ASICMINER Speculation Thread

[jimmothy]

i wonder if AM will again have a starving phase of low to no divs until sellout of gen 4, because that is one of my main concerns.

This is why I hope they set up at least one behemoth 20MW datacenter with cheap electricity so that we can have steady divs until gen4.

Don't they have a 500MW (or KW?) immersion cooled setup ready? Actually it is filled with gen1 chips, but no idea why aren't they using with their gen3 chips..

I wish it was 500MW. It's 500kw and not yet converted to gen3 hardware. (As far as I know)

It was turned off last month because electricity costs were more than earnings. At $0.15/kwh I don't really see the 2 current AM DCs being viable for long.

[RoadStress]

i wonder if AM will again have a starving phase of low to no divs until sellout of gen 4, because that is one of my main concerns.

This is why I hope they set up at least one behemoth 20MW datacenter with cheap electricity so that we can have steady divs until gen4.

Don't they have a 500MW (or KW?) immersion cooled setup ready? Actually it is filled with gen1 chips, but no idea why aren't they using with their gen3 chips..

I wish it was 500MW. It's 500kw and not yet converted to gen3 hardware. (As far as I know)

It was turned off last month because electricity costs were more than earnings. At $0.15/kwh I don't really see the 2 current AM DCs being viable for long.

Nice. Wasted money on the immersion cooling setup which isn't being used for the moment. Absolutely no preparing for the gen3 immersion cooled setup and 0 vision of the future since FC build his immersion cooling in a place where power is $0.15/kwh. Let's see on what else will FC waste shareholders money...

[jimmothy]

Nice. Wasted money on the immersion cooling setup which isn't being used for the moment. Absolutely no preparing for the gen3 immersion cooled setup and 0 vision of the future since FC build his immersion cooling in a place where power is $0.15/kwh. Let's see on what else will FC waste shareholders money...

You really have no idea what you're talking about do you?

The immersion DC has likely already payed itself off in electricity savings alone.

Btw if there is zero vision for the future at $0.15/kwh what does that mean for all of your customers who will be forced to pay that much or more?

[RoadStress]

Nice. Wasted money on the immersion cooling setup which isn't being used for the moment. Absolutely no preparing for the gen3 immersion cooled setup and 0 vision of the future since FC build his immersion cooling in a place where power is $0.15/kwh. Let's see on what else will FC waste shareholders money...

You really have no idea what you're talking about do you?

The immersion DC has likely already payed itself off in electricity savings alone.

Btw if there is zero vision for the future at $0.15/kwh what does that mean for all of your customers who will be forced to pay that much or more?

From cooling electricity saving you mean. Please show me the math behind it. How much did the immersion cooling setup cost? Is it safe to say that it was around 500k$ and 1M$? If yes then you really saved that amount of money from the cooling in just a few months of use?

FC has gen3 chips in hand for over 2 months and he is unable to use the chips in the immersion cooled setup. Right now the setup is doing nothing right? Why? Can't FC use gen3 chips at that electricity cost?

[jimmothy]

Nice. Wasted money on the immersion cooling setup which isn't being used for the moment. Absolutely no preparing for the gen3 immersion cooled setup and 0 vision of the future since FC build his immersion cooling in a place where power is $0.15/kwh. Let's see on what else will FC waste shareholders money...

You really have no idea what you're talking about do you?

The immersion DC has likely already payed itself off in electricity savings alone.

Btw if there is zero vision for the future at $0.15/kwh what does that mean for all of your customers who will be forced to pay that much or more?

From cooling electricity saving you mean. Please show me the math behind it. How much did the immersion cooling setup cost? Is it safe to say that it was around 500k$ and 1M$? If yes then you really saved that amount of money from the cooling in just a few months of use?

FC has gen3 chips in hand for over 2 months and he is unable to use the chips in the immersion cooled setup. Right now the setup is doing nothing right? Why? Can't FC use gen3 chips at that electricity cost?

Immersion cooling costs nowhere near $1-2/w. The datatank prospectus says about $0.45/w. Keep in mind that this was built in allied controls building so there are some cost savings. There would be another ~$0.1/w savings from not needing heatsinks/fans/cases. (There is also the savings from not needing to buy an AC but ill leave that out)

500kw * $0.35/w = $175,000 (I suspect it was less)

Electricity savings: 250kw *$0.15/kwh * 5000kwh = $187,500

Not sure why the immersion dc has not been filled with gen3 chips but I really doubt FC is simply choosing to wait.

[RoadStress]

Immersion cooling costs nowhere near $1-2/w. The datatank prospectus says about $0.45/w. Keep in mind that this was built in allied controls building so there are some cost savings. There would be another ~$0.1/w savings from not needing heatsinks/fans/cases. (There is also the savings from not needing to buy an AC but ill leave that out)

500kw * $0.35/w = $175,000 (I suspect it was less)

Electricity savings: 250kw *$0.15/kwh * 5000kwh = $187,500

Not sure why the immersion dc has not been filled with gen3 chips but I really doubt FC is simply choosing to wait.

Aren't the all miners requiring the same power to function no matter if they are in immersion cooling or air cooled? Where are the savings coming from? From the difference of cooling with air versus immersion cooling.

Where is the 250kw from? That's the air cooling for 500kw?

[jimmothy]

Immersion cooling costs nowhere near $1-2/w. The datatank prospectus says about $0.45/w. Keep in mind that this was built in allied controls building so there are some cost savings. There would be another ~$0.1/w savings from not needing heatsinks/fans/cases. (There is also the savings from not needing to buy an AC but ill leave that out)

500kw * $0.35/w = $175,000 (I suspect it was less)

Electricity savings: 250kw *$0.15/kwh * 5000kwh = $187,500

Not sure why the immersion dc has not been filled with gen3 chips but I really doubt FC is simply choosing to wait.

Aren't the all miners requiring the same power to function no matter if they are in immersion cooling or air cooled? Where are the savings coming from? From the difference of cooling with air versus immersion cooling.

Where is the 250kw from? That's the air cooling for 500kw?

No, miners can have a performance boost in immersion cooling as you can see with the datatank somehow managing 0.6w/gh at the wall with BE200 compared to the most efficient air cooled system I've seen (open source BE) is only ~0.9w/gh.

PUE in hong kong is 1.5 average so a 500kw system would use 250kw for cooling. In reality the savings are even more than that because of the performance boost.

[RoadStress]

Immersion cooling costs nowhere near $1-2/w. The datatank prospectus says about $0.45/w. Keep in mind that this was built in allied controls building so there are some cost savings. There would be another ~$0.1/w savings from not needing heatsinks/fans/cases. (There is also the savings from not needing to buy an AC but ill leave that out)

500kw * $0.35/w = $175,000 (I suspect it was less)

Electricity savings: 250kw *$0.15/kwh * 5000kwh = $187,500

Not sure why the immersion dc has not been filled with gen3 chips but I really doubt FC is simply choosing to wait.

Aren't the all miners requiring the same power to function no matter if they are in immersion cooling or air cooled? Where are the savings coming from? From the difference of cooling with air versus immersion cooling.

Where is the 250kw from? That's the air cooling for 500kw?

No, miners can have a performance boost in immersion cooling as you can see with the datatank somehow managing 0.6w/gh at the wall with BE200 compared to the most efficient air cooled system I've seen (open source BE) is only ~0.9w/gh.

PUE in hong kong is 1.5 average so a 500kw system would use 250kw for cooling. In reality the savings are even more than that because of the performance boost.

So by having an air cooled DC with PUE 1.3 in the same WA state where the immersion cooled Datatank is located I get this numbers (I'm not sure why you multiplied with 5000kwh):

150kW needed to cool 500kW * 0.025$/kwh = 3000$ for cooling costs per month to cool 500kW. Please correct me if I am wrong.

No idea why are you comparing an immersion cooled setup at 0.025$/kwh with air cooling setup at 0.15$/kwh. Disinformation and misguiding at it's best.

Having a DC with PUE 1.5 show exactly how bad FC planned for the future of it's mining operation.

[jimmothy]

Immersion cooling costs nowhere near $1-2/w. The datatank prospectus says about $0.45/w. Keep in mind that this was built in allied controls building so there are some cost savings. There would be another ~$0.1/w savings from not needing heatsinks/fans/cases. (There is also the savings from not needing to buy an AC but ill leave that out)

500kw * $0.35/w = $175,000 (I suspect it was less)

Electricity savings: 250kw *$0.15/kwh * 5000kwh = $187,500

Not sure why the immersion dc has not been filled with gen3 chips but I really doubt FC is simply choosing to wait.

Aren't the all miners requiring the same power to function no matter if they are in immersion cooling or air cooled? Where are the savings coming from? From the difference of cooling with air versus immersion cooling.

Where is the 250kw from? That's the air cooling for 500kw?

No, miners can have a performance boost in immersion cooling as you can see with the datatank somehow managing 0.6w/gh at the wall with BE200 compared to the most efficient air cooled system I've seen (open source BE) is only ~0.9w/gh.

PUE in hong kong is 1.5 average so a 500kw system would use 250kw for cooling. In reality the savings are even more than that because of the performance boost.

So by having an air cooled DC with PUE 1.3 in the same WA state where the immersion cooled Datatank is located I get this numbers (I'm not sure why you multiplied with 5000kwh):

150kW needed to cool 500kW * 0.025$/kwh = 3000$ for cooling costs per month to cool 500kW. Please correct me if I am wrong.

No idea why are you comparing an immersion cooled setup at 0.025$/kwh with air cooling setup at 0.15$/kwh. Disinformation and misguiding at it's best.

Having a DC with PUE 1.5 show exactly how bad FC planned for the future of it's mining operation.

Is this the datatank thread? I thought we were talking about the immersion DC in hong kong.

Of course if FC had access to a $0.025/kwh DC he would have used it.

And even at $0.025/kwh you would save ~$0.2/w per year from immersion cooling so it would pay off in less than 2 years.

Also the tanks don't depreciate in value and can be reused for next gen hardware so even more savings.

1.5 PUE is the average for hong kong. And 1.5 PUE is not even bad for an air cooled DC in a hot climate. Spongebobtechs expensive high tech cooling DC only has a PUE of 1.3 iirc and electricity costs probably not much below $0.15/kwh. (considering they are charging $0.27/kwh..)

BTW why are you now saying $0.15 is so bad when only a week ago you were saying that it is completely viable for your customers?

[RoadStress]

Is this the datatank thread? I thought we were talking about the immersion DC in hong kong.

Of course if FC had access to a $0.025/kwh DC he would have used it.

And even at $0.025/kwh you would save ~$0.2/w per year from immersion cooling so it would pay off in less than 2 years.

Also the tanks don't depreciate in value and can be reused for next gen hardware so even more savings.

1.5 PUE is the average for hong kong. And 1.5 PUE is not even bad for an air cooled DC in a hot climate. Spongebobtechs expensive high tech cooling DC only has a PUE of 1.3 iirc and electricity costs probably not much below $0.15/kwh. (considering they are charging $0.27/kwh..)

BTW why are you now saying $0.15 is so bad when only a week ago you were saying that it is completely viable for your customers?

We can mode the discussion to the Datatank thread if you wish.

You haven't replied where are the electricity savings coming from exactly because I still don't get it.

I'm not saying $0.15/kwh is bad. I only said that it's not fair to compare different setups and different power costs to justify a non-existent savings of 200k$.

Let's recap.

1MW WA DC at 0.025$/kwh:
- air cooled: ~18k$ for the power and another ~5.5k$ for cooling= ~23k$ total
- immersion cooled - ~18k$ for the power and another ~200$for cooling = ~18.5k$ total

A 4k$ saved with the immersion cooled setup for an investment of what? At least 500k$? Nice use of money!

[jimmothy]

Your calculations are not including the savings from performance boost.

Here are my calculations for a 1MW WA DC at 0.025$/kwh:

Air cooled: $200k(base cost) + $18,000/month

830TH * 0.8w/gh * 1.5PUE = 1MW

Immersion cooled: $350k + $18,000/month

1.66TH * 0.6w/gh * 1.01PUE = 1MW

Or

Air cooling

$24/gh at 1.2w/gh

Immersion cooling

$0.21/GH at 0.6w/gh

[RoadStress]

Your calculations are not including the savings from performance boost.
Here are my calculations for a 1MW WA DC at 0.025$/kwh:
Air cooled: $200k(base cost) + $18,000/month
830TH * 0.8w/gh * 1.5PUE = 1MW
Immersion cooled: $350k + $18,000/month
1.66TH * 0.6w/gh * 1.01PUE = 1MW
Or
Air cooling
$24/gh at 1.2w/gh
Immersion cooling
$0.21/GH at 0.6w/gh

It's too much disinformation in your post. It makes me puke!

I thought 1.5 PUE is in HK, not in WA state. Misleading again!

You also compare air cooling for miners with 1.2W/GH with immersion cooling for miners at 0.6w/gh even if in your example you have 0.8w/gh miners for air cooling. What the fuck?

Also I just don't think that the immersion cooling is only 350k for 1MW. That's not possible unless a very big subsidize from somebody else. Nice misleading again!

[jimmothy]

I thought 1.5 PUE is in HK, not in WA state. Misleading again!

That is the average in HK. There are plenty of 1.5 PUE DC's in washington. It costs extra money if you want 1.3 PUE but just for fun let's imagine it wouldn't:

Air cooled: $200k(base cost) + $18,000/month

960TH * 0.8w/gh * 1.3PUE = 1MW

Immersion cooled: $350k + $18,000/month

1.66TH * 0.6w/gh * 1.01PUE = 1MW

Or

Air cooling

$0.21/gh at 1.04w/gh

Immersion cooling

$0.21/GH at 0.6w/gh

You also compare air cooling for miners with 1.2W/GH with immersion cooling for miners at 0.6w/gh even if in your example you have 0.8w/gh miners for air cooling. What the fuck?

1.2w/gh is including cooling. (0.8 * 1.5 = 1.2 or 0.8 * 1.3 = 1.04)

Also I just don't think that the immersion cooling is only 350k for 1MW. That's not possible unless a very big subsidize from somebody else. Nice misleading again!

It costs about $450k per MW and as I said earlier you are saving ~$100k on fans/heatsinks/cases/etc..

[RoadStress]

It costs about $450k per MW and as I said earlier you are saving ~$100k on fans/heatsinks/cases/etc..

So if I get in touch with DataTank guys they will sell me a 1.2MW DataTank for 540k? That's a good price, but I really don't think it's possible for everyone.

Last numbers are closer to the reality finally.

Too bad AM chips need too much power for this difficulty. If we ignore hardware costs, then a 200k$ air cooled DC would beat AM chips in the DataTank. If I had the money I could easily fit 2.5PH in a 1MW immersion cooled setup at 0.4w/gh.

[jimmothy]

It costs about $450k per MW and as I said earlier you are saving ~$100k on fans/heatsinks/cases/etc..

So if I get in touch with DataTank guys they will sell me a 1.2MW DataTank for 540k? That's a good price, but I really don't think it's possible for everyone.

Yes I think you could get a datatank for about that price and that is exactly what their IPO is offering.

Even if you can't buy your own container from allied control for that cheap you can probably make your own for that price or even less. I somewhat doubt that allied controls 3rd try is the most cost/energy efficient possible with 2-phase immersion cooling. For example 3m was able to cool 4KW with only 0.2 liters of novec. If you could match that then you would only need 50 liters or $2,500 worth of fluid per MW.

Too bad AM chips need too much power for this difficulty. If we ignore hardware costs, then a 200k$ air cooled DC would beat AM chips in the DataTank. If I had the money I could easily fit 2.5PH in a 1MW immersion cooled setup at 0.4w/gh.

I'm not sure how 0.6w/gh at the wall transformer for a datacenter is too much power.

I am assuming you mean using SP 28nm chips you could get 2.5PH and you would probably be about right. Question is why are they not taking advantage of this technology?

[RoadStress]

Yes I think you could get a datatank for about that price and that is exactly what their IPO is offering.

Even if you can't buy your own container from allied control for that cheap you can probably make your own for that price or even less. I somewhat doubt that allied controls 3rd try is the most cost/energy efficient possible with 2-phase immersion cooling. For example 3m was able to cool 4KW with only 0.2 liters of novec. If you could match that then you would only need 50 liters or $2,500 worth of fluid per MW.

DYI is the wrong way to do a 1MW immersion cooled setup. Not that easy as it sounds. I would definitely let experts handle this.

I see that the DataTank guys aren't selling the DataTank separately so I don't believe that the 500k is a real price. Maybe I will e-mail Allied Control to ask them about the price, but I'm 99% that their price will be much more (at least 50% more)

I'm not sure how 0.6w/gh at the wall transformer for a datacenter is too much power.

I am assuming you mean using SP 28nm chips you could get 2.5PH and you would probably be about right. Question is why are they not taking advantage of this technology?

Yes I was talking about SP 28nm chips. Their future plans are unknown to me, but I remember that Bicknellski wanted to try that. Will see how that goes.

[LeanSixSigma]

PUE in hong kong is 1.5 average so a 500kw system would use 250kw for cooling. In reality the savings are even more than that because of the performance boost.

Actually, I'm curious about the average PUE of 1.5 in Hong Kong. May I ask where you've got the number from?

There are some very modern DCs with PUE of 1.5-1.6, but it's not the average AFAIK. It's the very best available after Immersion-2:
https://www.datacenterdynamics.com/focus/archive/2013/09/hong-kong-stock-exchange-builds-data-center ("many in HK running at PUE 2–2.5")

The various articles/studies I'm aware of (including above one) are actually pointing towards PUE 2.2 or even higher:
https://www.datacenterdynamics.com/focus/archive/2013/04/apac-data-center-survey-reveals-high-pue-figures-across-region ("average reported PUE is 2.42")

CLP (China Light and Power as Hong Kong's major electricity provider) has released at a data center symposium the results of their own survey in late 2013 with 100+ data centers that the average PUE is 2.2.

Hong Kong is warm even at some nights with more than 30C/86F, paired with 95%+ humidity which makes it very difficult to cool efficiently. So achieving PUE of 1.01 in HKG like for Immersion-2 is extremely low by any standards. This does not add to the discussion on comparison with US data centers, but it actually supports your case in terms of saved electricity costs with Immersion-2 even more.

[LeanSixSigma]

Even if you can't buy your own container from allied control for that cheap you can probably make your own for that price or even less. I somewhat doubt that allied controls 3rd try is the most cost/energy efficient possible with 2-phase immersion cooling. For example 3m was able to cool 4KW with only 0.2 liters of novec. If you could match that then you would only need 50 liters or $2,500 worth of fluid per MW.

There's economies of scale involved. 3M sells the fluid to several industries, including to law enforcement agencies in small quantities to recover fingerprints from extremely sensitive surfaces like facial tissue. 3M's selling price for high quantities is much lower than in small quantities. Same for everything else, including building 2-phase immersion-cooling devices in large scale. Buying a hand full of power supplies is quite different to buying thousands directly from the manufacturer for example. Metal fabricators easily charge twice or triple for low quantities, because a lot of labor is involved when setting up and programming machines or even custom create new tools for fabrication of special metal shapes.

Further, the 3M study was using copper boilers in a simulator. The capability to achieve densities of 4kW in 0.2 liters is inherently already possible right away with current setups. But there is still no hardware which has such power densities (hence only the simulation). It's like squeezing 16 of the latest 250W GPUs onto a PCB the size of two post cards. That's one of the reasons why there is the confidence that this technology can be still used for so many more hardware generations to come, while many other technologies are almost maxed out already.

[Anthony1985]

Guys has been a very long time I don't get anymore shares from AM100 is that normal?

[freedomno1]

Guys has been a very long time I don't get anymore shares from AM100 is that normal?

There hasn't been a dividend in months and BTCT closed a while ago so their is no drip either
At least I don't think anywhere AM is traded had drip's set up.