Topic: CoinTerra announces its first ASIC - Hash-Rate greater than 500 GH/s - page 90. (Read 231002 times)

So is this moving slowly to a fail?

Just came across CoinTerra's Google's Adwords ads.  "Easy Money"  I agree it is easy money for CoinTerra, not so much for their customers unless they meant easy come, easy go

i want to believe its worth it - but even 6 grand seems a waste now.

Id be fuming if id spent 15.

They haven't said a single word.   Last statement was "tape scheduled for early Oct" and then silence.

did Cointerra tape-out in the meantime?

Investing in a good 220V PDU and a 220V 30A circuit is a good idea if you're planning to run lots of equipment.  I'm really glad I did this.

Chip wattage =/= wattage at the wall.  Figure 90% lost in DC conversion, and another 90% lost ATX PSU and add in overhead for cooling and host.

If the chip is 0.55 J/GH no way you are looking at the wall wattage of only 1300W.


Agreed.  I doubt we will see higher than 2 TH/s and 1.0 or 1.5 TH/s may make more sense in the long run.   Continual loads need to be derated 20% (NEC) so that limits a 120V, 15A circuit to 1440W continual.  You can always use more units if you want more power.

I wish there was some news.  It's hard to feel good about silence.

agreed, we dont know the exact power til theyve taped out, though since everyone else is also in the same range, i think for the next few days we can assume its 0.6 ISH, since thats where most of the 28nm chips seem to be drawing... (hashfast, bfl, etc) and then when they announce the tapeout final figures, we will know exactly.  thats why i suggested 1200-1300 watts, allowing for 0.65 w/gh (same as hf).

no one else in the bitcoin asic world has got it so wrong as bfl did last time, so that was a monumental error on their part and you cant tar everyone else with the same brush.  there's no way that anyone else is going to make a mistake and get the power wrong by 5X!

but it does seem that 2 TH is about the max that you can power off of one household circuit in the 28nm generation (from any asic supplier), and even that is borderline and better to do it off two circuits.

i think its clear that the long-term future of bitcoin mining is in datacenters... even for individual ownership, or as part of hashing contracts... it makes more sense for these power hungry boxes to be hosted somewhere professional with good cooling and reliable power.

-- Jez

I care because I don't have 2 120V/15A circuits in close proximity to one another. I don't have 240V. I have no intention to do any 'rewiring' in my house. My datacenter charges me for each 120V/15A circuit I use. In other words, the device would be much less hassle to deploy (for many people, I imagine), if it could be run from a single and very common 120V/15A circuit.

Your calculations are incomplete until you know the power requirements of the ASICs. This is the big unknown. BFL was off by a factor of 5 (what they *thought* would be 60W turned out to be almost 300W). I don't expect Cointerra to make a mistake of this magnitude, but until they have actual silicon the reality is that we do not, and cannot, know what the power will be.

why do you care whether its 1440 watts or more?   there are two power cables, so you plug each one into a different circuit if youre using 120 volts), or could be the same circuit if youre on 240 volts.

lets say 2000 TH needs 1200-1300 watts (0.6 ish W/GH).  and the fans (there's 6 of them in the photo, 3 at the front, 3 at the back)... and the pumps (there's 4 of them, one for each chip) are probably 100 watts... and the controller board... (im guessing, 20?)...   so you're probably on the max end of your 1440 watts, without allowing for power supply inefficiencies.. and dc/dc power contverter efficiencies, so i think its going to be, realistically, above 1440 watts, even for nominal use.   And then wouldnt you want to overclock/overvolt them to make them run faster, at least initially, when its makes more sense to get more GH's out of it?

like everyone else in the 28nm world, im sure they will also offer a way to undervolt to save power, at the expense of GH's

-- Jez

I'm hoping the the 2TH Cointerra does not require more than 1440W. Right now that's still up the air.

In Europe we have electric heaters that run at 13 amps (3,000 watts).  some of our kettles, and sandwich makers run at 3,000 watts... so i think thats the usual max for a household circuit (13 amps at 240 volts)

and of course, our ovens have a 30AMP circuit.

-- Jez

Fat wires are better. 220 to 240 volts is better. I live in a country where everything is almost 220 to 240 volts, and every wire I've seen is noticeably bigger, even if they actually need a thinner wire.

Only appliances imported from 110 volt countries are 110 volts, and they tend to fry when plugged in a 220 volt outlet. hehehehe.

Thanks Epoch, I just sent you a PM with one last question because I didn't mean to hijack this thread. 

Yes, as strange as it may seem, you can easily transmit 288000W (15A, 24kV x 80% derating) through 14 gauge wire (though at that voltage you'd need to separate the conductors further to prevent arcing. That is a separate issue but let's ignore that).

The reason for this is that the power is not dissipated in the wire; it is dissipated in the load. If you can grasp that, you are well on your way to understanding what is happening here.

Your formula W = V x A is correct. The total power dissipated in the entire circuit is 288000W, that is true. But the key thing here is that 287999W of that is dissipated by the load (random device, or space heater, or ASIC mining farm, etc.) while the remaining 1W is dissipated in the wire.

The wire itself has very low resistance so the voltage drop across any part of it is very low. So applying the formula: V (very low) x 15A = a very low wattage. The wire hardly heats up at all. The bulk of the 24kV voltage drop happens across the load, and the power there is V (large) x 15A = a very large wattage. It is the load that has to withstand that huge amount of power, not the wire.

This is how a small wire can transmit very high powers. In fact, in the power transmission industry people try to crank up the voltage as high as possible in order to transmit more power along a fixed-gauge wire. That is why high power AC transmission lines are measured in kilovolts, or even hundreds of kilovolts. Transmission lines don't use 120V or even 240V because they would never be able to transmit enough power to be useful.

I probably won't do it justice but the lines are out of phase.  AC power is a waveform.   The phases don't peak at the same time they are offset by 30 deg.   The 1.73 comes from 2 cos(30 deg).

This diagram might help.



If you look at Phase-B it peaks at right below the letter "e" in Phase-B.  Now draw a vertical line down to the Phase-A.  You notice at the same point where Phase B has a positive peak, Phase A isn't at the negative peak, it is slightly less.   How much less well about 0.73x the peak.  If the distance from peak to centerline is 1 the distance from one phase to another when one phase is peaking is 1.73.  Using voltage, image the x-axis as ground (0 V) and each phase peaks at 277V.  So you can mark the top and bottom of the graph as 277V. If the peak voltage is 277V then the voltage between any two phases is 1.73*277V=480V.  A to B = 480V, B to C = 480V A to C = 480V.  A to GND = 277V. B to GND = 277V.  C to GND = 277V.

As for "why"?  It is more efficient in terms of power vs size and number of conductors.

BTW I believe it is 480V not 460V.  Each of the phases has a line to ground voltage of 277V.  They are tapped line to line producing 1.73*277V=480V between conductors.

I understand that, thank you.. I know you are correct, but it is still counterintuitive to me as to how you can send the same current through the wire yet somehow get double the power.  
As a thought experiment I tried exaggerating the scenerio. what happens if you use 24,000V?  Does that mean you can use 288000W through a 14 gage wire, as long as you stay at 15A?  Obviously not, but I cannot see why the situation would be different.

... sorry for sidetracking I just want to get a better understanding of this.  I've always known the formula W = V x A but I've never been able to fully grasp it in real life operation

can you explain why when you get into 460 three phase that 1.73 number comes in all of a sudden?   I cannot for the life of me understand that.

Simple answer is yes.   Wire guage is based on current.  By switching the circuit to 240V you effectively double the wattage of the circuit.   Remember in many cases there are multiple outlets on the same branch circuit so you want to make sure you change all the outlets (or at least mark them it wouldn't be code but it would keep you from doing something stupid).

You will need to rewire the branch at the breaker panel.  In the US a 120V circuit is wired neutral, hot, and ground.  240V is two hots and ground (no neutral).  

So
120V 15A circuit = 1800W * 80% (derate for continual load) = 1440W usable.
240V 15A circuit = 3600W * 80% (derate for continual load) = 2880W usable.
Same wire double the power.