Originally Posted by CingularDuality

Couldn't you say that about any sound card?

Absolutely - Just about every sound card available these days will have a DSP on it... the whole 'extra loading for the CPU' thing really isn't an issue these days provided if you're using an external sound card. Back when the earth was cooling, sound cards would only have the ADC's and DACs (digital<-->analog converters) on them.... Today, you generally have that situation on your motherboard. So, if you are using "on board sound" then you're using the CPU to render sound effects and the motherboard for conversion. Likewise, having Software [x] checked will use the CPU for rendering and just use the the sound-card's converters.

Originally Posted by CingularDuality

So, what does the X-RAM do, then?

The short answer: it allows "better" effects. X-RAM is just a marketing name for some additional local RAM they put on the X-fi cards for use during effects processing... (so it isn't bottlenecked by waiting for RAM on the other side of the motherboard)

The long answer: Most audio effects, especially spatialization/reverberation effects like CMSS, require a "history" of the last few seconds of sound to be available in memory.... for example, a simple reverb ("echo") algorithm works by combining what's playing now (the sound coming directly from the source) with attenuated versions of what was played 'before' (sound that was delayed getting to you because it had go up and bounce off the ceiling first). So at any point in time, you're listening to something like this:

sample(N) + 0.5*sample(N-5) + 0.25*sample(N-10)

For a more refined effect, you need to blend more samples, which necessitates a longer 'sound history' to draw from... Pictures here if you're curious about this kind of thing.