Back in 2000 I wrote some code to do some funky control systems stuff for a Tokomak nuclear fusion reactor as a research project. We happened to have a spare 486 with DOS and Borland C available that had a decent A/D and D/A converter board installed. This actually turned out to be really good for realtime code because you *knew* there was nothing else running on that system.
I had to rewrite the drivers for the converter board because it couldn't give the performance we were looking for--the settling time for the A/D conversions was too long. I figured out a way to interleave the A/D and D/A conversions so that the hardware delay for one also provided the required delay for the other, essentially doubling the sampling rate relative to the stock driver.
These days it'd probably make more sense to use an Arduino...