Quote from: DeathAndTaxes on August 29, 2013, 11:58:40 PM

Quote from: Dyaheon on August 29, 2013, 11:49:31 PM

Good list.

Bitfury is somewhat better than that though, I measured +116W for adding a 150GH/s Bitfury unit to my rig. Gold-rated 1300W PSU at 440W@230V total now so I'd assume about 90% efficiency. Anyway, <0.8J/GH at the wall for Bitfury. Another guy had similar results.

Was this on the reduced speed boards? The reason I ask is if the boards intended for 25 GH/s are getting 21 GH/s then they are essentially underclocked (even if unwillingly) and that will improve efficiency.
Was the 116W increase at the wall?

Thanks for the thread, D&T.

BTW, as I'm sure you know, if voltage of the chips is constant, efficiency is roughly independent of frequency(hashrate). While there is a very slight increase of efficiency at lower hashrate, due to better efficiency of DC/DC and main PS, we are talking of single digit percentages only.

Indeed if the voltage can be adjusted, and if you do this properly, efficiency is roughly inverse dependent with the square of frequency(hashrate). For a 10% increase in hashrate you pay a ~33% penalty in power consumption.

On 4 modules Avalon, where voltage is fixed, I have ~8.4J/GH@256MHz degrading continuously to ~8.6J/GH@360MHz. I assume efficiency is a little higher at lower frequencies due to better efficiency at lower power of the DC-DC converters and main PS.

On the other hand, on a 2x7970 rig, at the wall, including motherboard and HDD power consumption:
465 J/GH (2.15 MH/J) @ nominal voltages (1.175 and 1.112V) = 1300MH/s, 605W
322 J/GH (3.10 MH/J) @ undervolt 0.95V = 1100Mh/s, 355W