Silly merchandizing.

[grinvader]

If you don't like math or silly mental exercises, door's on your left.

So yeah... I thought about it a long while ago, but finally remembered to post it.
Basically I'm talking about the (major) difference between stupids using powers of ten instead of powers of two to count byte amounts.
You expect the usual, but they sell the reduced amount, and you get trounced savagely.

A kilobyte (2^10) and his moron cousin (10^3). 24/1024 == 2.34% lost.
A meg (2^20) and his retard little bro (10^6). 48576/1048576 == 4.63%. Shit.

I was bored one day (around... 6 years ago. Yeah, I'm really late posting this shit. Bite me.) and was wondering how ridiculously huge the trounce would get with size.

To evaluate that, you can easily compute (2^(10*x) - 10^(3*x)) / (2^(10*x)) and see where it leads. For small values, at least.
So you dive a bit in power manips and you spit an elegant 1-(125/128)^x . Much easier to hit the high values now.

Now for some fun results:
- you go grab a petabyte, you get trounced by 11.2%. Fun times.
- for a yottabyte (2^80), you'll lose 17.3%. At that size, that's a fucking lot. Way more than the whole petabyte just above.
- let's get a nice quexabyte (2^160). 31.6%, ouch.
- for the killing move, an amount of bytes noone made a prefix for yet. 2^300... is twice as big as it's decimal counterpart, 10^90. 50.9% lost to be precise.
So when you go get a nice whatchamacallitbyte, you'll actually get less than half that.

You can now resume your usual routine. Don't force yourself to flame me over how stupid this is, yadda yadda, not enough molecules in the universe to make that amount, yadda yadda, been there done that.

[Nach]

[Rashidi]

i remember encountering somethin like this:

• 10^(3*n) defenders assert that 2^(10*n) were merely programmer lazzy-ness to do divisions

i very tempted to call em shift haters

[Gil_Hamilton]

I wonder how many of his younger readers don't get the "Intel kilobyte" and ranted about him being an AMD fanboy... I mean, that was during the final days of DOS. In computer terms, it's ancient history.

[funkyass]

its was a significant bug.

xkcd is aimed at nerds, so most fanbois fly right under the jokes.

the 1000 vs 1024 thing is nothing more than caveat emptor.

[Gil_Hamilton]

its was a significant bug.

xkcd is aimed at nerds, so most fanbois fly right under the jokes.

Yeah, but there's a whole generation of nerds now that never touched a P1.

It's sort of like taking shots at the 487 "coprocessor."

[DancemasterGlenn]

The important thing is that someone, somewhere, thought that was fantastic.

[Gil_Hamilton]

Me, actually. I thought the Intel kilobyte was the funniest one, along with the imaginary kilobyte.

ANYWAYS... base 10 VS base 2. Never should've happened.
Especially not since it ONLY happened in hard drives. A kilobyte of RAM is still 1024 bytes. 650 MB on a CD is still made of 1024-byte kilobytes.

I'm rather amazed that the hard drive manufacturers haven't started using totally arbitrary definitions of byte too.

[blargg]

My take:

"KB" is the name of a unit. One KB = 1024 bytes, by definition. The name was chosen to suggest kilo byte since that approximates it, but the K does NOT stand for kilo. MB and GB are similar. These units were created to precisely measure the address space of digital systems, like the physical space of a processor or an amount of memory, without having decimals or rounding. That's their only advantage over using powers of 10; you can say 32KB instead of 32.768 kilo bytes, which is little better than just saying 32768 bytes.

For measuring information quantities, the proper solution IMO is to measure in bits, as has been done in communications all along. This is where SI prefixes are appropriate. kbit = 1000 bits, Mbit = 1000000 bits, etc. There's no reason to bring bytes into the picture, as those are an artifact of addressing, and only appropriate where address spaces or similar are involved and exact values are desired, in which case you use KB, MB, etc.

And trying to make dogfood-sounding units for 1000 bytes, 1000000 bytes, etc. is stupid and confusing, and serves little value.

[paulguy]

Using 1000 instead of 1024 for data rates (kbit) kinda sucks. Is this really how it's done? I've always assumed both were 1024 for conversions. The other way, conversions are much more difficult.

My opinion is that bits/bytes/whatever should always be base 2 instead of base 10, but that's bad from a business perspective, damn marketers.

[Gil_Hamilton]

Using 1000 instead of 1024 for data rates (kbit) kinda sucks. Is this really how it's done? I've always assumed both were 1024 for conversions. The other way, conversions are much more difficult.

My opinion is that bits/bytes/whatever should always be base 2 instead of base 10, but that's bad from a business perspective, damn marketers.

You never noticed your internet access was spec'ed in bits per second?

[blargg]

Using 1000 instead of 1024 for data rates (kbit) kinda sucks. Is this really how it's done? I've always assumed both were 1024 for conversions. The other way, conversions are much more difficult.

Use a calculator. Why do you need to be converting kbits to KB that often anyway? Maybe the desire for power-of-2 bases is like the one in the US to continue avoiding the Metric system.

My opinion is that bits/bytes/whatever should always be base 2 instead of base 10, but that's bad from a business perspective, damn marketers.

And it's bad because it's easier to mentally convert between the various SI-prefixed units. 1.234 Mbit = 1234 kbit = 1234000 bits. Just like 1.234 MHz = 1234 kHz = 1234000 Hz, when dealing with clock rates.

[Rashidi]

so, that remain fatihful to power 2-base, was only memory card, mainly caused by technical reason? (my 1 GB memory card)

heck, hardisk was made with somewhat similar addressing calculation as memory card does...

[blargg]

heck, hardisk was made with somewhat similar addressing calculation as memory card does...

Its size is dictated primarily by the disk, which isn't related to a power of 2, rather the data density and area of the platter. The capacity of a solid-state memory chip is almost always a power of two, and most memory devices use a power of two count of chips, each the same size, so the resulting total capacity is also a power of 2. But I'd argue that even solid-state storage device capacities should be measured in Gbits or Tbits, since the exact size is rarely important, and less than a power of 2 anyway due to formatting overhead, unlike with volatile RAM modules.