Removed the 20% estimate on faster bignum routines