To test power consumption we connected a Kill-A-Watt EZ and recorded the maximum and minimum numbers in watts for the entire system. Idle was recorded after the system was sitting on the desktop screen for 15 minutes and full load was recorded first during multiple loops of Aliens vs. Predator benchmark at 2560 x 1600 resolution with everything maxed out and then after 15 minutes of Furmark 1.9.1
The idle power has been pretty consistent among the cards run on this bench with a gradual increase as the amount of memory and active processing units increased on the GPU. About what you would expect that as you increase the GPU horsepower the idle load goes up a little bit.
On the full load side the HD 6970 showed that it was a bit better than the NVIDIA on power only hitting 276 watts during the AvP benchmark loop. This is impressive as it is competitive with the GTX 560 series for power consumption. In contrast the EVGA GeForce GTX 570 HD Classified hit 291 watts during the gaming load cycle proving that the delta in temperatures between these two cards was related to the amount of power hunger they had.
Under the synthetic Furmark 1.9.1 test things really solidified with the HD 6970 jumping up to 299 watts and the EVGA GeForce GTX 570 HD Classified running all the way up to 387 watts.
It was good to see that the 500 series was doing a much better job of power management and kept very close to the HD 6970 in power consumption under full gaming load but if you push the cards unrealistically hard the FERMI’s power hunger shows up. It is a good thing you would never run the cards that way because that extra 88 watts would add up to a nice charge on the utilities at the end of the month.