Quote from: Dexter770221 on June 17, 2013, 07:40:04 PM

Look at the whole perspective.
Those chips are designed to be powered in serial. So DC/DC converters can be simpler and cheaper (less current). And this leads to smaller PCB, smaller heatsinks, smaller cases, less noise from fans, higher density GH/cm2.... Ideally for datacenters where every watt counts. So this ratio 6/2 makes ALOT of sense...

In data center kW don't count that much, they have plenty of them, if you shop around. I got quotes for $0.12-$0.25kW including internet and rack space from multiple data centers for 10-20kW per rack. Full rack with 10kW power, 10Mbps internet from $864-$1,800

Ok, the cost of less powerful supporting electronics vs more of them may or may not balance to zero sum, but you still get virtually no benefit and only detriment in ruining chips at lower voltage, i.e. use more hashing chips to get same performance and chips are not free.

Would you rather make $86/day and pay $30/day for power = profit of $56 or
make $43/day and pay $5 for power = profit of $38

Quote from: kaerf on June 17, 2013, 08:08:38 PM

In about 18-24 months, the power savings could still allow the chip to be profitable when diff is closing in on 10 billion Grin

This chip will not produce any meaningful hashing beyond 45 weeks at 2% hash rate increase a day that is coming