Here and here are some audio files recorded from the output of various low end (including some very old) PC sound cards, while driving 250 Ohm headphones near the maximum volume; in all cases, the actual voltage on the headphones would have been in the range 0.93 to 0.99 Vrms with a 1 kHz 0 dBFS sine wave. The recording and editing format was 96 kHz/24-bit, but this was converted to 48 kHz/16-bit at the end to reduce the file sizes. The files are level matched, channel imbalance is eliminated, too (although it was always well under 1 dB to begin with), and sample rate conversion was applied where necessary to correct minor differences in the actual sample rates of the cards tested. One sample was processed with a FIR filter to equalize the effects of a low damping factor on the frequency response. The original files (borrowed from these threads) are included as well, with the sample rate converted to 48 kHz for consistency. I do not tell what the files are yet to minimize bias, but a text file containing the solution has an sha1 sum of 9c614de5110f151995380d62626aa08db0f3c67e.
These are probably not difficult to tell apart, and some are even very easy. Post any interesting ABX results, or your opinion on what sounds best and/or worst.