Originally Posted by ROBSCIX
True both cards are very similar for certain measurments you cannot just compare the head outs like that...also to point out a difference of say 113dB to 117dB for SNR..etc is quite alot as the scale is LOG, not linear.
So when some see a measurment that looks similar in all actuality it is not close at all.
-119 dB noise may be only half as much noise voltage as -113 dB, but if -113 dB is already inaudible, then the improvement is of little to no real practical value, but it looks good for marketing. Under normal listening conditions, people cannot even hear the noise floor of the 44.1 kHz/16-bit CD format, which is worse than -100 dB A-weighted. It should also be noted that when multiple uncorrelated noise sources are added, then it is the power that needs to be summed. So, for example, if there is 95 dB amplifier + ambient + source SNR (a figure that is even optimistic, a typical recording on a CD already has more noise than that), going from a DAC with -110 dB noise to another one with -120 dB only improves the overall SNR from -94.86 dB to -94.99 dB.
Very low DAC noise can, however, be useful for digital volume control. That is in fact used on many sound cards to reduce costs. But with an external amplifier that controls the volume after the DAC (which can therefore be at or near 100% digital volume all the time), the dynamic range of the DAC does not need to be that high to be good enough.
Of course, what is advertised by the manufacturer, might not always be achieved in reality. But it is not guaranteed that better advertised specs will translate to better real world performance, either. For example, the headphone output of the Xonar Essence STX is specified to have an SNR of 117 dB. That looks great on paper, and one would think that no one can hear that. However, the specs do not mention that the DAC is more noisy at 44.1 kHz (the sample rate used by well over 90% of available music), dropping the value to about 110-111 dB. Also, the card has fully digital volume and gain control, so it always outputs the maximum amount of noise, regardless of the volume or gain. Reducing the signal level from the maximum of 7 Vrms to something that is listenable with typical headphones will reduce the dynamic range accordingly, too. At 0.4 Vrms (which is plenty loud enough with efficient modern headphones), and 44.1 kHz sample rate, the headphone output has a less impressive dynamic range of 86 dB, or 14.3 bits. If the signal level is reduced further to 0.1 Vrms - for example for a sensitive IEM - then only 74 dB is left, and the hiss may actually become audible. The line output of a cheaper card combined with a low noise external amplifier, like the O2, will easily outperform those numbers.