He's talking about fixed point, not floating point. The decimal type in SQL is an exact, fixed point numeric type. It can avoid loss of precision, like integer, but without the scale conversion step.