The famous Pentium FDIV bug

One of my favourite CPUs from my CPU collection is the original Pentium 60MHz shown here. The reason that this model (SX835 with Gold top) is one of my favourites in not because it looks good, but rather that it was one of the Pentium models that had the famous FDIV bug.

The FDIV bug was a problem with the floating point unit in early Pentium CPUs that resulted in reduced precision of division operations (FDIV is the x86 assembly language instruction for floating point division).

For example, if you performed the following division operation: 4195835 / 3145727, you should get 1.33382044, but on a Pentium with the FDIV bug, you would get 1.33373906. This wasn’t a problem when running Windows since nearly all operations are integer calculations, but if you were using a Pentium for science/engineering applications (e.g. regression analysis), the problem was huge.

It was discovered in 1994, and caused quite a bit of bad PR for Intel since people were already complaining that early Pentiums generated a very large amount of heat (which they did). To appease everyone, Intel agreed to exchange the FDIV CPUs for new ones. Regardless, the original Pentium FDIV bug went down in history as one of Intel’s biggest blunders - even Douglas Coupland wrote about it in Microserfs:

After lunch, we went to see the Pentium movie at the theatre Intel put up in the main lobby. It was about how interactivity was going to take your life better in the future, and we couldn’t stop giggling because of all the Pentium jokes about decimal points being spammed around the Internet. You knew that every single person watching the show was, too.

“0.999999985621,” I whispered, setting everybody off into spasms again, and finally we had to leave because were annoying too many people with our giggling.

I guess if you find jokes about decimal places interesting, then you truly are a geek.