Blind listening tests
In my view, the best way to evaluate a sound card is to listen to, well, how it sounds. Plenty of folks insist they can hear a difference between integrated and discrete solutions, so we lined up the Xonars against our motherboard's Realtek codec for a series of listening tests. These were blind tests, so our subjects had no knowledge of which audio config they were hearing at any given time.
After several years of sound-card testing, I've found that doing back-to-back listening tests with relatively short 30-second song clips is the best way to get listeners to pick out specific differences in playback quality. The listeners wore our Sennheiser HD 555 headphones for these tests, and each card's output level was normalized to within 0.1 decibels using RightMark Audio Analyzer to measure the volume of a test tone. The ALC892, DG, and Xense all faced each other in a series of head-to-head matchups with each song clip, and the order of those matchups was randomized for each track and test subject.
Our music snippets were all ripped directly from the original audio CDs and saved as uncompressed WAV files. We played these tracks back using Windows Media Player 11, the default media player for Windows 7. Below, you'll find our listeners' impressions of how each card sounded versus the competition. However, before we get into the results, it's worth taking a moment to introduce our listeners. I asked each to rate themselves on an impromptu audiophile scale between 0, which considers Apple's stock iPod earbuds to be awesome, and 10, which classifies as garbage anything that isn't piped through multi-thousand-dollar speakers via gold-plated Monster cables from an original vinyl source.
The first listener to run the gauntlet was Matt Trinca, a new writer of ours who lives conveniently close to the Benchmarking Sweatshop. Until recently, Matt relied on a mini-stereo system for audio playback. He only rates himself a 6 on our scale. Cyril also lives within range of my lab, and he was kind enough to not only sit through the listening tests but also run them so that I could become a blind subject. A bit of a connoisseur, Cyril uses Sennheiser HD595 headphones and puts himself at an 8 on our audiophile scale. I'd give myself an 8, as well, and I spend most of my time listening to music piped through a pair of Abit iDome speakers. Finally, I bribed my girlfriend, Mo, to sit through a round of tests with the promise of a backrub and an empty dishwasher. She had truly atrocious speakers before I migrated her to a pair of cheap Logitech units that she thinks sound phenomenal, so I'm going to give her a 4 on our audiophile scale.
Arctic Monkeys Fluorescent Adolescent
A reasonably straightforward rock track, Fluorescent Adolescent combines just enough British attitude with the usual mix of radio-friendly guitars and drumming.
Matt had a hard time discerning the difference between configs with this track. He found the Realtek solution's drumming to be more forceful than that of the Xonars, and thought the DG sounded a little muddier than the Xense. Mo said they all sounded roughly the same, but she preferred the DG for reasons she couldn't put her finger on.
Cyril and I both detected a low hiss in the background with the Realtek codec, which I suppose made the rest of our testing somewhat less than perfectly blind, at least for us. He thought the integrated solution sounded a little artificial and preferred the DG for its crisper bass and superior separation, even versus the Xense. I singled out the Xense for better separation, but I still favored the sharpness of the DG's percussion and vocals. To my ears, the background instrumentation sounded a little lost on the Realtek codec.
Dan Le Sac vs Scroobius Pip Thou Shalt Always Kill
Think a taller version of Eminem, add a beard, more intelligent lyrics, better beats, and, well, forget Eminem altogether. Dan Le Sac is about as urban as this middle-class white boy gets.
Scroobius Pip's sarcastic vocals came across clearer for Matt on the Realtek codec, and he found the drumming on the Xonars a little dull in comparison. Mo didn't agree, and thought the ALC892 had an air of static. She liked the DG overall, and preferred its vocals to those of the Xensean opinion shared by me and Cyril. Cyril found the Xense a little timid overall, but we both agreed it and the DG sounded much better than the integrated audio, which sort of mashed things together under a heavy bass line. I quite liked the sound of the Xense overall, but the DG tickled my weakness for a bit of vocal bias.
Nine Inch Nails Ghosts 14
My favorite track from Nine Inch Nails' free Ghosts instrumental collection, number 14 is heavy on strings, with a healthy dose of industrial flair.
You learn interesting things while conducting this sort of testing, one of which is that different people listen for different things. Matt, for example, consistently commented that drumming hit harder on the ALC892 than on the rest. Percussion with a little more punch is also the reason he preferred the DG to the Xense with this track. He liked the DG best overall, noting that its background instrumentals were a lot clearer than with the hard-hitting crab.
Indeed, we all preferred the DG with this Ghosts song, and for much of the same reasons, even if we used different words to describe a general sentiment. Mo thought the DG had more range than the Xense, which Cyril said sounded a little muted. I reached the same conclusion from the opposite direction, praising the DG's brightness versus the Xense, which had more of an even keel.
Radiohead Weird Fishes_Arpeggit
Another online release, Radiohead's critically acclaimed In Rainbows spawned this oddly named vehicle for Thom Yorke's beautifully delicate voice.
Score three for the Xense. Matt found it more vibrant than the DG, which was clearer than the ALC892. Cyril praised the Xense's natural sound versus the other solutions, whose vocals he found a tad sharp. I detected the same vocal bias, and noted that the DG at least preserved the rest of the spectrum, while the onboard codec made the background instrumentals sound a little compressed. But the Xense was my favorite, as well, thanks to a more even separation between the various elements of the track.
Although she didn't agree with the rest of us, Mo was at least consistent in picking the DG as her favorite. She did think that all the configs sounded very similar, but also that the Xense was a little tinny when compared to the budget Xonar. The DG, she said, sounded just a bit better than the ALC892.
Tori Amos Cornflake girl
A little something for the ladies... or I suppose from the lady, Cornflake Girl layers strings, piano, and Amos' intoxicatingly breathy delivery.
With our final track, the results were nearly split down the middle. Matt and Mo thought the DG sounded better than the others. He detected more separation with the DG than with the Xense and thought the former was clearer than the ALC892, whose drums were too high in the mix. Mo called the differences more subtle, but noted that the Xense's vocals sounded weak versus the DG. I had similar sentiments, characterizing the DG's vocals as punchier than those of the pricier Xonar. The DG's piano stood out a little more for me, too, although the Xense had a more even balance with better separation. Only because I like a little extra kick did I prefer the sound of the DG to the Xense.
Cyril isn't as easily swayed by a little extra oomph, and he sided with the Xense's superior separation, even if it sounded just a little bit quieter. The ALC892 didn't fare as well to his ears, which heard something more metallic and artificial than what was produced by the Xonars. I complained about a blurring in the background with the Realtek codec, and Matt thought it lacked a crispness present in the discrete cards.
Some time has passed since these listening tests were conducted, and I've since been able to gather a some additional context that I think will help to explain the results. More often than not, the Xense had the best separation and most even sonic profile of the bunch. It tended to give each element of the track equal attention, and in back-to-back comparisons with implementations that have specific biases, I can see why we often thought the Xense sounded less exciting than its competition. As it turns out, Asus programmed a little extra excitement into the DG. The card isn't tuned to sound exactly correct, as is the case with the Xense, but to give things like percussion and vocals a little extra pop. And our listeners liked pop. I'm a total sucker for pop, too, especially since it often amplifies my favorite elements of a given song.
The ALC892 has some extra kick of its own, but not like the DG. Here, the question isn't whether there's bias, but what, if anything, that bias costs you. On the DG, I'm left with the impression that that Asus has nudged up the volume on a couple of instruments without messing with the rest of the band. With the Realtek codec, it sounds like the drums and vocals have been turned way up, robbing focus and bandwidth from the rest of the spectrum. Both solutions are guilty of massaging the sound a little before passing it to your ears, but the DG's touch doesn't disturb the surroundings, while the onboard audio can trample on background instrumentals and other subtleties.
|AMD reveals the full specs of the Radeon RX 460 and RX 470||4|
|Nvidia will pay GeForce GTX 970 owners $30 over memory snafu||22|
|Gigabyte's GeForce GTX 1080 Xtreme Gaming graphics card reviewed||17|
|Microsoft's free Windows 10 upgrade offer ends tomorrow||72|
|ASRock H110M-STX mobo puts the 5x5 platform in builders' hands||15|
|Asus' slim ROG G20CB desktop gets in on the Pascal party||7|
|Apple sells its billionth iPhone||36|
|TT Premium Edition RGB LED radiator fans play better together||7|
|Toshiba's latest BiCS flash is stacked 64 layers high||11|
|Now you can install Crysis directly on the video card!||+63|