Wind power seems to have a lot less regulation for smaller setups than hydro does - turns out that if you connect into the power grid on a "grid-tie" setup the feds can AND DO require you to licence pretty much any hydro setup, citing the "commerce clause" of the Constitution as their authority.
Might not apply if it's a really SMALL setup on a stream that can't be "navigateable" (sp) but not many hydro setups are going to fit those limits in the 1MW+ range.

 1 MW is large for wind power - the largest single windmills I am aware of are in the 6-8 MW ballpark - but it shouldn't be hard to go with windmills in the 100KW range and scale up gradually if the economics are comparable.

 LARGE windmills do have some licensing stuff to deal with though - FAA lighting regulations get involved due to their height - but I don't think they normally have other federal licensing requirements unless they get into the range that the EPA requires an "environmental impact statement".

 They seem to have radically changed their position since the last I saw.
 This ruling makes dealing with mining income a NIGHTMARE - how the heck do you know which date you "mined those coins" on and WHICH price during that date are you supposed to use?

 I know that the IRS is a bureaucracy and loves their paperwork, but this is INSANE even by their standards.

 GTX 1080 for a short while was running the same to a hair LESS than the GTX 1070, during the peak of the "price gouge" period this summer.
 500 sol/s out of a 1080 is pretty trivial, and they seem to provide a few sol/s more at the same wattage settings than a GTX 1070 once you get above 130 watts or so.

 Depending on price, ZEC hashrate/$ is generally about the same on a GTX 1070, GTX 1080, and GTX 1080 ti especially when you are comparing at the "system" level - individual sale prices and individual card variations and how you set them up to run (efficiency, or max production) all seem to have more input than the actual card model.


 The other thing I have to factor in on my setup is "infrastructure" of how the existing outlets are set up, as I don't have the option of doing significant changes to this place - but on the positive side, a lot of the existing outlets are clustered and almost every existing "duplex outlet" is on it's own dedicated 15A circuit, so I don't really HAVE to change things around to make them work viably well.
 I do have one cluster running from a subpanel I set up to plug into a 220V 30 Amp "drier" circuit, but for ease of management I set that up as 4 x 110V 15 Amp circuits near some of the existing outlets.

 Basically, I aim each rig (except the ASUS B250 mining one) to use 6 amps of power at the wall, which ends up being about 540-550 watts of GPU power draw plus the rest of the system (the CPU does BOINC work, and I have the hard drives set up for BURST mining, so all of my dedicated "mining" rigs are working at least 3 things at a time EXCEPT the ASUS and other "experiment" Intel rig I built, due to Intel drivers not playing well with other drivers for OpenCL usage).

 I may end up switching the CPUs over to mining Monero at some point though - FX8xxx aren't too bad at that, and pay off a LOT faster than Ryzens do (70% or so of the hashrate for about the same power draw on a 8320E vs a Ryzen 1700, but more like 35% of the cost).


 Most of my rigs were ORIGINALLY built to run Folding@Home, where using a bottom-end CPU is a "bad thing, kills your production VERY badly" and using anything less than a 16 lane PCI-E 2.0 slot also hits production noticeably, so they're not as optimised for pure mining use as they could be.
 That's also why most of my rigs are NVidia based - I do have some AMD rigs as well, but right now they're hammering on the Distributed.Net project via the BOINC "Moo Wrapper" project, and generating enough Gridcoin to pay the power bill for them (and not much more any more).
 I sometimes impress myself with the thought of what kind of producion I could do on D.Net/MooWrapper if I swapped all of my NVidia rigs over to that for a couple days....

 You left out the Gigabyte Aorus - any of the "3 slot" GTX 1080 ti models should be able to run pretty quiet while staying comfortably cool, though a hybrid or water-cooled setup might be cooler for the same noise level.

 You're not limited to minining skein or groetsl with those cards - you have tons of options, the most common ones being ZEC and ZEN right now, and potentially that Bitcoin Gold thing if they ever get it going.

 I'd like it a LOT better if it offered at least one more inch of space between the slots - as it is right now, it's VERY marginal for cooling on mid-to-high level mining cards.

 Forget running high-end GTX 1080 ti cards in it at all - they won't physically FIT.

 The economics of "cost per card" are slightly better on a massive rig than on multiple smaller rigs.
 It can also be a little easier managing a farm of massive rigs, as you have fewer rigs to keep track of.

 The tradeoff is the much bigger loss if a massive rig goes down, and the generally higher probability of something in a massive rig going wrong and causing it to crash.


 Some folks just like the "bragging rights"" too.

 Doesn't really match up right on that, as BOINC projects use a mix of "heavy on DP FP", "heavy on SP FP", and "heavy on integer" work - depending on which project - while the Top500 supercomputer list is strictly FP heavy (and does not SPECIFY if it is SP or DP but most folks that publish that benchmark are probably running it in SP).

 Moo Wrapper, for example, does not floating point work at all - like cryptocoin mining (for the SAME reasons since it IS work on a specific crypto algorithm) it's all integer operations, mostly integer addition and integer rotations (aka barrel shifts) which makes it impossible to compare using a floating-point benchmark.


 I've never used a GRC pool - and it's not like most cryptocoin mining where you are in a "lottery" and have to earn a block to get any earnings at all.
 A pool might make sense for folks with low GRC magnitude, so you don't have to keep months worth of your production in the wallet to get your earnings on a reasonably often basis.

 I would NOT go with a 850 on a rig with 6 of those - it leaves NO headroom and almost no margin whatsoever.
 You might get away with it if you consistently and reliably drop the TDP to under 100 watts per card with Afterburner in Windows or the applicable command-line control in LINUX BEFORE you start your mining software up.
 A good 1000 should be enough.

 The 1200 you list should be PLENTY.

 Both are out of production, but there is a "rev 2" NEWER version of the H81 that is widely available and APPEARS to be in current production.
 I have no idea where ASUS is getting the chipsets to make it with though as I'm pretty sure Intel discontinued production of the H81 chipset years ago.

GTX 1080 ti rig mining DGB-Groetsl or DGB-Skein can be built pretty quiet, if you use something like the Gigabyte Aorus cards.

 I don't, but that motherboard makes multi-PS rigs a lot easier to manage than the other current alternatives.

 I'm aiming for probably 11 or 14 cards on mine long-term.

 I paid about $150 at Newegg about a week back for the ASUS 19-slot, and have seen mention of it on Amazon around the same price.

Pretty much any GPU mineable altcoin EXCEPT FOR ETH and possibly ETC are still mineable by 2GB cards.

 Your GT 630 are pretty low end by current standards though, they're not going to mine much.

Can't speak to Canada.

 In the USA, IRS rulings are that Bitcoin mined does not become taxable 'till you either sell it for cash or spend it on "tangible property" at which point it becomes taxable at the cash value of the amount of Bitcoin you spent.

 There's a lot more verbage involved, but that's the short-form basic information you need.

 I suspect state and local income taxes (where applicable) have to follow the IRS rulings on the subject.

1-10 MW isn't "small" by hydro standards - more like medium.

 I suspect most areas you're going to have permitting stuff you have to do to build enough of a damn to feed that size of plant - unless you can set it up as a "run of river low head" type design, or your property has a sizeable stream going through it and a fairly large elevation change ON that stream within your properly.


I don't know if you would need FEDERAL permitting for that size of project - but given how crazy EPA regs have gotten in recent years, you might.

It's entirely possible to get very cheap electric in areas outside of China.

 Even in North America, there's the 3 counties of East-Central Washington state, Labrador, possibly Quebec (I've heard conflicting info about rates there), possibly British Columbia (they DO have part of the Columbia River available)....

 For all of the 1070 models I've worked with, 110-120 watts seems to be the "knee" where sol/s starts dropping off fast with TDP limit.

 This is going to VARY as a "percent" figure because the factory default TDP of 1070 models varies WIDELY - the lower-end ones like the FE and the Gigabyte ITX/Zotac Mini/MSI Aero Mini are 151 watts, many higher-end models are 180, and there is one MSI model with 2 power connectors and an official TDP of 240 watts (higher than most 1080 models and VERY VERY CLOSE to many 1080ti models).

 "80% TDP" on a Zotac Mini is the SAME power level (give or take a fractional watt) as "50% TDP" on that MSI gaming model.

 This issue also applies to GTX 1080 cards to some degree, and I believe there are GTX 1080ti cards that do not have the same default TDP as the FE cards do but I have not seen one to date.


 Sweet spot on the GTX 1080 ti models I've worked with seems to be between 175 and 200 watts, very close to 4 sol/s in that entire range and doesn't start dropping off fast 'till somewhere a bit over 200 watts - but my sample size there is small, 2 "base" model Aorus and a Windforce.

 If that's the 8+6 pin dual power connector connector 240 watt "official TDP" card, I suspect you're not seeing it actually get to that +120% TDP - I've never been able to get a ZEC miner to push mine past about 200 watts ACTUAL draw (per NVidia-SMI and confirmed pretty closely by my at-the-wall measurements) no matter what I set it to.
 I DID get it to suck 240 watts in a short test with the distributed.net client - but it ran QUITE hot doing so.

 They ARE a beast for a 1070, but I don't think they're worth the extra $$ or the pain of needing dual connectors on a 1070 - they're getting into 1080 price range most of the time and the 1080s I've worked with to date will outmine them a bit at the same power draw.

 Depends on power cost - for those of us with CHEAP power, it makes some sense to push more for max sol/s vs max efficiency.

 Also depends on how much power you have available - if you are pushing your available power limits, more efficiency makes sense even with cheap power.

 Lower TDP also helps keep the cards cooler, so they are more likely to last longer.

 Base Aorus can match that 780 sol/s with minimal tweeking and runs cool and LONG term reliable doing so (the one in the machine I'm typing this on is at 60C right now, but it's the only GPU in a gaming machine in an air-conditioned environment and a case with good cooling).
 Not sure if the higher-end model does any better, it's got a higher factory overclock but I suspect it might be limited by having the same cooling system.

 I suspect ALL of the "3 slot wide" GTX 1080ti models from the various makers will end up in the same temp/performance range.

 The RAID is redundant - you'll actually read them faster as individual drives, if you are using a CPU-based miner - all of them I am aware of assign one thread to each drive.
 I'm not sure on the JAVA-based GPU miner if it's faster as individual drives, but I suspect it might be.

 Yes, I'm aware that RAID0 AKA JBOD doesn't lose space to redundancy.