nesdev.org

Even the 8086 had MUL and DIV instructions, and while it was slow, it was still faster than doing it longform.

Ultimately, multiplication of integers can be implemented in the same amount of time as the corresponding addition, by using a good deal more space. Modern machines seem to not bother, e.g. http://stackoverflow.com/questions/2181 ... modern-cpu

The PC-FX, released in 1994, used a NEC V810 which had an FPU that could do single-precision IEEE-754 math(no transcendental functions in hardware, though).

...that'd probably make it the earliest then. I don't think the 3DO nor the Jaguar had floating point support, right?

Is it still super slow compared to addition and subtraction?

This document contains the relevant numbers for just about every Intel x86 processor starting with the Pentium, plus some AMD/VIA x86 processors.