The original IBM PC had a slot for the 8087 floating-point coprocessor. This was a somewhat esoteric feature at the time – previous microcomputers had done all their floating-point in software – but it was sometimes used, and quite a few programs supported it; there is some discussion here: http://www.vcfed.org/forum/archive/index.php/t-38292.html
- Unlike nowadays, games never used floating point; even 3D calculations were always done as fixed point approximations.
- AutoCAD, as expected, could use the FPU.
- Ditto mathematical analysis and statistical programs.
- Somewhat unexpectedly, so could some databases.
- So could spreadsheets like Lotus 1-2-3. This is a tricky one. Spreadsheets should really use decimal arithmetic, because they are commonly used for financial calculations. VisiCalc correctly did this. Lotus 1-2-3 chose to use binary floating point. One could criticize this decision, but having made it, then it clearly made sense to support the FPU.
- And of course scientists compiling their own number crunching code in Fortran would benefit from the FPU.
What was the reasons for people to buy the 8087? Given that spreadsheets were considered the killer application of personal computers, I would guess that to be the answer, but I would be interested if it turns out to be something else.