We measure the power consumption during two scenarios: idle (where we determine an average based on 5 minutes of idling) and under load, where we measure the maximum power consumption during Cinebench 15.
We test the sound quality of the boards with the loopback test of Rightmark Audio Analyzer. Do note: this means we test the quality of both the input and output of the on-board sound cards simultaneously. The 'weakest link' determines the result.
The dynamic range and noise level test are very comparable and show nearly identical results. The dynamic range test measures the difference in volume between the loudest and softest observable signal, the noise level test measures the difference between the loudest sound and the noise level.
To put the numbers in perspective: the highest dynamic range c.q. noise level theoretically possible with 16 bit sound is 96 dB. 99.9% of the sound you listen to on your PC (CDs, MP3s, YouTube, games, ...) is 16 bit and the graphs proof that the hardware is no bottleneck in almost all cases. We can only achieve the higher scores by testing with 24 bit audio. By the way, we cannot imagine someone actually able to distinguish between 91 dB and 100 dB signal/noise ratio.
There are only a few boards that truly offer audio quality that is below average. With the MSI B250M Mortar and the two Gigabyte B250-boards it was clear that this was not a priority in the design process. The ASUS Strix-boards and the ASRock Fatal1ty H270 Performance perform very well in these tests.
The stereo crosstalk tests to which extent sound from one channel (left or right) resounds on the other channel.
Lastly we measure the total harmonic distortion, the average distortion in the frequency domain, using RMAA.