Mining Farm Insight

[dballing]

I'm in the research phase for putting together a 7 figure mining operation in January,17. In this area the electric cost is only 7.48¢/kWh along with dirt cheap warehouses.

I've got everything worked out in terms of power distribution, networking and the heat/cooling for the farm.

What worries me though is the type of mining farm based on year to year changing dynamics. Bitcoin is obviously out of the question at this point, so I'm debating between Scrypt ASICS or a GPU multi-pool farm (I've done a little bit of both already). Lastly, I have no interest in cloud-based mining and I'm already researching different solar-based systems as well so no need to mention it.

Here are my thoughts so far:

-----
Scrypt Pros:
Coin value stability. Consistent and stable hashrate. Most efficient mining hardware at the time of acquisition. ROI in a year is feasible. History of 1-2 years for new ASIC development cycle.

Scrypt Cons:
New ASICS being developed making current hardware inefficient or irrelevant. Dependence on ASIC manufactures for upgrades or part replacements. It wouldn't take much for China to expand Scrypt farms to the size of their SHA-256 farms dramatically affecting the hash difficulty. 

GPU multi-pool Pros:
The ability to always be mining the most profitable coin at any given time (Ethereum today, XYZcoin tomorrow if anything bad happens). Overall more flexibility (even down to the components acquired). The ability to simply switch out the old cards for new cards and potentially the PSUs (if needed).

GPU multi-pool cons: 
The multi-coins mined, hashrate and coin price are less consistent compared to mining Scrypt. Development cycle for new GPUs occurs more often than ASICS. An ASIC could be released at any time for the most profitable coin you consistently mined (as your bread and butter).
-----

So those are my thoughts currently and I would greatly appreciate any additional insight anyone has between the multipool vs scrypt farm debate. The other options is a 50/50 hybrid farm if I can't make up my mind.

Thanks guys! 

[MarkAz]

Having had both a relatively large BTC and LTC setups, I had far less problems with the LTC equipment than with the BTC - but that has more to do with InnoSilicon (the vendor I used gear from) vs. Bitmain.  I also had a decent amount of Avalon 6's, and I quite liked those plus service here in the US was great.  In hindsight, we could have continued mining LTC without much issue - so we sold too early on that gear, but we made the right move on BTC.  Since then we've been doing GPU, so it's not too far off of what you're looking at.

The plus of LTC is that the market is small enough it isn't really attracting much competition, and the machines could be viable for insanely long times (in mining-terms).  Consider that the KNC Titan is *STILL* the most efficient scrypt miner (better than the A4), and it's several years old.  Obviously the A4 is right on the horizon, so it's going to move the needle some, so get out your crystal ball to figure out it's effects.  In general Innosilicon has been a pretty conscientious manufacturer, and doesn't just flood the market with gear and basically implode it, like you've recently seen with X11 hardware.

On the BTC front, for me it's a no go - none of the equipment is profitable enough to warrant the cost, at least at the moment.

GPU is great right now, and I've dabbled on and off with it - the problem is it's very hit driven, so there can be MONTHS there's very little to mine that works for a large scale miner.  The flipside is that your gear generally doesn't plummet in value, so just liquidate cards, hold onto the rigs, and wait until something else pops up.

In terms of your location - 7.5c power isn't anything to write home about, but it's not terrible either.  One thing to keep in mind is how difficult it is to get a usable large amount of power to these dirt cheap warehouses you've found.  The dirty secret that most people don't talk about with large scale mining is that the challenge isn't the gear, or even the economics - it's finding a location that has sufficient usable power.  I can't tell you how many 'heavy' power buildings I looked at that would have taken $100k+ to get them usable, or after contacting the power company and getting an update site survey found out that what it's wired for isn't actually available there anymore.  Just because they have a 2MW feed from the power company doesn't mean the power company still has that capacity within that portion of their network - when load remains underutilized for a certain period, they take it back and use it elsewhere, and getting it spun back up could be all but impossible. 

The only other tip I'd give is pick a dry environment and use evap or industrial misting - the money you spend in water will be far lower than you will in moving air, and the temperature of the air directly corresponds with the amount you need to move.  Also keep in mind that if it's a region that freezes, if the system goes down, be careful about it coming back up - the freezing can cause condensation and when it warms up again in liquefies and it's bad news - you need to scale things up slowly, even using heaters in some cases.

[dballing]

Thanks Mark! I appreciate the insight.

That's exactly the intel I was looking for and you're right about the crystal ball. There's clearly aspects I like about Scrypt and GPU mined coins, but at the end of the day we don't know what the future holds for either option. It might be wise to diversify then at this point and get the best of both worlds.

These warehouses are located right next to a coal fired power plant and while they currently aren't wired for this type of operation, they can be (getting estimates currently). Another more expensive option I'm looking into 15min away is a massive underground facility that allows data centers and maintains a year round temp of 62F (basically a man-made cave). However, I've never seen or heard of an underground Bitcoin mining farm before, so I will assume there could be some challenges (such as ventilating heat).

[Amph]

i think everything under 0.1 cent is good int erm of power, right now you can do profit wiht nvidia even in the case you have 0.5 cent, yes it's not a typo, it's the reality, so 7 cent it's a really good rate

[Mzie]

What i like most about GPU Mining is that you can switch at any time to mine the most profitable coin at the moment. In your case having calculated cheap electricity and other costs, now what you need is Good gpu's that can sustain mining for the next two years. Having said this it means withing the next two years for example the DAG size for ETH/ ETC will rise and some gpu's will not be able to mine the coins. A rational decision on which choice of gpu to use is important. I would suggest RX480 8GB Version or any latest AMD Cards. This is just my opinion. Wish you all the best with your farm.

[MarkAz]

Thanks Mark! I appreciate the insight.

That's exactly the intel I was looking for and you're right about the crystal ball. There's clearly aspects I like about Scrypt and GPU mined coins, but at the end of the day we don't know what the future holds for either option. It might be wise to diversify then at this point and get the best of both worlds.

These warehouses are located right next to a coal fired power plant and while they currently aren't wired for this type of operation, they can be (getting estimates currently). Another more expensive option I'm looking into 15min away is a massive underground facility that allows data centers and maintains a year round temp of 62F (basically a man-made cave). However, I've never seen or heard of an underground Bitcoin mining farm before, so I will assume there could be some challenges (such as ventilating heat).

No problem - being located next to a power plant could be a big plus, what you might want to look into is whether or not your state is actively doing Smart Grid stuff.  Unfortunately my power company stopped supporting it externally, but if you're in one that does, then in many case if you're consuming 1MW or more and can shut down the load on demand, then they'll give you a WAY better rate.  With their contracts they can't do it more than a couple times a year, and no more than 8 hours at a time, so it more than works out - typically this is only when they're under particularly heavy load.

As far as underground goes, I really wouldn't count on it being able to overcome the heat load of a normal mine - I suspect you'd run into a bunch of problems underground, especially from the raw airflow point of view.  Keep in mind that mining has a VERY high power density, which translates to heat - so it generates much more heat than your equivalent sized datacenter.  Most of the datacenter tech you'll find is going to be a dead end for mining - it will cost you substantially more than the benefit it will gain - unless you're planning on filling some empty datacenter and get replaced by normal customers as they come online.  Just call up some of your local datacenters and get colo rates from them, and it will give you some insight into the overhead they have, in terms of power, cooling, etc.  Most profitable crypto currency mines are a bit more ghetto.   

Take a look at what they do for large scale manufacturing facilities - such as industrial misting systems.  You'd never see one used in a datacenter, but they work like a champ in mining.  I got a chance to roll through a 10+MW facility entirely cooled using them, and it was impressive how well it worked and it's relative cost.  Keep in mind this isn't like the mister in your back yard, they're under high pressure and atomize the water.  Once again, only useful in a dry climate.

[dballing]

Thanks! I completely agree that GPU mining has its perks. So then let me ask you, In terms of heat dissipation do you prefer the blower style or the bi/tri card cooling configuration. If I decide to go GPU mining or a hybrid with Scrypt the plan is to do open air aluminum frame builds with 6 GPUs per build.

With enclosed cases I prefer the single blower for cooling, but I was thinking it wouldn't matter in an open air configuration if I have box fans blowing directly on the GPUs.  

Lastly, do you guys have a backup plan if the ETH or ETC gravy train comes to an abrupt halt? The only thing that worries me currently is if these forks happen to fork again for whatever reason or if they decided to go POS ending the POW party.

Do you guys speculate another coin consistently producing the same ROI or having similar potential if that occurs?

More crystal ball theories

[dballing]

Yea, these dirt cheap warehouses I'm referring to appear super ghetto from the outside (which is perfect to keep a low profile). I'll be stressing every expenses to create the most efficient system, which is why I don't mind the facilities appearance. Unfortunately, this area does have the 4 seasons and in my own personal living experience I've dealt with water pipes breaking, so this industrial misting you've mentioned might not be possible (unless it's simply shut off in the winter). I recently saw photos of a BTC mine in China using this cooling system you're referring to and it looked amazing and efficient.  

Exactly, data centers aren't running at 100% full load 24/7 like a crypto mine would. ROFL! I was imaging this underground facility kicking me out after they notice the temperature of their 1+ million sqft facility increasing a few degrees a day...

[greaterninja]

Right now I am looking at a  place in San Francisco Bay Area and 2 places in Arizona.
Costs are 5 cents to 7.5 cents per kwh.


Ah...choices.   BTW I may be interested in taking on some partners.

[MarkAz]

We just convert as we mine to BTC - so we do speculation separate from mining, just to keep things easy from the accounting standpoint.  So I don't really care one way or the other whether it's ETH or ETC...

In terms of machines and airflow - the more controlled you can make the airflow, the better.  We design all our own cases, you can see a couple of them in some of my other posts.  I'm a big fan of plenum designs, so the exhaust air is ducted and managed in some way, and then I just generally cool the rest of the air.

In terms of CA or AZ, I can tell you that none of the big AZ power companies (APS, SRP, etc) have the smart grid programs going anymore, which is unfortunate.  No idea if they do in CA or not.  AZ is awesome for evap, CA not as much but some parts are decent.  There's also some other interesting things, like in Mesa AZ they have a chiller loop that you can get on pretty cheap, which would be wicked for cooling - but their power costs are on the higher side, although parts of the city are supplied by city power, and they're pretty flexible.  They're also pretty big on co-generation, but I didn't go too deep into it.

[QuintLeo]

7c/KWH is on the VERY HIGH side for a major farm operation.
5x/KWH is kinda marginal long-term.

 Most of the MAJOR farms I am aware of are in the *UNDER* 3c/KWH range.


 CA has large areas that are good for evap, but they're pretty much all in the southern half - Mohave desert and south, specifically including the Imperial Valley and San Diego County (except Palomar) in my experience.
 San Fran bay area is a very poor evap cooling area, too humid too much of the time.


 If you have plenty of space between the cards and good airflow into them, optimally in an open-design no case setup, multi-fan works better for cooling.
 Blower is intended more for tight spacing setups IN a case.


 Innosilicon A2 (and Lketc Dragon A1, SAME basic design) units are rock solid - it remains to be seen if the upcomming A4 units are as solid and reliable.
 Bitmain Antminer stuff *can* be reliable, but the S7 and S9 don't appear to be as well designed for reliability as the S5 and earlier units were.
 GPU based gear varies - if you set it up well, it CAN be rock solid reliable as well, but cooling is VERY critical.


 Define "dirt cheap" warehouses - $1/foot, 50c/foot, 20c/foot?

 In addition to the power availability, you also have to factor in cost to COOL - or cost to move massive quantities of air and just live with the ambient temperature, likely BOTH. Misting systems, like any evap system, REQUIRES large amounts of airflow to be effective.

[MaxDZ8]

If you are going to burn 7 digits and you decide to go GPU be sure to get in touch with some GPU developers. If there are game developers around odds are they can get you in touch with someone doing GPU stuff. I'm not following development those days but I'm still inclined to believe some algos hold surprises. With your budget you can probably get someone really badass.

[adaseb]

You do realize ETH is going POS within about 6-12 months or so.

Then there is the difficulty to worry about.

Might as well save yourself all the headache and just buy the coin directly.