1) Solar energy is NOT "free". It takes a MAJOR investment to build a solar plant, and the average lifespan and cost turns out to be fairly expensive.
   You also have to deal with the "no sun for half the day" issue, which kicks the costs up a lot more.
   See Point 3 for why "grid tie" is pretty much impossible in most of that area.
2) COOLING. Try keeping a miner COOL in the normal daytime environment in that area.
3) NO INFRASTRUCTURE. Net is possible via sat (Iridium at least covers that area, though it's bandwidth might be marginal for mining).
4) Political instability is very common in most or all of that area.

Some of the server-based power supplies should have a nice long lifetime, but unless you're getting them used (and likely to have shorter lifespan) they tend to be VERY expen$ive.


 IMO of standard ATX power supply makers, Seasonic is the best.
 EVGA's Supernova G2 series has generated a pretty good reputation though, and are usually a bit less expensive.


 A standard 15 Amp US power outlet/circuit can handle a 1600 watt power supply comfortably (ignore the "derate by 20%" crowd, NEC already rates wiring and outlets VERY conservatively - it's the POWER SUPPLY that commonly wants to run at 80% or less for efficiency and better longevity).

 In theory, a standard 20 Amp 115VAC circuit and outlet (common in commercial buildings and sometimes in home kitchens or as a washer circuit) could handle over 2000 watts easy, but nobody makes power supplies for that as it would be too small of a niche market - most PS makers figure if you need that kind of power you can run 220V or already have 220V circuits available.

Nothing AMD can match the hash/watt of the Maxwell based Nvidia cards on X11.

 Perhaps the next AMD generation will fix that.

They already announced delivery date of around 30 January for their next batch - and FINALLY dropped the price some, though not enough to matter.

Somehow I don't believe that Mr. Lee knows EVERYTHING about all Chinese farms, especially when he doesn't operate any of them.
 I've seen a couple comments from at least one Chinese farm operator about how cheap (NOT FREE) their electric was and why - I'll take THAT comment over the comments of someone that DOES NOT OPERATE A FARM AND HAS ZERO DIRECT KNOWLAGE.



 Try crunching the numbers - even the S7 has ZERO chance of achieving RoI unless your electric is VERY VERY cheap due to the high Bitcoin difficulty increases, while A2 units just keep plugging away on pretty much FLAT with small bounce-around Litecoin difficulties that have not significantly increased since July and only went up a little in the 3 months before THAT.

 No, I don't "believe in Litecoin" - but right now mining gear for it is a TON more likely to achieve RoI than Bitcoin gear, which is ALL overpriced and almost all of it almost certain to become unprofitable even WITH very very cheap electric mid-summer when the halfing occurs, or in the case of the S7 and Avalon 6 (and the B-Eleven if it actually shows up this month or next) probably a couple months later after having been barely better than break-even after the block reward halfing.

CRUNCH THE BLOODY NUMBERS.

 I've done it repeatedly - and they keep coming up "3 cent electric you might have a faint prayer of RoI on an S7, *IF* the diff increases don't start jumping even more in a few months when 14/16nm gear hits the Bitcoin market" - AND if Bitcoin's currently inflated pricing holds up long enough.

 A2's will be profitable on cheap electric even if the difficulty TRIPLES - which I don't see any real probability of happening in the next 6 months, and fairly low probability of happening in a year.

Most of my S5s had a 2-3 degree difference, but one (the one that runs the BEST oddly enough) shows 8-9 degrees all the time.

 I suspect part of it is just "how the sensor got positioned", some of it is "how well the heat sinks are conducting heat from the chips", and the rest is just normal manufacuting tolerances in the parts.

 As far as ASIC for Cryptocoin mining goes, the next generation (14/16nm full custom) will be the last for 3+ years, as ASIC will have finally caught up to current "state of the art" for semiconductor manufacturing.

 Currently have 5 A2 units, 4 of them 110 Megas the other an 88, mining Litecoin.
 Currently have 5 Antminer S5 and a Spondoolies SP20, but just sold one of the S5s (will be shipping it out later today) and working on selling the rest of that gear.

 Investment in the Bitcoin gear (not counting power supplies, which I can recycle to next-gen gear) was about $3600.
 Investment in the Litecoin gear was about $3700, though I had to replace one power supply in one of the A2 Megas which added about $200 for a good Seasonic X1250 unit bringing the total in the Litecoin gear to about $3900.

 I consistantly bring in a little more than 4 Litecoin a day over the last 2 months (appx. $14 at recent pricing) with small variation due to the diff bouncing around a little.
 Bitcoin is somewhat more variable but has been averaging about 4 days to generate 0.1 Bitcoin over the last couple weeks (about $45 at recent pricing) but the last diff increase will drop that about 10%.


 While the KNC Titan is the most efficient Scrypt miner in existance right now, the reliability on those things tends to be VERY problematical - many many reports of cores DYING even when not overclocked.
 I'd definitely go for an A2 unit over a Titan even though the A2 is somewhat less efficient.
 The Alcheminer 256 is almost identical efficiency to the A2, I'd consider getting one of those if the price/MH was comparable or less AFTER including the cost of the power supplies to run the unit.


 There are *2* companies that have announced they are working on next generation Scrypt miners - Innosilicon and Alcheminer.
 The Innosilicon project is quite a bit further along, and while both companies have good records of actually delivering hardware Innosilicon has a better record on that score.

 Technically, SFARDS still has that SF100 dual-miner thing of theirs supposedly available, but it's not actually been buyable (except PERHAPS for large industrial-scale purchaces) for quite a few months now.


 Chinese mining farms do NOT have "free" electric, though most of them seem to have VERY VERY CHEAP electric.
 The hashrate didn't drop because most of the hashpower is on A2 units THAT ARE STILL PROFITABLE EVEN WITH NOT ALL THAT CHEAP ELECTRIC.
 The A2 will still be profitable if you have 3c/KWH or cheaper electric even if the hashrate tripled (it would be break-even at appx. 4x the current hashrate) as long as pricing on Litecoin stays in the post-halfing $3+ range.

 This is a very BAD time to invest in SHA256 ASIC hardware - the probability of ANY of the currently-available gear ever achieving RoI is pretty much zero, unless you have VERY VERY cheap electric and even then it's iffy. I'd WAIT 2-3 months for the next gen gear to start hitting the market on anything SHA256-based.

I'm currently selling off my Bitcoin mining gear, in anticipation of the 14/16nm generation showing up and buying some of THAT gear later this year.

At 0.07 Cents/KWH, it is VERY VERY iffy if you will ever earn enough back on an S7 to pay for it.

 I do my own figuring at my appx. 7.5 c/KWH incrimental electric price, and have NEVER had a calculation indicate that an S7 would pay for itself - close but not quite at best and recently not even all that close with the high difficulty increases.

 I don't anticipate difficulty increase level dropping anytime soon - might see a short-term DIP right after the halfing, but very doubtfull that it will drop much if any under 10% increase per diff adjustment before than.

 Does not exist nor has it existed for over 2 years.

 ASIC on Scrypt have pushed so far past GPUs that there is no such thing as a GPU that can profitably mine Scrypt - even with FREE electric, you'll probably have the card die before it manages to mine enough to pay for itself.
 For that matter, the Gridseed GC3355-based ASIC gear is to the point it's not profitable unless your electric is VERY VERY cheap - despite being a lot more efficient than any GPU ever was at mining Scrypt-based coins.


 If you look into non-Scrypt non-SHA256 coins, THOSE can still be mined for a profit in some cases by some GPUs.
 DASH would be a good choice for Maxwell-based NVidia cards given recent X11 optimisations for them, not so good of late on anything AMD though probably profitable enough on free or VERY VERY cheap electric.

The price isn't actually bad, if it actually exists AND meets that 10TH at 0.1J/TH efficiency specs.

Unlike existing miners, it should remain profitable for a long time after the halfing this summer, even with the competition it should face.

 Longer of late, but why are you ASSUMING they didn't start working on 14/16nm quite a few months back?


 S3+ and S4+ were totally different concepts from the S5+, they don't really count the same way - though I had forgotten about them, and they are + designated models, just not the S5+ type like the S7+ speculation was talking about.


 I will NOT be supprised if there is an S9 model released before the halfing, in response to releases or announced releases from other manufacturer(s).

Was the A2 a full custom design?

 Bit more than a node shrink if it wasn't.

 Bit more than a node shrink if it was - FinFET is a significantly different technology than what has come before.

Don't bother trying to mine Litecoin (or any other Scrypt coin) with anything but an ASIC.

 GPU mining of Scrypt coins for a profit became impossible a couple years ago.

 Right now, the most efficient Scrypt miner is the KnC Titan, but reliability on those units is VERY iffy in the long-term.
 The Innosilicon A2 based units are fairly close on efficiency, along with the Alcheminer (and OEM varients), but the Innosilicon has a longer-term proven track record of reliability over the Alcheminer.


 Things are going to change sometime in the next few months though, as 14/16nm tech arrives on the Litecoin front.

I doubt that 2016 will be boring for Bitcoin. Too much going to happen - the block halfing, the introduction of 14/16nm tech, the continuing China economy meltdown, the EU talk about cracking down on "terrorist" usage of Bitcoin (and probable badly written law that will either be too narrow to have any effect at all, or will be too wide and will fuck with folks with ZERO connection to terrorism), etc.


 2017 on the other hand I suspect will be boring.

The S5+ was the first and so far ONLY " + " model out of Bitmain.

 It is SPECULATED that they might release a S7+ model (the controller board supports enough boards) but nothing official has ever been mentioned.

 I suspect that 28nm is going to become obsolete before Bitmain gets around to releasing an S7+ model, as they seem to be selling as many regular S7 models as they can make currently.

 2U and 80mm fans are NOT a SP50.
 Might be an SP31 or SP35 - if the units are real.

 It would be nice if they would resolve their new inconsistancy - 100W/TH is 0.1 J/GH not 0.25

 On the other hand, just makes me more inclined to think "SCAM" when they can't even be consistant on basic specs.

 250 watts per PCI-E 6-pin connector is within the connector specs - the SP20E runs up to 288 watts/connector quite reliably. It's the morons at KnC that were trying to push over 300 watts/connector EXCEEDING the connector specs that had meltdown issues.

Innosilicon doesn't have and never has shown the "hype" mentality of most cryptocoin ASIC producers.
Consider how they introduced the A2 - AFTER they had it in full production.

 I suspect it's that they don't RELY on cryptocoin chips, they had a long-established chip business before they came up with the A1.


 On the other hand, they have specifically hyped that the A4 is "proven technology" in their press release about it, and SPECIFIED it had already taped out.