ASICMINER: Entering the Future of ASIC Mining by Inventing It

[smilence]

immersion cooling is useless..

Care to elaborate or are you just blurting things out of your ass?

Just 1P ,  also allied control doesn't belong to Firedcat..

http://www.allied-control.com/about-us/allied-control

It was my understanding that those immersion cooling tanks WAS Asicminer.  We've seen photo's of a shareholder touring the Hong Kong high rise and taking pictures of himself between the racks.

That is a currently hashing Gen 1 bitcoin mining operation by Asicminer.

https://bitcointalk.org/index.php?topic=99497.msg5867787#msg5867787
Not fast enough...

In order for the first wafers to have been delivered already, tapeout happened 6+ weeks ago.

immersion cooling data centre was Asicminer, but this company "allied-control" is not owned by Friedcat,

also the total hash power of that data centre  is just ~1P


and other companies are mining btc by using 28nm chips ,   Friedcat is still testing their chips.

[smilence]

only problem is TIME, nothing else, the Gen 3 chips just taped out ..

Source?

https://bitcointalk.org/index.php?topic=99497.msg5867787#msg5867787


Not fast enough...

Nowhere does it say in that post that they just taped out. In fact it says packaged chips will be ready by tomorrow.

And how exactly is 10 times space density + saving $65,000/month in electricity useless?

It was my understanding that those immersion cooling tanks WAS Asicminer.

Video from allied control clearly says customer: asicminer

What I am trying to say is that TIME is more important than electricity cost

Also the immersion cooling data centre cost too much

[jimmothy]

immersion cooling data centre was Asicminer, but this company "allied-control" is not owned by Friedcat,

also the total hash power of that data centre  is just ~1P


and other companies are mining btc by using 28nm chips ,   Friedcat is still testing their chips.

Nobody claimed allied control was owned by FC only the datacenter.

AM 40nm chips will be cheaper and more efficient than those 28nm chips.

I would go with 1PH in the size of a shipping container using only 500KW over a massive datacenter using 1500KW any day.

What I am trying to say is that TIME is more important than electricity cost

Also the immersion cooling data centre cost too much

Not true. Immersion cooled datacenter cost $100,000 and saves $65,000 per month.

Time is only important if you are using hardware that is going to be replaced.

At 0.2w/gh AM hardware will be doing the replacing.

[KS]

They're both important.

Time to market because of the crazy high diff change (more profit per GH/s when you deploy sooner rather than later).

Immersion cooling because of the electricity and space savings (which can also be mitigated by moving to places where electricity is cheap and space is not an issue but AM's datacenter space is limited and its electricity costs are not the lowest).

[smilence]

They're both important.

Time to market because of the crazy high diff change (more profit per GH/s when you deploy sooner rather than later).

Immersion cooling because of the electricity and space savings (which can also be mitigated by moving to places where electricity is cheap and space is not an issue but AM's datacenter space is limited and its electricity costs are not the lowest).

My Chinese friend rent some hydropower station for mining , better than expensive immersion cooling

[jimmothy]

They're both important.

Time to market because of the crazy high diff change (more profit per GH/s when you deploy sooner rather than later).

Immersion cooling because of the electricity and space savings (which can also be mitigated by moving to places where electricity is cheap and space is not an issue but AM's datacenter space is limited and its electricity costs are not the lowest).

My Chinese friend rent some hydropower station for mining , better than expensive immersion cooling

Why not both?

[smilence]

They're both important.

Time to market because of the crazy high diff change (more profit per GH/s when you deploy sooner rather than later).

Immersion cooling because of the electricity and space savings (which can also be mitigated by moving to places where electricity is cheap and space is not an issue but AM's datacenter space is limited and its electricity costs are not the lowest).

My Chinese friend rent some hydropower station for mining , better than expensive immersion cooling

Why not both?

Not sure what kind of "3M™ NOVEC™ ENGINEERED FLUIDS" they use  , but check this link , http://www.copsplus.com/prodnum5327.php

Is it cheap ?

[antirack]

They're both important.

Time to market because of the crazy high diff change (more profit per GH/s when you deploy sooner rather than later).

Immersion cooling because of the electricity and space savings (which can also be mitigated by moving to places where electricity is cheap and space is not an issue but AM's datacenter space is limited and its electricity costs are not the lowest).

My Chinese friend rent some hydropower station for mining , better than expensive immersion cooling

Immersion cooling is cheaper, that's the point of it.

A little bit of fluid in a bathtub with bare mining boards costs less than cases, heatsinks, fans, shelves, cables, screws, nuts, bolts, thermal grease, house, all that and people to put it all together.  You only need 200cc to cool 4kW.

If you have 1MW power and spend X kW on cooling that's X kW more mining gear you can deploy if you use immersion. No matter if you use hydro or burn coal for your power it's unrelated. Power is valuable.

And it saves possibly hundreds of board developers time to develop/test/source/buy/produce/ship all that. Miners are reduced to chips on boards, that's it.

Boards can be made smaller and have more chips on them because now there are no thermal issues.

All that in a shipping container instead of a complete building. If the IRS comes after you, you take your container and move ;-)

When new chips (28nm) come along, you throw away the old ones and stick the new ones in.

The boards make more Bitcoins because they run faster (performance increase).

The boards make more Bitcoins because they some online weeks earlier (board developers can focus on board and chips and get them out ASAP instead of doing all the thermal and other design and start looking for parts).

Unless you like Ikea shelves of course and/or you just deploy a few TH.

http://www.immersion-cooling.com/publications/Analysis_of_Large-Scale_Bitcoin_Mining_Operations.pdf

Example 2.4PH 1.2MW Mining Container (ASICMiner Gen3):

●   6 tanks per container, 200kW-240kW each
●   400 boards each tank, 500W per board, 2,400 boards total
●   Equivalent to 600 4U boxes
●   Shipping container footprint instead of 60 racks in high density facility
●   2-3PH/s at estimated hashrate (based on reports from China )
●   No thermal throttling, possible performance increase
●   Fixed and predictable cost, no matter if fully or only partially populated

[necro_nemesis]

Immersion cooled datacenter cost $100,000 and saves $65,000 per month.

I don't understand why Rockxie stated immersion cooling is too expensive in his Q/A. Start up costs versus operating costs maybe? Q. 6.

https://bitcointalk.org/index.php?topic=528464.msg5872404#msg5872404

[bitcoin.newsfeed]

I have that feeling that this new guy here smilence actually knows what he's talking about    better than us ...

[jimmothy]

Immersion cooled datacenter cost $100,000 and saves $65,000 per month.

I don't understand why Rockxie stated immersion cooling is too expensive in his Q/A. Start up costs versus operating costs maybe? Q. 6.

https://bitcointalk.org/index.php?topic=528464.msg5872404#msg5872404

After doing some calculations I don't think the stated 65k per month in savings is accurate.

At $0.15/kw
50kw = $5,400/month (air)
5kw = $540/month (immersion)

Not sure how much you would save monthly by renting a datacenter 1/10th the size.

If the savings would be more than $5000/month then the immersion cooled datacenter would pay itself off in 10 months.

[antirack]

Immersion cooled datacenter cost $100,000 and saves $65,000 per month.

I don't understand why Rockxie stated immersion cooling is too expensive in his Q/A. Start up costs versus operating costs maybe? Q. 6.

https://bitcointalk.org/index.php?topic=528464.msg5872404#msg5872404

After doing some calculations I don't think the stated 65k per month in savings is accurate.

At $0.15/kw
50kw = $5,400/month (air)
5kw = $540/month (immersion)

Not sure how much you would save monthly by renting a datacenter 1/10th the size.

If the savings would be more than $5000/month then the immersion cooled datacenter would pay itself off in 10 months.

How much would the air cooled data center cost?

[jimmothy]

How much would the air cooled data center cost?

I have no idea the rates of renting a datacenter but it cannot be cheap to rent 10 times the amount of space.

[antirack]

If you look at this it doesn't look that cheap and it's just 1.4MW
http://arstechnica.com/information-technology/2014/03/meet-the-manic-miner-who-wants-to-mint-10-of-all-new-bitcoins/

Here's the video that was around for a while:
https://www.youtube.com/watch?v=5CjldZLXiAU

Energy saving he probably doesn't care much but it would make a huge difference in operations.

You can put 1.2MW of AM gen3 in a container that is remote controlled via a smart phone ;-)

[jimmothy]

just found this:

If you look at this, it doesn't look that cheap and it's just 1.4MW
http://arstechnica.com/information-technology/2014/03/meet-the-manic-miner-who-wants-to-mint-10-of-all-new-bitcoins/

Energy saving he probably doesn't care much but it would make a huge difference in operations.

You can put 1.2MW of AM gen3 in a container that is remote controlled via a smart phone ;-)

The two buildings in Washington are about 16,000 square feet

How much does a 16,000 square feet datacenter cost vs a 1,600 square feet one.

[jimmothy]

Immersion cooled datacenter cost $100,000 and saves $65,000 per month.

I don't understand why Rockxie stated immersion cooling is too expensive in his Q/A. Start up costs versus operating costs maybe? Q. 6.

https://bitcointalk.org/index.php?topic=528464.msg5872404#msg5872404

After doing some calculations I don't think the stated 65k per month in savings is accurate.

At $0.15/kw
50kw = $5,400/month (air)
5kw = $540/month (immersion)

Not sure how much you would save monthly by renting a datacenter 1/10th the size.

If the savings would be more than $5000/month then the immersion cooled datacenter would pay itself off in 10 months.

How much would the air cooled data center cost?

To add to my calculations, I found that $10/square foot/year is about the average for renting a warehouse in my area at least.

So that equates to:

Immersion cooled: $26,000 electricity for cooling + $16,000 rent = $42,000/year
Air cooled: $260,000 electricity for cooling+ $160,000 rent = $420,000/year

[KS]

Let's say 700mm wide * 1000mm long, that's 0.7m² for a 42U rack with 2x (10A x 230V)= 4.6KW. A common price would be about 1000-1500 USD per month, electricity not included (at around 0.15 $/KW/h), or 2000-3000 USD, electricity included. Both have a 10Mbits internet connection.

DC's are really power (heat) limited and the power + heat dissipation cost is usually about 1.8x the power cost (1.2 in the very best ones).

So, in a regular DC, you need twelve (11.7+) 42U racks for a 54KW setup, or 8.4m² or about 90.4 sq. feet.

If you own the DC, the electricity you actually use will be 1.8x54KW= 97.2KW.




Example 2.4PH 1.2MW Mining Container (ASICMiner Gen3):

●   6 tanks per container, 200kW-240kW each
●   400 boards each tank, 500W per board, 2,400 boards total
●   Equivalent to 600 4U boxes
●   Shipping container footprint instead of 60 racks in high density facility
●   2-3PH/s at estimated hashrate (based on reports from China )
●   No thermal throttling, possible performance increase
●   Fixed and predictable cost, no matter if fully or only partially populated


The equivalent in a DC would cost about:
60-90K USD per month for rack space
21-26K for electricity

or 120-180K with electricity included.
(If you OWN the DC, you need to pay less than 0.083 USD/KW/h to match the heat dissipation + power cost to your power cost in another DC charging 0.15USD/KW/h).

A regular 20 feet long container is 13.72m² or 147.7sq. feet. If you pay 10USD/sq foot/year in a warehouse (assuming zero security cost):
123.08 USD per month for rent
21-26K power cost
17-21K cooling (0.8x the power cost, since a regular DC will have power + cooling = 1.8x power -> cooling = 0.8 x power).

Does that look about right?

edit:
NB: not taking parts costs into account

DC: 71-116K USD/month
Container: 38-47K USD/month

so, container vs DC savings: 33-69K USD/month more profit for container
(no reason to take into account the "power included" DC)

+typos

[antirack]

So, in a regular DC, you need twelve (11.7+) 42U racks for a 54KW setup, or 8.4m² or about 90.4 sq. feet.

I think you need to double that because doors need to open on those racks and you need aisles.

[bitfair]

In the next round of shareholder questions, I would like to ask how the crackdown on Chinese exchanges will affect the operation. I imagine the lack of Chinese exchanges would be an impediment to moving in/out of BTC/fiat that might complicate the operation.

And also, does the crackdown mean increased business risk for AM/Bitfountain? What is the risk, for instance, that an overzealous bank closes or freezes AM bank accounts?

[KS]

So, in a regular DC, you need twelve (11.7+) 42U racks for a 54KW setup, or 8.4m² or about 90.4 sq. feet.

I think you need to double that because doors need to open on those racks and you need aisles.

Good points. Though it doesn't change the rack + power costs in a DC you rent, you need to keep it in mind if you build your own DC without the containers.

When you think a container can host 5 times the power in a space 20% smaller, it would be crazy to go for a regular DC. I bet Rockxie has a lot of space and is planning his own "ghetto" DC instead (he'll need 180sq feet per 54KW, 900 sq feet per 270KW, or about ... 666 sq feet per 200KW - though 667 is closer, it's less fun ).