| Date: | September 19, 2003 / year-entry #71 |
| Tags: | other |
| Orig Link: | https://blogs.msdn.microsoft.com/oldnewthing/20030919-00/?p=42413 |
| Comments: | 20 |
| Summary: | Slashdot is in an uproar over a lawsuit charging computer manufacturers for misleading consumers over hard drive capacity. The manufacturers use the ISO definition, wherein a "gigabyte" is one billion bytes, even though most people consider a gigabyte to be 1024 megabytes. This is a tricky one. The computer industry is itself inconsistent as to... |
Slashdot is in an uproar over a lawsuit charging computer manufacturers for misleading consumers over hard drive capacity. The manufacturers use the ISO definition, wherein a "gigabyte" is one billion bytes, even though most people consider a gigabyte to be 1024 megabytes. This is a tricky one. The computer industry is itself inconsistent as to whether the "kilo", "mega", etc. prefixes refer to powers of ten or powers of two. The only place you see powers of two is when describing storage capacity. Everything else is powers of ten: Your 1GHz processor is running at one billion cycles per second, not 1,073,741,824 cycles per second. Your 28.8K modem runs at 28,800 bytes per second, not 29,491. And your 19" monitor measures only 17.4" inches diagonally. There do exist IEC standard designations for power-of-two multipliers. A kibibyte (KiB) is 1024 bytes, a mebibyte (MiB) is 1024 KiB, and a gibibyte (GiB) is 1024 MiB. Good luck finding anybody who actually uses these terms. |
Comments (20)
Comments are closed. |
I say we stick with the traditional powers of ten for terms like kilo, mega, and giga as much as possible.
My vote would be for storage manufactures to begin using the gibibyte system. This would make their labels correct, and most people won’t even notice the difference. If asked, sales people will probably just tell them it’s the same thing, and anybody who has to ask won’t know enough to tell the difference anyway.
Bah, hard drives are so damn big these days, does it really matter if it’s only "in the neighborhood" of 120GB?
If I ever hear someone say "gibibyte" I will either assume they have a speech impediment, or are talking to a two year old in baby speak.
"Does wittle timmy-wimmy wanna waptop with 1 gibibyte of WAM? Yes, he does! Yes he does!"
*shudder*
Yeah, I’m not at all a fan of the "Gibi" prefix (assuming it’s pronounced like ‘gibby’)
And Monban…The storage guys do use 10 ^ x to calculate the size (the drives seem bigger that way), it’s the OS (and RAM marketers, as well) that uses 2 ^ x and would therefore be ‘wrong’ if the correct way to name 2 ^ x prefixes is with that baby-talk notation.
If I recall correctly, hd manufacturers give unformatted size information. The actual free space available will vary depending on how its formatted.
For me the indicator has always been the use of casing of the denominator, where lowercase refers to the 2^10xC and uppercase refers to 10^C.
It useless to pretend that kilobytes and megabytes are metric units. Metric units let you use any metric prefix; but you can’t do that with bytes. There is no definition of decabytes or hectabytes. And there certainly aren’t millibytes or nanobytes. Since bytes aren’t metric there is no reason to assume that the prefixes in use have their normal meaning.
leppie:
Uppercase abbreviations like KB or GB don’t refer to the number. The uppercase B refers to Bytes and the lowercase b refers to bits. 28Kb modem, 256MB RAM.
MilesArcher:
If the MFG is already misleading you about the starting point, you are in really bad shape when formatting takes more space away.
DavidK:
The truth always matters.
The exact size of a drive does matter in many applications. For example, a 2^30-byte drive (1 gibibyte –ptewy–) and a 10^10-byte drive (1 gigabyte) are argued to be close to the same size but you can’t combine them in a RAID system.
If you need to make a back-up of all your data, the 2.4% difference between Kilo=1024 and Kilo=1000 adds up (well it multiplies up 1.024^3 is 1.074). Imagine having to run out to the computer store to buy the 11th DVD-R because "4.7GB" on the box means "4.3GB" to Windows.
The vendors need to wake up and die right, or they’ll have the FTC on their backs.
1) No, hard drive manufacturers don’t quote unformatted capacity. They used to, quite a few years ago.
2) "Giga" meant one billion long before it meant 2^30. How many Hertz are in a gigahertz?
"How many Hertz are in a gigahertz?"
Silly me, Raymond already mentioned that.
Here’s a wild thought: What if the lawsuit succeeded, and HDD manufacturers started measuring in 2^30 gigabytes… Of *unformatted* capacity! (Not likely, but it would be an interesting case of the law of unintended consequences.)
HDD manufacturers _do_, AFAIK, still quote unformatted capacity. It would be difficult for them to do otherwise – you’ll get a different capacity if you format with FAT32 than that you get from NTFS.
Actually, I don’t know if NTFS includes the size of its disk structures in the ‘capacity’, because they resize according to how much they’re used. FAT32 has two, fixed size File Allocation Tables on the disk, plus the root directory has some minimum size.
"HDD manufacturers _do_, AFAIK, still quote unformatted capacity. It would be difficult for them to do otherwise – you’ll get a different capacity if you format with FAT32 than that you get from NTFS."
This is funny–here we have yet another case of one word meaning two different things.
Before you can even think about partitioning and "formatting" a drive with NTFS, FAT32, or any other filesystem, the drive must first have a low-level format that divides it up into sectors.
Back in the old days of MFM and ESDI drives and the like, hard drives were shipped truly unformatted, and you would see both the unformatted capacity and a "typical" formatted capacity quoted. You could format a disk with different sector sizes, and on some disks you had a choice of hard sectoring or soft sectoring. The formatted capacity would vary depending on those choices, so it made some sense to talk about unformatted capacity–which was constant regardless of your formatting scheme.
For example, the classic ST506 had an unformatted capacity of 6.38 MB and a typical formatted capacity of 5 MB.
This had nothing to do with any partitions or filesystem you would put on the drive–we’re just talking about how many sectors are available on the drive.
Nowadays, IDE and SCSI drives come already formatted with 512 byte sectors, ready for you to put partitions and filesystems on them. So you don’t see unformatted capacity quoted any more because it would be meaningless.
What with having two different definitions for MB/GB and two different definitions for "format", it’s no wonder that we’re all confused! :-)
"Your 28.8K modem runs at 28,800 bytes per second"
I wish they did :) Although on the subject it’s interesting that at least data rates are what are generally quoted – gigabit ethernet is 10 bits per cycle at 125 MHz, but the 8b/10b encoding drops the data rate from 1.25Gbps to 1.0Gbps.
The rule that makes the most sense is "what smallest units are generally dealt with". For time-based things (network speeds, CPU frequencies, etc.), the units that make the most sense are metric divisions of seconds (1ns, 1ms, etc.)
For the world of hard drives, the units you deal with are almost always sectors with each sector being 512 bytes in size. Yep, that’s a power of 2. Therefore, hard drive sizes should be quoted as relative powers of 2 for the base units, not powers of 10.
The critical "third party" to make this case? Memory sizes. Memory is addressed in sizes that are powers of 2 (64-bit or 128-bit or whatever size words). Therefore, the overall size of the stick is a power of 2 and should be (and is!) labeled as such.
Yup, HDD manufacturers are being deceptive. They know their product is a number of 2**9 sectors, but they also know that most people won’t understand that, so they go with powers of 10. In reality, this wouldn’t make any difference if HDD sizes were available at only powers of 2 like memory tends to be, since labeling 128 as 137 doesn’t matter if the numbers below and above are 64 and 256
There was once an agreement that B is bytes, b is bits; K is 1024, and k is 1000.
So, it’s a 33,6kbps modem, but a 256MB memory stick.
It would be interesting to know how many people are aware of those "finer points" and stick to them in their writing :-)
Living in the country that once had a law that pi=4(?!) I find it worrying that binary quantities are being decimalised.
While it makes sense for consumer marketing to use powers of 10 (I buy the arguments that far, for all I dislike them <g>) I worry about where the boundary will stop. I really don’t want to be targeting binary devices with decimal values, and traditionally legislators don’t let ignorance of technical matters hold them back…
> Living in the country that once had a law that pi=4(?!) I find it worrying that binary quantities are being decimalised.
That’s actually an urban legend. Check snopes for the details.
I’m surprised nobody’s mentioned one factor here which ought to be considered fairly decisive. Most consumer disk management software (I’m thinking specifically of Windows Explorer here, but almost everything else is in agreement) uses the power of 2 definition. Therefore, when somebody upgrades their disks, they’ll look at the report of how much they’re currently using in the power of 2 base to evaluate how much they need.
Also, the difference isn’t exactly small in the GB range like some people have suggested. It’s actually approximately 7% according to my quick maths. And when we start seeing consumer terabyte disks, it’ll be nearly 10% difference.
Quick point of information: what about floppy disks? A "1.44MB" floppy is actually 1440KB (using the power of 2 definition), so that’s using neither system!
"Does wittle timmy-wimmy wanna waptop with 1 gibibyte of WAM? Yes, he does! Yes he does!" heh read that without laughing
how much it’s a gygabite
it’s not right for mainstream consumers to be told the same name for different quantities. it just aint right!
I agree.