20 hours? Are you sure your algorithm (or the random generator) is correct?
Long time ago, I wrote a bruteforcer for CRC32 (even before I calculated how to compute it directly
) - and on Pentium 3 1.5 GHz is takes slighly over a minute to check the 2^32 possibilities.
It generates binary data (not just "numbers and letters")... but a simple algorithm that tries random texts should be about the same, i.e. getting the result in about half a time (in average).