As someone who actually does a lot of DSP work I have to disagree. Hardware channel mixing, audio latency, resampling, and so on are all often handled very poorly by multi-purpose processors. 3D and HRTF's although 'cool sounding' are not the most resource intensive functions a card, or audio application will perform. 'Perfect' polyphase or bandlimited interpolation techniques will consume far more resources than they will. Most any processor today, including current i7's, xeons, etc can be bogged down by resampling alone. Let alone noise shaping, dithering, hard limited, and so on. We're a long way off from no longer needing sound cards. The only reason they wouldn't matter is if audio quality isn't something that matters. Which is sometimes the case. Also onboard sound cards are generally atrocious. I'd put them about on par with your average onboard graphics adapter in quality. No sample rate switching, no hardware mixing, no hardware dac, high latency. I realize this may seem pedantic, but really it's not. The onboards aren't crappy compared to some studio level expectation. They're crappy by most standards. In many cases when you're listening to lossy audio, sent through a crappy resampler, with a ton of gain added through some relatively low end speakers of course you're not going to notice. When you start talking about high end headphones, or monitors however, it often will be noticeable in more ways than one. Your audio setup is only as strong as its weakest link, and if you're using an onboard... it's not your monitors/headphones.