Function:

MUL multiplies ax with a 16-bit operand to form a 32-bit result in dx:ax.
The 32-bit version multiplies eax with a 32-bit operand to form a 64-bit
result in edx:eax.

Some 386es have a problem which redirects output from the 32-bit MUL
to the wrong parts of the wrong registers.

Typically the following happens:

Properly operating 32-bit version:  Properly operating 16-bit version: