Instead of generating floating point buffers, we permit the data to remain at its original precision.