I disagree and for good reason. While you are correct that a lot of the low-end sound cards are junk, the upper end is also the only way you can get any kind of 3D position processing to work well. It is possible that your use of the word "most" is correct, but only because I do not know precisely how many junk cards are made. Also, the sound buffer sync issue is a Windows issue. I know for a fact it is not an issue with Mac, and probably not an issue for Linux. If you overclock you would probably be better off learning how to modify the buffer for better performance. The only real way to avoid the buffer problem is to not use Windows. Since that is not an option the best thing a gamer can do is spend a little money on a sound card that offloads the sample conversions from the cpu to the sound card or sister-audio processor. I am not aware of an onboard sound chip that does that. You can also completely avoid the buffer issue even with Windows by moving to an ASIO interface, but then there is no 3D position processing at all (except of course the standard DirectX CPU bound processing). The best solution I have found is using a top-end Sound Blaster card that offers 3D position processing on card (the latest also offer separate 128 voice positioning) for sound output, and an ASIO interface for vox input. Doing that removes the vast majority of latency issues.
I tested the P-47D25 engine sound using a Realtek ALC1185 without an issue. It runs at 500MHz, which is a slight improvement over the ALC892. The reason I suspected a complexity issue is because ammo reports a 10 second interval, but the sound is only eight seconds in length. I really want to explore the track interval compressing further, but if it leaves onboard users out then I have to let it go.