And this is why O() notation drops constants. 0(bits) or O(bits/8) aka bytes are the same thing. It's also worth pointing out that in standard comparison sorts, the comparison itself is technically linear to radix too, but is treated as constant. I get why it's dropped but it's worth knowing.
Exactly. You can try to compare two bytes, but really you are comparing 8 bits, if you thought about it algorithmically. Maybe those steps are 100% parallel. But it's irrelevant to the big O. You are measuring number of steps, and it's intended to be hardware independent.