http://bugs.winehq.org/show_bug.cgi?id=29359
--- Comment #15 from Andrew Eikum aeikum@codeweavers.com 2012-02-13 15:43:33 CST --- This is a problem of our buffer size calculations in dsound, like described in Bug 14717, comment 260. PES uses 8 channel audio, causing large frame sizes (nBlockAlign = 16). In DSOUND_fraglen, we calculate the fragment length to be 512 * 16 = 8192. We prebuffer 10 fragments (= 81920 bytes) into mmdevapi, but the buffer size is defined to be 65536. So, we don't have enough buffer space to even hold our 10 prebuffered fragments.
Since we're in code freeze, I'm probably not going to submit a patch for this (fixing dsound isn't trivial). You can work around the problem by opening regedit and setting:
[HKCU\Software\Wine\DirectSound] HelBuflen=131072 (Decimal)