Originally posted by TDvN57 In PS environment 8bit means 8bit per channel x 3 channels = 24 bit colour.
True - but whilst that sounds like a lot, it only allows for 256 tonal variations per individual colour, hence why banding can occur in gradations or other areas with multiple tones of the same colour. 16-bit processing, by comparison, allows for 65,536 tonal variations per colour, allowing for much smoother tonal transitions.
DxOMark's colour sensitivity score for the 645Z is 26 bits at base ISO, which - as with all cameras - is way less than you might expect given the 14-bit raw files, but considerably more than a 24-bit colour space can represent. Maintaining the highest possible bit depth throughout the workflow prior to final export ensures the best possible colour and tonal reproduction, and reduces the impact of the rounding you refer to...