phlip wrote:General rule of thumb:
If there is some technical reason why the size of the thing you're referring to should be a power of 2, then it'll usually be a binary prefix... otherwise it'll usually be a decimal prefix.
The way RAM is built, it will always be a power of 2 bits. You technically can do other values, but it'll be just as cheap to just round up to the nearest power of 2. RAM is always measured with binary prefixes... a 1GiB stick is easier to talk about than a ~1.07GB one.
Also note that there have been problems with at least some attempts at making, for example, 384MB memory sticks. Issues I've noted included some motherboards only recognizing the last 1/3 of the memory if the stick was in the "last" memory slot, some motherboards would seize up if you stuck two such memory sticks in it, some motherboards just wouldn't recognize the last 1/3 of the memory, and one motherboard I had would decrease the memory bandwidth to 50% of ideal if one or more of these chips was present (that motherboard wanted 'matched' memory in every pair of slots. If any slot was filled with a different type of memory than its pair, or if it had a stick which did not contain an integer 2^n bits, it accessed memory in a degraded mode.)
I suspect that other motherboards could easily have other issues when presented with such memory, but fortunately, my time in hardware support ended, and I encountered no others.
phlip wrote:CPU speed is always measured with a decimal prefix... there is no reason why it should always be a power of 2 in Hz, so there's no reason why 1kHz should be 1024 Hz.
Except when it wasn't. I'm pretty sure I remember seeing adverts for two different processors which *happened* to have just over an integer number of 1024*1024 Hz, and the manufacturer tried a small run of commercials pointing out their 1024 kilo compliance. They, um, didn't do so well. Apparently, 5% isn't a significant win - especially when everyone's rounding to the nearest tenth anyway.
phlip wrote:Hard disks are a little more complicated... namely because the disks are generally decimal prefixes, but the sizes of the files on them are generally binary prefixes. This confuses people to no end, and certainly should be changed... but it's hard to change de facto standards.
Even when the defacto standard is no standard. If memory serves me correctly, hard drives were actually inconsistently marketed on this until somewhere between 40MB and 250MB. Note: I also missed the first hard drives; it's possible they were consistently labeled with the 2**20 MB rather than 10**6. (I was alive for the first hard drives, I just wasn't paying attention yet.)
Also: Very yes to the "thank god baud is gone" motion.
It seems most people would like to forget all about stop bits and all of the moments of silence by which Ethernet lives, as I don't believe anyone's mentioned them before me.
So, um, anyone up for changing our numerical base to better work with computers? I'm fine with octal or hex - although I think if we're going to seriously change to using hex, we need six new symbols, rather than just continuing to overload A through F...