HD unless the GPUs are VERY MUCH REDUCED on TDP, and it's going to be iffy even THEN.



 A 2400 watt server PSU would be plenty, presuming it has all the ATX connections you need to run the MB with AND enough PCI-E and "riser power" connectors - AND presuming you have 220 volts to run it on.

 I've never used nvOC, but that's because I had an already-working setup I could clone before I heard of it.
 It's a Ubuntu-based LINUX distribution, so it should have standard (VERY GOOD) LINUX reliability once you get it configured and working.