The gap between one IEEE float value and the next one is larger for large values than for small values.