Can't answer the original question (although my opinion from reviews i have read and personal experience with on-board sound and PCI cards is DON'T WASTE YOUR MONEY), but yes, sound cards can affect FPS in some games.
The reason is that if the CPU is having to spend time calculating sound as well as the frames, then it clearly can't spend all it's time calculating frames. Or by the same principle, if you ask the CPU to do heavy AI work, again, there will be some reduction in frame processing. If you can off-load those calculations strictly to the soundcard, bingo, higher FPS. But only a few (the graphics card clearly has a strong involvement here - don't ask me exactly what the relationship is).
That makes sense, but I don't think changing audio cards will affect FPS at all. Sure, the FPS may be lower on a computer without any sound card, but I doubt the FPS differs much on computers that have a sound card, no matter the brand or type.
It can make a larger difference if you have heavy 3D sound in useage with lots of objects.
For instance, for a onboard solution to emulate EAX2, let alone EAX3 would put quite a bit of additional pressure on a CPU, depending of course on the load already on the CPU from other components.
Lots of times though, you won't see significant FPS loss because those cheaper solutions (onboard), its simply not feasable through the CPU and cannot emulate or don't have a license to attempt to recreate certain 3D processes such as EAX 3 (Advanced HD). If they did your CPU would most likely choke and die, so you don't see to much advanced 3D sound usuage onboard.
Getting to the point, you still get sound in every game with cheaper solutions and not loss massive FPS because you are losing sound quality, effects and richness to the environment.
Well it's hard to argue with the article (thanks for posting the link by the way).
And even the theory behind seems pretty watertight to me.
A CPU (speed irrelevant) can only run at 100% of it's maximum.
We can say that this is allocated to making a game run. Let's say (very simplified) that 50% is dedicated to computing the AI and physics etc, and 50% to feeding the graphics card (which in turn will produce what gets seen).
If you then ask it to also compute sound, with for example an on-board sound chip, let's say 20%, you then have to reduce the amount available for the rest (so you now have 40% for AI etc, 40% for graphics).
If you can off-load some of this to a sound-card with it's own processing, the CPU will still have to do SOME work (let's call it 10% now), but leaving 45% each for AI and graphics.
So you can clearly see how turning off sound altogether gives the best possible frame rates (although would detract from the overall experience in my opinion), and having a soundcard is the next best step.
I found one major distinction between the Audigy 1 and the Audigy 2 that applies if you use a stereo setup with your computer rather than typical computer speakers. The Audigy 1 requires the front and the rear speakers to plug into separate jacks, but the Audigy 2 is able to run both signals through a single jack.
I use a receiver and four speakers with my computer, and I have been deeply disapointed with the Audigy 1 because of this.
One thing you are failing to remember about the new NForce2 motherboards is the inclusion or exclusion of the Nvidia APU (Audio Processing Unit). That APU is just like a soundcards processor. It does calculations so the CPU doesn't have to. The big problem with the Nvidia chip is the lack of soundstorm cert on all mobo's with the APU. The ones with just the APU and no soundstorm cert lack a couple of key features like Dolby Digital 5.1 encoding and no SP/DIF. They all do EAX2.0 and 64 3d voices. So the question is it worth it to go from Audigy 1 -> Audigy 2. IMHO Yes, but you should look at other options as well because there are some just as good that will be less expensive.