Topic: Bitfury Containerized Plug and Play Datacenter (Read 7386 times)

2 million dollars investment :-)

Well I looked at facts and made the theory .

China has a unique thing in that many really big ass dams are going on line as I type.

And bitmaintech can be Johnny on the spot for the next five months filling that need.

Unless you buy into Philipma's interesting theory expounded here

https://bitcointalksearch.org/topic/bitmain-is-killing-the-business-1362106

i.e. a roadtrain of Bitmaintech knockoff versions of the Bitfury container, charging around China from one free short term electrical supply to another, as new hydro comes online and requires load testing.

Actually I think you are wrong. Exactly same density can be achieved with plain single loop water cooling, without the expensive two-phase fluid immersion cooling.

The target market is the buyer without enough engineering skills to design its own mining hardware based on mining chips individually mounted on small PCBs.

The actual strength of the two-phase immersion cooling is for example when you really have to use large PCBs, like e.g. CPU/GPU chips connected to large arrays of DRAM. With speed-of-light limitation you cannot design around that constraint and the 3D designs used in aircraft or missiles would be more expensive than the pricy immersion fluid.

Bitcoin mining has no such unusual engineering requirements. So their target market is somebody who really has no way of purchasing regular engineering skills but has large amounts of cash burning in their pockets. Only the usual suspects are left: money laundering for organized crime or some secret departments of small nation-state governments.

This first point is 100% true. Once you find the best location you can there will never be a need to move.

This second point is just as straight forward and obvious from a business perspective. Its is much more costly to use immersion cooling than it is to just have a bunch of high powered fans in a larger, spread out facility. Land is not that expensive that you need to keep your footprint low at the cost of efficiency. I could easily see bitfury paying as much per Ghs for the facility/fluids/ect as they did for the miners.

It kinda depends on what aspect of the "Miner in a Box" you focus on. In my opinion, the whole immersion cooling stuff is squarely aimed at density (i.e. TH per square foot). The rest of the container business has been done by others and was aimed at simplification of installation and setup. Both of those have additional costs over and above the actual mining hardware. We've even seen a "Mining Trailer" that looked like it could easily be move in a day if the connections were right. While interesting, there is precious little value in making a substantial mining "operation" mobile to any significant degree. Yes it make the initial install easier, at a cost, but after that.

I focused on the immersion cooling aspect of this, and it seems fairly costly compared to a more "spread out" facility with lots of air flow. Pumps, radiators, and 3M NOVEC (or Flourinert), aren't cheap at all. They also have a long term run cost compared to fans and a more hospitable natural environment (e.g. a cool/cold climate).

 Standard shipping containers are more like 9' (appx x 9' (appx) x 40' (there are also 20 foot long and 53 foot (appx) long standard sizes).

 With the added cooling stuff on top, the Bitfury container would be closer to 9x15x40 at a guesstimate.

 Dunno where you came up with 40x100 for it.


 There have been a few previous "container" concepts floated, but very few that ever saw actual hardware built much less sold.


 I suspect the container concept from BitFury is aimed more at "simplify setup and installation" than at "save real estate".

Yes, the concept is nothing original. I know a Chinese manufacturer built one years ago and has been lying in the factory gathering dust ever since because of poor ROI. Let me try to find some photos that I took from my hard drive.

It's kind of interesting to see what drives this kind of product. Obviously it's all about density in terms of hashrate per square foot (or meter if you prefer). I am wondering though, where is real estate for a mining enterprise expensive, but has low electricity costs? Sure you could decide to put mining farm in downtown Tokyo, but wouldn't the electricity cost also be large?

It seems to me that there has got to be a fairly limited market, and for a truly cost efficient mining enterprise, you need cheap electricity, modest to cheap real estate, and reliable Internet. So far it seems that this will be the mining "holy grail" for the future, and that you could go ahead and optimize one, but completely lose on the others (e.g. cheap real estate and electricity, but poor Internet).

Either way, I think it's unlikely that any of the forum members will be trying to purchase one of the containerized mining farms anytime soon. I would be interested though on what it costs, and the surrounding infrastructure items and costs (e.g. 480VAC? Chilled water? What?).

The proof of concept obviously did exist even in the days of ASICMINER. They had some pilot installation with 2 or 3 half-rack vats of Novec cooling then current ASICMINER boards and power supplies. The video used carefully staged angles (and possibly VFX) to make it look that there are whole rows of full racks in operation. The pilot installation was located in Singapore or some such are with extremely limited and valuable real estate.

The containers (steel cages) obviously do exist. The second stage water pump and water cooling radiator & fan combos are also standard industrial equipment.

I do share your doubt that they will ever close a sale or lease contract. With their sales tactics the whole thing stinks too much of snake oil. It will be hard (but not impossible) to find somebody stupid enough to plunk the money for the full installation. I was actually somewhat disappointed when I read that Bitfury acquired Allied Control.

I'm not sure they want to disclose anything really.  I view this as the bitfury lightbulb.   They put something out there to make some news, no doubt some sites showed it. 

The lightbulb they said they would sell.... and just never came.  I predict this container will not even make it as far as the lightbulb.  I think it will remain a rendering.  If they make one even as proof of concept I will be surprised.

I think you may be lowballing the cost and lead time for the required quantities of Novec fluid.

Before I knew about Novec as a cooling fluid I though that this was just another reagent used in medical industry. There I overheard people talking about borrowing a 50 gallons drum of some used Novec to keep their medical test processing line up. Novec isn't some neutral or noble fluid. This is some sort of super-solvent that can leach nearly anything out of nearly everything. Keeping it reasonably clean is a tightly controlled information, one that Allied Control/Bitfury folks will not disclose.

Anyway my take is that the Allied Control folks are their own biggest enemy. They sales tactics are such that if they were selling illicit drugs to junkies the Narcotics Anonymous would have had a banner year.

In the absence of real news, how about a repost from 2013 about ASICMINER DragonFire and ChainForker?

I thnk that Bitfury are very unlikely to sell these containers, on their webpage blurb they mention the phrase 'maintained by Bitfury' which to my mind suggest a leasing agreement. Look at it this way, if they rent oiut the equipment then they get the best of both worlds - they can get someone to essentially fund it, charge them monthly maintenance fees then take it back at the end of the lease period and upgrade it with new chips, also throwing in terms that if Btc rises in price then then get a cut of the extra profit (but not a cut in rental if it falls).

Imagine this scenario:

Btc = $400, Network hashrate is at 1500PH, block reward has halved to 12.5Btc.

1 PH earns approx US$ 11.7k per month after electricity cost of 3 US cents per kWh, for 16PH it's $187k

Network hashrate rises linearly by 100 PH per month, over a year the 16 PH container will earn a total
of roughly $1.56 million, again after electricity costs.

I estimate that Bitfury can build this containerised system for about $900 k, including 11kV input capability. Of that amount around $550k will be 160,000 chips on plug in modules, so the actual infrastructure costs about $350k. If they build the container with borrowed money at, say 8% then after a year then will have paid for the container and have around $590k in the bank.


Is it better to keep the container and mine with it themselves or charge a third party a rental fee?

From the investors perspective they might pay anything up to $1.35 million in the hope of making 15% on their money, where else could they enjoy short term potential profit at this rate? At $1.35M Bitfury makes a $450k profit on the deal less their 'maintenance' costs - if charged, but they will be using someone else's money to build the container in the first place and that might be a very inportant factor in their decision - more containers built means lower parts costs AND deters competitors. They also still have the container at the end of the years contract. The investors could sell their hashing power onto others and so increase their effective margins without using so much (if any) of their own money.

Naturally, all of this is highly speculative and depends on many factors but I though it was worth taking a stab at how things might shape up.

Any other theories / conjectures? All comments welcome

 

They are not near as impressive and are kinda dated as I think the idea of container just sounds better on paper then it turns out to be.    I don't know if either of those companies had sales.   

The second company had price of like 35k for just it being wired up, and cooling.  No miners were in that price which seems insane.   I think bitfury will stay in renderings as sales just are not there on these, unless they do it much cheaper then others did.

Some nice pictures in that thread, not quite as impressive looking as Bitfury, but at least they are real....


Rich

Took me a while to find but here is one: https://bitcointalksearch.org/topic/m.10770794

And there is: http://www.cryptokube.com/

Those are only ones I have seen actually go from rendering to actual one. I think data centers win out over these things.

Good to know, but still just an, impressive, render, or has anyone actually seen one of these containerised Miners?


Rich

Allied Control is not "other people". Bitfury acquired Allied Control some time ago.

I had a scout round to see what I could find and there are a couple of similar Containers / Data centres from other people.



Impressive looking with 1.2MW, but just another render, so perhaps just a concept?

The only other Container I could find was an Icelandic Data Centre, but at least it's real. 




Rich

I believe it is just rendering with bitfury.  I don't think it exists or they would show it off.  And I'm guessing if they do have miners they are more worried about switching in regular data center.  I dont see to many buyers of this if any.

The one company that sold containers before for mining I believe had 1 or 2 they showed off.   I think it was SP gear in it.  I can't find a link.  It was also not near as impressive as this rendering on cooling.

Do these Containers actually exist, surely there ought to be a real picture of one somewhere?

Rich

You do have a good point on B11.  Some reseller posted in hardware saying it was coming soon.   Only thing is I don't think we have seen anything besides renderings.  So they are really good at keeping secrets or they don't have it perfected yet.   

On this container there was another company that sold containers they were rediclous on price.   It did not take off they showed 1 or 2.  I suspect bitfury will be the same they don't sell many of these.   And it could be like the light bulb.... were still waiting for it to sell to customers, which looks like it will never happen.

 At very low hashrate per chip.

 Definitely a tradeoff on "cost of chips" to run more efficiently vs. "lower power/cooling costs".

 Consider the S7 and the BM1385 - if Bitmain had put more chips-per-string in the S7, it would be noticeably more efficient - but would have cost quite a bit more per TH or they would have had to accept quite a bit less profit per unit.


 BW.com should be starting deployment of the B-Eleven (B11, whichever it is THIS month), might be part of the hashrate jump the last week or two.
 The current large hashrate jump doesn't HAVE to be all about BitFury.

Disagree. I suspect the mining manufactures are not selling hardware. They are mining with it.  The best the general public can buy is the S7, which is unlikely to ever give an ROI.

The better miners, from the SP50 onwards, are not available to the general public.  No manufacturer has yet released details of a 16nm miner for potential general sale despite lots of evidence they have been around for some time.  And when they are available they are likely to only be available to large buyers.

In summary if there is one thing the miner manufactures are not doing it is selling the latest miners to the general public.

I don't think we should expect there to be any significant slowdown soon. The ASIC manufacturers want to exract every possible $$$ (mu fiat of choice), while they can. None of them know what the halving will do to the price of mining gear and their ability to sell it. Right now it's a sellers market, and they will try and feed the appetite, along with transferring any risk, to the willing buyers.

Bitmain lit the fuse back in August, and re-ignited the ASIC arms race. They make their money on the front end, and give up some possible appreciation and earnings, to let the buyer take the risk.

Sell while the selling is good.

Interesting it would make a lot of sense as difficulty has went crazy it seems right now.  Just price staying sub 400, and huge jump coming is not good for miners.  If they get them in before having ... that is huge of these new chips.   

I would think we would see them pushing hard on getting gear as Bitfury seems to have beat all other makers on time and specs, at least that we know about.  I hope it slow's down but I'm not sure what will happen.

please be aware that this chip and equipment based on it is ALREADY deployed and in use by large mine farms.  i can state for almost certain fact that the bitfury 16nm has been deployed in a farm in the nyc area for a few months now.  the size takes less than 128u rack space and is hashing 400th.  the reason i say that i am pnly "almost certain" is i have not seen it personally but was told about it by a good friend that has seen it first hand.

I think there will be a lot of that going on.  They will replace miners replaced and they might try to sell old miners to another location.

A possible example is over here: https://bitcointalk.org/index.php?topic=1231822.120   armedmilitia noticed that a operation in Venezuela's gear looked just like the one's MegaBigPower wanted to sell in bulk. 

What we see in that thread I predict more of.  Send a container of old gear to some country where electricity is super cheap and high risk, but if since you already replaced gear... high risk is still a bonus if you get more ROI.

Agree. I see technicians in large scale mining farms slowly moving along racks replacing 28nm miners with 16nm miners.  The new miners look identical to the old miners and draw the same power.  The only difference is that they have 5 times the hashing rate.

So the network hash rate grows 500% compared to say June 2015 with no extra space or power needed.  It could rise from 400 Petahash to 2,000 Petahash just from upgrading of existing miners.

In practice some new mines will open as miners move to cheaper power areas.  But mammoth new mines over those previously announced are unlikely.

Everything old is new again.

The attached article gives a detailed discussion on Bitcoin mining in a 1.2 Megawatt container data center.  

It also has details of KNC, ASICminer and Megabig power large scale mining.  

The interesting thing is that the article is from 2014 when the total worth of mining was $845 million pa and expected to grow.  Today total mining earnings are about $525 million pa and if Bitcoin stays around $400, possibly just $260 million pa after July.  

Back then mining attracted big money.

http://www.allied-control.com/publications/Analysis_of_Large-Scale_Bitcoin_Mining_Operations.pdf

I think a lot of the potential clients will be less then the .07 even.  I see this most likely going to some of the biggest operations and I would guess around half of that price they pay.

The only thing I wonder is if some of these mega clients really will want the already setup box.  I see a lot of the big players already with investment's in data centers, so likely they will want to put in a spot they already have.   I predict most of clients will just want miners.   But I could be wrong if someone is a new player in the game then they might like this.

And if it does .11 watts per gh  before cooling it is poorly designed.

the chip has demoed at 0.06 watts a gh 

 So  this does close to 2x what the chip can do on the low end.

I don't see it as a good design unless they don't sell any chips to any one else.

If they sell chips and some one builds a .08  watt   s-9  it will hurt this machine.

Now if they back off on the power or allow this to run at a solid down clock down volt.

Ie if you can run it at .07 or .08 watts a gh as an option  it is  a fine  piece of gear.

I would love to find a client to host one of these units.  This has been in the works for years, good to see it as a reality.  Immersion cooling is a growing technology, and Bitcoin Mining is a near perfect test-case for the viability of the underlying technology.  It is a very efficient package overall and in the times of rising difficulty, efficiency is key. 

-Dalkore

Nicely put.  And if a chip can do a low of 0.06 to a high of 0.11

 gearing it in In a container that produces 8,000,000 BTU of heat  and needs 400,000 CFM of fans is just plain stupid.



As for Roi.

All this setup makes a chip that can do 0.06 or 0.07 air cooled do worse then .11 since I have to fire the fans up and pumps to cool.

So I buy the monster and I find it does .14  watts a gh  with cooling costs and works great.

Bitmaintech tosses a ton of money for just the chips at bitfury and builds a .08  watt air cooled s-9  .



So by buying gear that with cooling approaches  .14 watts

I could be hurt if bitmantech puts out a .08 watt miner in small gear.

Right now bitfury needs to design the   container to do under 0.09 watts  not

.14watts with the fans pumps and cooling.

If I could come up with the coin for this I don't want it since it is designed to go high watts per gh on the chip not low watts per gh on the chip.

and bitfury could simply undercut me with a significantly better design for the same chip.  And no one will want the beast. for a second hand miner.

I assume that this harks back to bitfury CEO past. He is from Latvia/former USSR growing up during 1990 USSR collapse.
Lots of bad people, lots of various scams and swindlers, so people's mistrust, perhaps.
He probably does not want to deal with tens/hundreds of thousands of small miners, hence the "go big" mentality.
Honestly, why not just put up the chip for sale directly on the internet, then institute a volume discount for large customers.
This way, anyone can build a miner and there is no need for some byzantine integrators communication pathway scheme.

I believe they are doing it like this: http://www.datatank-mining.com/
http://www.datatank-mining.com/files/DataTank_Mining_Ops_Prospectus.pdf

If this is the case, you definitely wouldn't be a moron for buying it, if you could afford it and the ROI math checks out..

this gear is moron on BITFURY'S PART  and only a moron would buy it.

We all saw the video of these chips doing 0.06 watts per gh on a down clock..

so if they build a box at 0.075 watts per gh net  that uses 1000 watts it is a 13.3 ph box

They are complete moron for pushing this idea for now and pushing the light bulb idea  last year

When a 1000 or even an 800 watt box makes perfect sense. You know the one in the middle of the two idea's above.


This container will produce close to 8 million btu's

That is around 80 home furnace's

I estimate that the 16 fans on top of the box need to move around 400,000 CFM

true insanity.

When they can simply make 800 watt boxes = to a pair of s-7's.

I'd buy that in a heart beat actually!

one idea would be to shrink that container (maybe 100-200 fold in volume) from 40X100 ft to 8.6X21.5.4 ft or 6.8X17 ft box respectively that could fit in the backyard.
It would do 160Th or 80Th and use 20 or 10kw. Such mini-box would sell well to bitcoin mining prosumers.

I think you need to add cooling costs and the cost of creating a room for hosting many S7's.
I'm pretty sure cooling 16 PH of Bitfury's immersion cooled chips, will require less energy than cooling 16 PH of S7's with air circulation. (Pure speculation on the immersion part, but only realistic way to cool 2MW in a container IMO.)
Property cost/rent should also be in there!

well, this is borderline price assuming maybe two more months wait.
Why? S7 is $789+shipping=~840 for 4.7Th, but I will go with $789 number because there are discounts for large purchases, which will negate shipping
16000Th (container)/4.7=3404 S7 equivalents
3404X789=$2685957
It would take Bitmain to simply lower S7 price to ~$605 to match the container price per Th (although without counting better efficiency).

Conclusion: Bitfury has to lower the container price to 1-1.2 mil to have a significant competitive advantage and deploy en masse vs Bitmain.

it could cost 2.200.000$, aprox, but the most amazin thing is that you can "transport" your mine center over world with only a truck... 

The 0.1j/GH figure was air-cooled with a heatsink (and IIRC was actually closer to 0.11)

 Immersion cooled figure they mentioned would be a little less efficient, the .125 actually sounds about right for a fairly close to max pushed miner setup under immersion cooling.

can government start buying those in the future to build his own muga farm and start mining? i can see this if bitcoin get very big

if not government some unknown company that want to jump in the mining scene, it can maybe bring some decentralization

Just a couple of points:

12PH @ 1.5 MW = 0.125 J/(gh/sec)
16PH @ 2.0MW  = 0.125 J/(gh/sec)

Yes, you get power supply losses, but not 25%. I'm referring to the famous 0.1 J/gh figure.

Bitfury are also vague on the 'plug and play' angle, does this thing take 440 V 3 phase or 11kV? If it's the latter I would be surprised although they mention 'transformers' so maybe it does have about 10 tonnes of power transformer, circuit breakers, pf caps and other bits inside.

Why they are publishing and pushing this stuff is anyones guess, but you can bet your bottom dollar (or bitcoin) that they will only sell this if they can make enough profit on it to offset losses they would make from creating another competitor. Maybe they just want to dominate the market in asics by decentralising through leasing equipment to competitors.

See anything wrong with this picture?

This is purely amazing. It's also a sign that Bitcoin is going to be huge. Just think about how many people invested in this technology...

Wow .. One of these can hit 16PH!! Home miners are doomed.

Intereted to see how much THESE cost!   

http://bitfury.com/products#container-datacenter

Just hoping this isn't what they meant when they said "Selling to the public". I don't think my neighbors would like this in front of my house with a big power cable running to it....