|
I've just added new stocks/bonds to the swap list: PIMP, BTCMC, SYNERGY, MOORE, and MERGEDMINING
|
|
|
|
That certainly looks interesting. You could just stick those on a concrete slab with 4 walls and not need any insulation or fans at all  Yes, and then I drive up in a flatbed truck and cart a multimillion dollar container off. Nothx. |
|
|
|
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.
If Im not mistaken you can also get refrigerated containers ? You have to build them. No one is currently selling an all in one unit that handles this. This is expensive and requires knowhow, but see the above post about the self-water cooled enclosed racks that Liebert makes, which may be an equivalent solution. |
|
|
|
|
So, not only did we cross 2000 shares sold today, we crossed the 4.5 ghash line.
Fuck yeah.
|
|
|
|
Something like that, but APC doesn't seem to offer a "heat chimney" type of solution (its the other half of cold aisle containment). I want a partially closed loop solution. Cold aisle containment + heat chimney, and heavy use of economizers most of the time. However, I think there might be another solution: http://www.emersonnetworkpower.com/en-US/Products/PrecisionCooling/HighDensityModularCooling/Water-Based/Pages/LiebertXDK-W.aspxIts a water cooled enclosed rack. I know I'm not fond of water cooling in data centers, but the entire product family looks rather interesting. It basically means we can scale cooling per rack, and for customers other than mining customers it would manage cooling based on their load individually. They have a 25kW model that, effectively, I could put the entire mining company in two racks. |
|
|
|
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.
This is why I was planning for basically 4 racks of my own in the data center, and the racks won't be packed full. And yes, I have more plans. The plan listed is just phase 1, and it must be completed before I consider anything else in full. I don't want to waste money by buying too small, and who the hell is going to build a tiny hut in the middle of nowhere with a single rack in it. I don't need all $1m to start building the DC, I just need enough to keep it from financially collapsing. Depending on how BTC prices continue, we could be looking at building the solar farm first THEN the DC, I consider this an acceptable outcome. I have been considering effectively building a container storage facility, however it just doesn't seem to be worth it yet, not enough hardware. Its easier to build, say, four walls of rebar reinforced concrete with extra sound proofing measures and fill it with a bunch of locked security racks and just renting out the excess space. What I think is interesting about containers is, literally, people could just ship them in already loaded and I just have to park them in the facility. Small customers wouldn't, but the Amazons and Netflixes and whatnot would easily consider it. Although, at that point, I'd probably have to pull my corner of the Internet to me and thats pretty expensive. Containers make it a lot easier to do airflow containment, but it requires a lot of custom tooling. For what we need now, its just too much. However, a bunch of racks doing hot aisle/cold aisle with cold aisle containment and high airflow in the hot aisle would do just as well. Several manufacturers have complete containment rack solutions that prevent hot air circulation in the rack and forces it out into the hot aisle, so for now, I think this is what I'm going to pursue. Edit: Actually, I wanna ask our resident experts on something. Is there a complete solution to do both ducted hot air and cold aisle containment that is worth looking at? I know they exist and a lot of the high density DCs are starting to use them. |
|
|
|
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.
Yeah hes that guy. |
|
|
|
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.  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. |
|
|
|
|
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. |
|
|
|
|
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 (~n0n00dz4u@99-10-180-46.lightspeed.wepbfl.sbcglobal.net) has joined #dmc
[11:33:45] <n0n00dz4u> guess what
[11:33:49] <n0n00dz4u> releasing a vlog today
[11:33:56] <n0n00dz4u> discussing how much i hate this company
[11:34:06] <n0n00dz4u> and how you placed an order for 100 btc at .50 each
[11:34:10] <tsukino> oh are you that guy with the death by lollipop stuff
[11:34:12] <n0n00dz4u> you are trying to drag the btc down
[11:34:17] <n0n00dz4u> go fuck yourself
[11:34:19] <Diablo-D3> n0n00dz4u: who did what?
[11:34:19] <n0n00dz4u> good day
[11:34:22] <-- n0n00dz4u has quit (Client Quit)
[12:11:58] --> n0n00dz4u (~n0n00dz4u@99-10-180-46.lightspeed.wepbfl.sbcglobal.net) has joined #dmc
[12:12:05] <n0n00dz4u> okay here is the deal
[12:12:08] <n0n00dz4u> you're paying me
[12:12:12] <n0n00dz4u> 1000 BTC
[12:12:15] <n0n00dz4u> right now
[12:12:22] <n0n00dz4u> or I'm slamming your company on my vlog
[12:12:26] <n0n00dz4u> whats your answer
[12:12:39] <Diablo-D3> how about you explain wtf you're talking about
[12:12:45] <n0n00dz4u> no
[12:12:47] <n0n00dz4u> 1000 btc
[12:12:55] <n0n00dz4u> or I am slamming your company
[12:13:05] <Diablo-D3> if you slam my company, it will just drive people to buy more shares.
[12:13:16] <n0n00dz4u> go fuck your self you old whore
[12:13:18] <n0n00dz4u> your done
[12:13:23] <n0n00dz4u> and your getting slammed cunt
[12:13:25] <n0n00dz4u> real mean
[12:13:28] <-- n0n00dz4u has quit (Client Quit)
So, DMC has now passed the "being blackmailed for cash" milestone.
|
|
|
|
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. 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. Well, I budgeted for 40 tons. I think thats enough for 100kw of hardware, right? 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, 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. |
|
|
|
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. 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. Well, I budgeted for 40 tons. I think thats enough for 100kw of hardware, right? |
|
|
|
|
Okay, I updated the cooling study with new numbers. Standard AC wins, and it win even harder when factoring in green power generation.
|
|
|
|
, this is why I publicly post numbers,
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 . 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. |
|
|
|
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. |
|
|
|
I would guess the solution would be to provide a mix of cooled air and airflow. Water cooling might also be an option in this case, since it's cheaper, long term, than running A/C. You can also custom design it so that you're just keeping cases cool, which will suck a LOT of the excess heat out and ease the pressure on the A/C. Actually in a case like this, custom heatsinks should be considered. A block of copper which extends out of the case and connects to a central heatsink is a phenomenally good idea. You only need to blow on the main block and provide reasonable airflow in other areas. 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. |
|
|
|
I have a serious performance problem with my mining setup that I'm hoping someone here can help with.
First of all, it isn't a dedicated miner - it's my desktop PC at work, running Windows 7 64-bit. I have a pair of HD 6670's in it, one is driving two monitors, the other is dedicated to mining. Before I screwed it up today, I was getting decent performance - about 110-115 Mhash/sec from both cards (the card driving the monitors was set at dynamic, the other card at 9). I was running Catalyst driver version 12.4 and after doing some reading here decided to downgrade to 11.6, the recommended version for mining.
So I uninstalled 12.4, rebooted, installed 11.6, rebooted again, and suddenly my performance plummeted to about 30-40 Mhash/sec on the card driving the monitors and about 13-14 on the secondary card! I found out that I needed to disable ULPS, so I did but that didn't make any difference.
So then I decided to revert back to the driver version that I was running before that was getting decent performance (even though it isn't the recommended version). I went through the whole uninstall, reboot, reinstall, etc. -- making sure I removed every trace of Catalyst I could find (including all "ATI" and "AMD" keys in the registry) in between. I'm still getting shit mining performance. I'm at a loss on what to try next. Any ideas? I'll be happy to post my config and any more info you need.
On a side note, does anyone have the FirePro V4900, and does it work with cgminer? How's its performance? I'm probably getting a new PC at work in a few weeks, and that's what's on the specs sheet for it.
Sad to see yet ANOTHER person stealing employer resources. Hope you get fired ... A couple people reported this post to me. Honestly, I can't see anything wrong with it. Don't steal resources from your employers, the rest of us will be more than happy to turn you in. |
|
|
|
I think you missed a digit. To cool a 100kW load you would need around 28-29 tons of cooling. I might have missed a note about this somewhere in the wall of text, just thought I'd let you know this though. Cooling 100kW with a/c is not cheap.
Thats not what wolfram alpha thinks. http://www.wolframalpha.com/input/?i=tons%20of%20heat%20to%20kwh |
|
|
|
|
(Update: I had the cooling factor wrong, lets try this again)
Lets get back to the subject of the green facility.
Geothermal heat pumps are a combined heating/cooling system that dumps heat into the ground and take it back out in the winter (presumably for heating, although nothing stops you from just using it as a heatsink only, and dumping the heat from the system outdoors in the winter).
Around here, ground temperatures 4 feet deep (approximately the frost line) peaks at 50f, or about 40 degrees below air temperature in the hottest of the summer, and drops down to 20f or about 10 degrees above air temperature in coldest of the winter.
To cool the facility for the 3 hottest months of the year (July, August, September) and given 100kw (or twice what the plan calls for) for the facility and 1 ton of heat is 3.5kw, we would need 29 tons of cooling, and factoring in cooling loss/heat gain through the building itself and peak efficiency of a unit is around 3/4ths of its maximum capacity we need about 40 tons. The other 9 months can be cooled effectively using just minimally conditioned air exchange. 8 hours a day of cooling is assumed. Also, $0.10/kwh electricty is assumed.
Cost of air exchange system and ducting is not included, and would be identical for any of the cases. Also, this represents the absolute worst case scenarios for each, actual cooling costs should be much less. These numbers are very rough, and probably slightly wrong.
EER: Energy Efficiency Ratio, or the efficiency of the system at 95f outside temp and 80f inside temp.
EER to kW/h per ton: 12000 BTU/h / EER / 1000
Standard air conditioning
Large scale high efficiency (13 EER) commercial air conditioning is about $800-1000 and 930 watts per ton. For the three months of summer, that would be about 670 kwh per ton, or about $67.00.
Total cost of ownership for 10 years per ton is $1570, or $62800 for 40 tons.
Vertical closed loop, horizontal closed loop, horizontal closed slinky loop
A typical vertical closed loop geothermal pump drilled 125-250 feet down (250-500 feet of pipe) would sink one ton to two tons of heat. It costs about $3 per foot for the complete cost of drilling and piping, or $750 per ton.
A typical horizontal closed loop geothermal pump is done using 6 foot deep ditches with solid copper coils to distribute the heat. 150 foot long 3 foot wide installations of coils (about 600 feet of copper coil) with 3 feet in between them, and can sink 1 ton of heat. Installation cost is about $1500 per ton.
A typical slinky loop geothermal pump is done using 4-6 foot deep ditches with slinky loops to distribute the heat. 1000 feet of loop sinks 1.5 tons of heat, and are put in ditches 150 foot long and 3 foot wide (or about 1.5x more per surface square foot than copper coils), with 3 feet in between ditches. It costs about $1300 per ton.
The pump costs about $900 per ton, but efficiency goes up to at least 21 EER, so 572 watts and about 411 kwh per ton for the three months of summer, or about $41.10.
Total cost of ownership for 10 years per ton for vertical is $2061, or $82440 for 40 tons
Total cost of ownership for 10 years per ton for horizontal is $2606, $112440 for 40 tons
Total cost of ownership for 10 years per ton for slinky is $2611, $104440 for 40 tons
I'm not going to consider closed loop geothermal pumps because they tend to form leaks around pipe joints and leak all their refrigerant away, especially in northern regions that get significantly cold.
Evaporative cooling
At least one company makes an evaporative cooler that does not humidify the air and runs at 40+ EER and uses about 300w per ton and costs about $1200 per ton. That'd be about 216 kwh for the 3 months of summer, or $21.60.
Total cost of ownership for 10 years per ton is $1416, or $56640 for 40 tons.
However, evaporative cooling does not perform well in humid weather, and although Maine doesn't get like, say, Florida does, its enough to decrease the efficiency of the systems.
So who wins?
Surprisingly, standard AC, assuming any of these numbers are realistic. Even over the whole green battle, I can just install more solar/wind to compensate.
Now, since we're already installing solar and/or wind, heres the numbers for standard AC with solar, assuming the cost of 670 kwh of solar (spread out over the whole year, using 1200kwh per 1kw of panel) is $391:
Total cost of ownership for 10 years per ton is $1291, or $51640.
Now, as I stated above, this would be for 100kw, or twice what we need. I specifically ran the numbers for such in the case we start renting out excess space (which I really want to do). If I want to meet the $150k soft estimate, the building and land is going to have to cost around $100k.
|
|
|
|
|
Show Posts
