To readers: disabling SpeedStep
rectified the OP's issue.
The problem in question has to do with how PC hardware actually works at a very low level, and it's virtually impossible for any end-user to diagnose this reliably (in turn they choose to start spouting off nonsense on forums, saying things like "I removed my Big Dog Snake USB/CF card and now everything is great!" -- lies).
Event timers are driven by a series of dedicated chips which have different degrees of accuracy and latency (ex. HPET vs. ACPI timers vs. i8254). All of these are affected in some way or another by certain universal clock rates
SpeedStep adjusts the CPU clock rate, which has a direct effect on event timers as well as a multitude of other things -- many things that require/rely on a stable/set/constant rate of occurrence (sound/audio is the biggest one) get affected by this. Some drivers handle this better than others, and some hardware handles this better than others.
Likewise, other CPU features like C3/C6 P-states can cause fluctuations in these rates, resulting in the above.
The more "power-saving" features get added to hardware (regardless of where -- at the CPU level, at the bus/protocol level, or anywhere else), the more fluctuation there tends to be when it comes to event timers as well as interrupt rates. The applications (that also means drivers) have to be written to handle these situations cleanly/correctly -- many don't.
For example, here's a thread
where I got into a lengthy discussion with a NES emulator author about his choices to force-disable some of these power-saving features citing "audio latency issues". He provided responsiveness graphs which I also questioned. After pushing him for a while, he dug through his code and found that the root cause was actually a design flaw and upon fixing it there was no need to force-disable the power-saving features. I rest my case.
Realtek in general has a very well and long-established history of making (generally speaking) crap. Their audio ICs and NICs tend to be sub-par, ditto with their drivers. I've ranted about Realtek NICs on my blog in the past (use Google to find it). The reason Realtek is everywhere is because their ICs are cheap, thus allowing the board manufacturers to make a larger profit.
For example on all Realtek audio drivers that I've used (ALC887 and ALC889), there is a downright 100% repeatable bug pertaining to frequency playback rates when using non-DirectX for audio that is 22kHz or lower -- the playback frequency table is wrong (audio plays back at a slightly lower frequency than it should).
I also tend to avoid VIA anything (CPUs, ICs, whatever -- doesn't matter what) given their years and years of horrible, horrible chip bugs and driver bugs. Anyone who remembers the VIA 4-in-1 drivers and the 686B will know what I'm talking about.
Sadly I've tried add-on audio cards that have their own series of horrible bugs too -- I've found just as awful bugs in Asus' Xonar DG drivers, Creative Labs X-Fi drivers, and in M-Audio's Revolution drivers. I also tried some other brand (I forget the company name, I still have the card in some box) and found even different bugs.
That's really all I have to say on the matter.--
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer.