ok. Most computer programs run on 32 bit 'fixed' number systems. Which basically means that all of the values they can represent are equally spaced. This results in linear response. The decimal point remains constant.
Floating point numbers are generally better at recreating analogue traits because they aren't restricted to linear response. This is achieved by moving the decimal place (won't explain the scheme to you, its complicated) to maintain optimal resolution.
For example, a 24bit fixed point system has a theoretical range of 144dB. A 24bit floating number system has a theoritcal range of around 1550dB.
You'd never want that sort of range for recording, but the flexibility is good for analogue modeling.
If memory serves me rightly though, floating point numbers have a massive impact on cpu's because of a problem with denormal numbers. It doesn't matter how many cores you have, because they'd eventually suffer from the same issue. I suspect that's why your 8 core got upset.
