We measured the level of noise using an electronic noise-level meter CENTER-321 in a closed and quiet room about 20 sq. meters large. The noise measurements were taken outside the system case, when the only noise source was the cooling system and its fans. The noise-level meter was set on a tripod at a distance of 15 centimeters from the graphics card cooler fan rotor. The mainboard with the graphics card was placed at the edge of a desk on a foam-rubber tray. The bottom limit of our noise-level metering device is 29.8 dBA whereas the subjectively comfortable (not low, but comfortable) level of noise when measured from that distance is about 36 dBA. The speed of the graphics cards’ fans was changed with the help of a special controller supporting 0.5 V voltage adjustment increments.
The vertical lines on the graph indicate the fan rotation speed range for automatic mode during our temperature tests. Let’s see which graphics card of the three turned out the quietest:
The graphics cards from EVGA, Gigabyte and MSI are actually very similar in terms of noisiness, their graphs being close to each other in the diagram. The fans of the EVGA and MSI cards even have the same speed range, although the MSI is somewhat quieter. However, the numbers and graphs do not agree with our subjective impression in this case. For example, the Gigabyte seems to be the quietest of all while the MSI is the noisiest to us (in the automatic fan regulation mode) – just the opposite of what we see in the diagram. Well, we must confess that none of these three cards will please users who want a silent computer. Each of them becomes audible against the background noise of a quiet computer system as soon as you launch a 3D application. Every card is quiet in 2D applications, though.
Now we can build a diagram to compare the noise and temperature of the cards in three modes: 1) automatic fan regulation mode and default clock rates, 2) maximum speed of the fans, and 3) automatic regulation + overclocked frequencies. The graphics cards are sorted in noise ascending order.
As you can see, the difference in noise isn’t large in the automatic fan regulation mode, as opposed to the difference in GPU temperature. Taking the temperature into account, it’s clear that the GeForce GTX 660s from MSI and Gigabyte are preferable to the EVGA.
We measured the power consumption of our testbed equipped with different graphics cards using a multifunctional Zalman ZM-MFC3 panel which can report how much power a computer (without the monitor) draws from a wall outlet. There were two test modes: 2D (editing documents in Microsoft Word or web surfing) and 3D (three runs of the Metro 2033: The Last Refuge benchmark at 2560x1440 with maximum image quality settings, but without antialiasing).
For comparison purposes we also added the results for ASUS GeForce GTX 660 Ti DirectCU II TOP, ASUS GeForce GTX 680 DirectCU II TOP, Sapphire Radeon HD 7870 GHz Edition OC and Sapphire Radeon HD 7970 OC Dual-X. Let’ see what we got:
The configurations with four different GeForce GTX 660s and one GTX 660 Ti don’t differ much in their power consumption, yet the one with the reference Nvidia GeForce GTX 660 is the most economical since that card is not pre-overclocked. Next goes the Sapphire Radeon HD 7870 whose peak power consumption is 402 watts. The configuration with the top-end GeForce GTX 680 from ASUS needs 8 watts more. The two GeForce GTX 660s (at 1033/6008 MHz) in 2-way SLI mode require almost 500 watts of power, i.e. 110 to 115 watts more than the configuration with only one such card. So, Nvidia seems to have played it too safe by specifying the power draw of one GeForce GTX 660 at 140 watts, but we should take the efficiency of SLI technology into account (90% in Metro 2033) together with the fact that the test load wasn’t 100% high.
It is the Sapphire Radeon HD 7970 at 1050/6000 MHz that needs the most power in this test. Its configuration requires as much as 522 watts. On the other hand, we can note that any of these configurations can be powered by a 550-watt PSU despite the overclocked six-core CPU.