If you're not sure what I mean by this precision, it's the method by which the video card determines the color of each pixel displayed on your monitor. As in most situations, the higher number means the better result. To visualize this, think of a highly lit scene, with light coming through stained glass windows. The light is brightest when you stare right at it, but as you look around, the light that is casted shows varying shades of the color of the stained glass. What color is displayed is a matter of the precision with which it's calculated. In this case, the higher precision is going to deliver the truer color.
However, much like support for SM30, the FP32 math is not really in use in current or near future games. Additionally, and unlike SM30, that extra precision involves extra math that will slow performance somewhat. Most games use FP16 precision at this point, which ATI translates up to FP24 and NVIDIA runs as straight FP16, giving a small performance boost to NVIDIA. The result between these two types of precision is difficult if not impossible to see in current games. In fact, the demonstrations we saw were technical demos only; they just aren't currently in use.
It should be noted that FP32 has been a standard in movies for over 20 years and most people that deal with such things are very familiar with it. FP24 is sort of a weird middleman that game developers have not exactly latched on to, although ATI's drivers do most of the work for them. However, it's another forward-looking feature that won't really be noticeable for the near future and by then, assuming history repeats itself, both video-card companies will have new cards out.