Topic: Diablo Mining Company - page 42. (Read 96422 times)

I didnt know Trane did enterprise/industrial stuff. I knew Mitsubishi and Siemens did, but I somehow missed seeing their DC-centric models when I was wandering around.

See if you can ask someone what the typical cost per ton is, I bet its higher than I originally planned out by like 30%. I've asked for ballpark quotes pretending to be a customer but very few have responded.

He was highlevelminer before right? I guess my "Ignore" instincts were spot on.  I only watched like 30 seconds of his vlog, but it didn't look like it was bashing DMC to me.  Weird.

Something to consider would be a modular approach through the standardized shipping container. You need three connections: power, connectivity, water(technically two, in/out). You can seal these units up so you won't have to cool an entire building. A 20' container has 19' by 8' by 8' of internal volume. Internally run a water to air heat exchanger, have a small network stack and treat each container like a mini datacenter.

In the end, if you were to achieve your $1m IPO, you'll still be looking at not even filling out one of these containers. Just putting 8 42U racks in to this container and filling all that capacity with FPGAs would easily blow through $1m. For giggles, if the BFL Mini-rig were a 4U unit costing $15k consuming 1.2 kW and generating 25 GH/s, that lets you put $150k of gear in a rack with 2U available for watchdog/switch. $900k gives you 6 racks with this stack leaving $100k for the container, power distribution, networking stack and heat exchanger. That generates 1.5 TH/s, sucks down down 72 kW of juice and another 7 kW of juice for cooling(COP of 10:1).

So yeah. When talking about datacenter stuff, $1m doesn't really buy a whole lot when 1 rack of gear costs $150k. This was 5 minutes of armchair thinking and it's already sounding like you have much larger plans in your mind than what is capable.

Of course, for a long term plan it makes all the sense in the world.

Liebert stuff is the gold standard for datacenter and telecoms, but it also happens to be the most expensive by a bit of a margin. But there are certainly other options: from Trane, from Mitsubishi, and even from Siemens, although they would probably get Stuxnet'ed lol.

Today is certainly a red letter day.

We have sold our 2000th share.

What I was looking at was for continuous operation, but I didn't see anything specifically for a datacenter.

Oh God, do want. But if I'm going to buy it, I'm going to buy brand new stuff. I'm weird like that.

lol, you should have negotiated!!

It would have made for a much funnier conversation..

I bet you could have got him down to 10 or 20 btc if you really tried..

You should have started off with 2 satoshi's..

I've been debating posting this to this thread, but might as well, #DMC is publicly logged anyhow.

n0n00dzr4u is the deathbylolipop guy.

[11:33:35] --> n0n00dz4u ([email protected]) has joined #dmc
[11:33:45] guess what
[11:33:49] releasing a vlog today
[11:33:56] discussing how much i hate this company
[11:34:06] and how you placed an order for 100 btc at .50 each
[11:34:10] oh are you that guy with the death by lollipop stuff
[11:34:12] you are trying to drag the btc down
[11:34:17] go fuck yourself
[11:34:19] n0n00dz4u: who did what?
[11:34:19] good day
[11:34:22] <-- n0n00dz4u has quit (Client Quit)

[12:11:58] --> n0n00dz4u ([email protected]) has joined #dmc
[12:12:05] okay here is the deal
[12:12:08] you're paying me
[12:12:12] 1000 BTC
[12:12:15] right now
[12:12:22] or I'm slamming your company on my vlog
[12:12:26] whats your answer
[12:12:39] how about you explain wtf you're talking about
[12:12:45] no
[12:12:47] 1000 btc
[12:12:55] or I am slamming your company
[12:13:05] if you slam my company, it will just drive people to buy more shares.
[12:13:16] go fuck your self you old whore
[12:13:18] your done
[12:13:23] and your getting slammed cunt
[12:13:25] real mean
[12:13:28] <-- n0n00dz4u has quit (Client Quit)

So, DMC has now passed the "being blackmailed for cash" milestone.

If you need A/C units, get ones designed for continuous datacenter usage. They can be had used and refurbished for excellent prices. Liebert makes some of the most efficient models available, if you get their newer CW models, since they use a set of VFDs for continuously variable cooling capacity. http://www.emersonnetworkpower.com/en-US/Products/PrecisionCooling/LargeRoomCooling/Pages/LiebertCWChilledWater-basedPrecisionCooling26-181kW.aspx

CW models use a chilled water arrangement, but they also offer the DS models with standard refrigerant.

Finally, I have a friend who has four 10-ton VH125AUAAEI units sitting in storage, with date codes from 2001 that still work great, although they might not be up to modern efficiency standards. If you are interested, PM me and I'll see if he wants to sell them. They also have a pair of TVSS for incoming line power surge protection, and some other stuff. Pics:




TVSS


Fire suppression

http://imgur.com/a/MtQaQ

Well, the plan was for 50kw of hardware, not 100kw. I'm running the numbers for 100kw for farther in the future. 50kw would be modular purely out of the fact I'd end up buying 20 ton units, which seems to be the largest size for most of the commercial AC manufacturers.

And you have the floor space about right. 24 by 24 would give me about 4 feet between between each of 4 racks in a single room and the walls. The entire DC itself could easily be fitting in the footprint of 40 by 40 if I stick with the hardware in the plan.

The reason I've set aside $150k isn't for the DC entirely, I need quite a lot of land for the green power end of the plan.

Yup.  You may want to think about how to build out in a modular fashion, it seems unlikely you will acquire enuf funds in time to build the large scale facility your envisioning.  

If you started with say a <1k sq ft building with a 3-5 ton unit on it and put 10kw of equipment in it, built in such a way that you could add on bascially the same exact footprint/wiring/hvac to it, you could build out in small chunks to the size your thinking about.  Like a 25'x25' insulated garage with a 200A panel on it would be a good starting point..  

Well, I budgeted for 40 tons. I think thats enough for 100kw of hardware, right?

actually you were off by 24x.  wolframalpha used days instead of hours for one of the numbers, I sent in an error report about it.

Okay, I updated the cooling study with new numbers. Standard AC wins, and it win even harder when factoring in green power generation.

Cooling is 120% an issue in space. Get your facts right.

And no, this isn't a cost benefit analysis. I need AC, and its in the $150k soft budget. I'm just trying to decide on WHICH kind of cooling I want. It just doesn't seem cost efficient to use "green" cooling if I can just generate my own power anyways.

Also, before I build any of this, I would have obviously hired a company to run all the numbers for real to begin with. I'm just running these numbers here for fun, essentially. I want this company to be as green as possible, but not more green than legitimately worth it.

Data uplink (and especially downlink) would have too high latencies though, also maintenance costs are relatively high.

A moon based operation on the other hand (with mooncoin, a 1:1 bitcoin clone, just local to the moon to have lower latencies) could work out great!

which doesnt explain why you wouldnt even have done the most basic cost benefit analysis before IPO-ing your so called $1M company.

In that sense, Im surprised you havent considered the most obvious solution for housing, cooling and electricity; put your rigs where no one can steal them, where solar energy is far more efficient than even in the Mojave desert and where cooling is a non issue: low Earth orbit.
Space-x is aiming for $1100/kg, if you want to be ambitious, do it right .

So 1 ton == 3.5 kw. That means I need ~29 tons, or I was off by 10x. I'll have to rerun the numbers. Blergh. Although, this is why I publicly post numbers, someone always catches a mistake.

The original writeup already includes the fact that air exchange from the outside can pretty much handle it most of the rest of the year. This is a worst case study.

As for watercooling... you still need to dump the heat somewhere, which is almost always done using air cooling at the reservoir. Watercooling just is not suited for a DC unless you're doing large scale complete immersion tanks, which I don't thing is ready for prime time yet.

Custom ANYTHING drives cost up.

wolframalpha is wrong.. for the kwh to tons of heat conversion

  a ton of refrigeration is 12,000 btu/h

http://www.wolframalpha.com/input/?i=btu+to+watthour

12,000 x .293 =~ 3500 watts per ton of ac.