# You say megabyte, I say mebibyte - Page 4

•  Subject
• Author
• Posted on

## Re: You say megabyte, I say mebibyte

So you agree that the term mega is the standard that has been defined for
decades. Finally you agree that mega is 10^6 - the defined standard that has
been around for decades, even centuries. Can we stop arguing now?

Just search on google for anything mathematical - you'll find millions of
references to mega and other prefixes, dating back for many years before
computers even existed!

No, *you* are trying to claim that everyone has been using the wrong term
for centuries and only now have we realised that mega should not mean the
value it has for all those years. You have yet to show us a single
authoritative document indicating the change of use of the term mega fom the
long standing 10^6 standard.

## Re: You say megabyte, I say mebibyte

On Tue, 12 Feb 2008 10:24:36 -0000, "GT"

You know quite well that I don't agree mega is defined as
10^6 in the computer industry.

Yes, except in the computer industry, computer sciences
where it is a representation of a binary value.

Everyone has for years, if you were too young to realize
this it's not my problem.

## Re: You say megabyte, I say mebibyte

Many people agree with you. Many agree with me. Let's get on with more
important things!

## Re: You say megabyte, I say mebibyte

Hardly.

Standards authorities can help smooth the way, but they can equally well
just get in the way and botch things up.  We should shun them when they
do the latter.

-Miles

--
Success, n. The one unpardonable sin against one's fellows.

## Re: You say megabyte, I say mebibyte

On Fri, 08 Feb 2008 13:31:19 +0900, Miles Bader

Controlling bodies with authority are seldom the ones that
actually create a standard, and NIST is not a controlling
authority in the computer industry.

GT doesn't want to accept that other bodies with supposed
authority, like Webster's dictionary, don't typically invent
new words but instead incorporate them after a large enough
group of people have settled on a standard definition.

## Re: You say megabyte, I say mebibyte

More than happy to accept this. I even agree with this. The problem I have
is when somebody or some company tries to change a standard that is already
well established. This change goes against the majority and is not valid.

## Re: You say megabyte, I say mebibyte

On Fri, 8 Feb 2008 10:54:22 -0000, "GT"

Show us any evidence of a person using the term megabyte by
the definition you claim is correct prior to, oh let's pick
a date out of thin air like 1968, 40 years ago.  Actually I
don't think you can find anyone using your defintion prior
to '98, a full 30 years later except after the hard drive
manufacturers switched... and by the way there have been
several class action lawsuits about their doing so since a
hard drive is a binary storage medium.

IF you can't find one example of someone using the terms
before they were considered a standard (decades ago) you
have no evidence there was even any minority, let alone a
majority decimally-defined value for megabyte.

There was no majority defining megabyte the way you insist
we should.  Never.   once you tack byte on the meaning is a
very specific one.

## Re: You say megabyte, I say mebibyte

Several points here:
1. 10^6 for Mega is not my definition - its a standard.
2. It doesn't matter how many people get it wrong, the standard is there -
live with it.
3. All of the law suits you quote were thrown out of court and the people
trying to claim that 1024 = 1000 were laughed at.

Finally, will you please stop trying to bamboozle everyone by saying that a
hard drive is a binary storage mechanism. This is true, but irrelevant - you
are getting confused between contained values and quantity! I have asked you
this question many times and you always refuse to answer because it blows
your argument out of the water...

A coin toss is a binary result - head or tail. Therefore, by your logic, I
have to wait for 1024 tosses before I can use the SI prefix K, which means
1000a K tosse

Kony, please tell everyone how many coin tosses (binary results) are there
in 1K toss? According to the SI unit K, there are 1000, but it is a binary
value that I am counting.

The majority in question is the rest of the world, past and present. Mega is
defined as 10^6. You can tack any unit on the end of mega and you have a
'mega' of those things. 1 Mega byte is therefore a mega (10^6) bytes. End of
story. Anything else is wrong.

## Re: You say megabyte, I say mebibyte

On Mon, 11 Feb 2008 09:48:38 -0000, "GT"

Nobody declared Mega a valid prefix for -bits, -bytes, until
be representations of binary values, in a binary number
system where base 2, not base 10, determines the values.

That's not what a standard is.  A standard doesn't wait for
some group to invent it, when it has already been used for
many years.  Nobody used the term to mean what you suggest
because those using the term realized it is a decimal
representation of a binary value.

Because you don't understand that it is a decimal
representation, you don't understand why it is necessary it
be a specific binary value not based on powers of 10.  It
was clearly the standardized way to express values long
before NIST tried to step in and muddy the waters.  Nobody
was using the term as you suggested until recently, doing so
contrary to the majority use for many years, making it
impossible that the standard is actually what you claim.

It does matter, it was all the people using the term,
including scientists who realized it was a decimal
the standard when nobody was using the term the way you
insist must be the standard.

How long have people called dogs, "dogs"?
We don't need to know the exact date this began, but we see
it is a fairly standard term.  There is a more scientific
name for a dog too, but if some group demanded that from now
on, the term "dog" means  a horse, does that change what the
term means?  No.

## Re: You say megabyte, I say mebibyte

Mega is a standard prefix for any units. The computer industry can't change
that. End of story.

Yes it is.

Wrong, 99% of the population has used the term mega in its correct 10^6 for

Nice example of why a standard shouldn't be changed. We don't want to change
the standard term for dogs, just because a small proportion of society think
dogs are small black 3 legged wooden horses! A dog is a dog - 4 legged
canine. A mega is a prefix meaning 10^6. Lets leave the standards alone -
they have worked for centuries. If you want to represent things that are
2^20, then come up with a term for it, but you can't steel a long standing
SI unit and change its value! That is ludicrous and leads to confusion and
these kinds of silly discussions!

## Re: You say megabyte, I say mebibyte

Nonsense. Ever heard of ISO?

## Re: You say megabyte, I say mebibyte

On Fri, 8 Feb 2008 10:53:03 -0000, "GT"

Hmm.  Have you read any of their documents?
Here's one, now in it they wrote:
"Ultra DMA 133 Megabyte/s transfer mode"

of mega?  No.

## Re: You say megabyte, I say mebibyte

Grinder wrote:

The table on page 23, extends even further than shown. There weren't
any large DIMMs to test when the board was released.

The 82845 three slot SDRAM based controller is here. Table 12 on
page 109, shows it can handle a 1GB SDRAM. Your 256MB should be
well away from any density issue. (I have 3x512MB on my 82845
based board, installed when it was new, and 1.5GB is reported.)

If you can boot a version of Windows that can run it, I'd try CPUZ.
It has a tab in the display for memory, and selecting the slot where
the particular stick is located, should dump info on it. If
nothing is displayed, it means there is no SPD chip (small, 8 pins or
less, serial PROM with speed and capacity info). Or it could mean
that CPUZ couldn't manage to find the SMBUS, to do the necessary
reads. That doesn't seem to get reported too often. so this should
be a good tool. At startup, be patient. It takes about
30 seconds on my current machine, before the CPUZ window appears.

http://www.cpuid.com/cpuz.php

not precise enough for figuring out what is going on. If you
really need to know, take the part number off the top of a
memory chip, and Google it. You may discover its base
capacity, like 16Mx8, and 16 of those would give 256MB.

The density issue has to do with available row and column
address bits, and comes into play when modules larger than
were intended, show up. On the 440BX, where 128MB DIMMs might
have been expected, you can run 256MB DIMMs if they are
"low density". If the 256MB DIMM had 16 chips, the number
of row/column is one less, than if the DIMM had 8 chips.
Each of the 8 chips would have double the amount of memory
("higher density") and need an extra address bit when
accessing the memory array inside. Failure to solve the density
issue, may result in only half the memory being detected. (BIOS
detects by attempted probes to memory, and some of the probes
failing due to the mismatch on addressing.)

An example of another chipset with a density issue, might be
an Nforce2. A 1GB with 16 chips (two banks), using 64Mx8 chips
would work properly. The "Ebay memory" config of 1GB DDR, which
is 16 chips (but only one bank), using 128Mx4 chips, is only
half detected. The "128M" dimensions in that case, require the
extra (missing) address bit. Note that this case doesn't work
like the one quoted for 440BX. In this case, both 1GB DDR have
16 chips, so unless the seller is honest and tells you there
are 128Mx4 chips being used, you'd actually have to get the
memory chip part number, to figure it out. (Nobody includes
SPD contents, if you decoded them, would also be a dead
giveaway.)

Paul

## Re: You say megabyte, I say mebibyte

Paul wrote:

CPU-Z 1.41 reports:

Memory tab

General frame
Type: SDRAM
Size: 256 MBytes
Bank Interleave: none

Timings frame
Frequency: 100.0 MHz
tCL:  3.0 clocks
tRCD: 3 clocks
tRP:  3 clocks
tRAS: 6 clocks

all other fields on this tab are disabled

SPD tab

only Slot #1 shows as populated, with SDRAM
Module Size: 128 MBytes
Max Bandwidth: PC133 (133 MHz)
Manufacturer: Samsung
Part Number: M366S1723BTS-C75
Serial Number: FFFFFFFF
Correction: None
Registered: no
Buffered: no
Week/Year: 255 / 255

Timings Table frame
Frequency: 133 MHz
tCL:  3.0
tRCD: 3
tRP:  3
tRAS: 6

- - - - - - - - - - - - - - - - -

How is it that my only memory module has a "Module Size" of 128 MBytes,
yet my system memory reports (and memtests) as 256 MBytes?

## Re: You say megabyte, I say mebibyte

Grinder wrote:

Not a clue. I ran that part number, and the first hit says it is
a 128MB stick.

http://www.pinnaclemicro.com/frames/computer/parts.php?g=M366S1723BTS-C75&page=1

The actual Samsung site is hard to find stuff on. Not all the archived
pages work right. Again, that part number seems to be a 128MB, complete
with a typo for that particular entry (chip type and number are not
correct). One poster in Google with that part number, mentions the
RAM is single sided, so it is likely (8) 16Mx8. (I.e. Eight pieces
of 128Megabit technology.) Now, you say it has 16 chips, so something
isn't right. Maybe the module has the wrong SPD chip soldered to it ?

http://web.archive.org/web/20010709003225/http://samsungelectronics.com/semiconductors/search/datasheet.jsp?family=241

The chips themselves have part numbers on them, but my odds of finding
them on Samsung aren't very good. I was lucky that the above
archived page even rendered. Many of them do not.

Paul

## Re: You say megabyte, I say mebibyte

Ok, I officially regret my choice of title for this thread.

## Re: You say megabyte, I say mebibyte

On Wed, 23 Jan 2008 20:32:39 GMT, Grinder

But did Paul's link help?  We were kinda wanting to hear if

If not, what are the markings on the individual chips on the
memory?  IE - I'd assumed you gave the module label markings
already but not the chips themselves... thinking back it was
16 on a double sided module IIRC.

The main thing is, there's not likely anything you can do to
make this module show up as 256MB in that board if it's only
appearing as 128MB now.  The CPU-Z report was very strange
as if the detected amount and SPD amount were backwards.

## Re: You say megabyte, I say mebibyte

kony wrote:

As to the first issue, about what technology is in use, we're at a bit
of a stalemate because of conflicting descriptions on the web, and even
from CPU-Z itself.

As for the second, some general background, I think I've gained some ground.

That's correct, there are 16 chips--eight on each side.  The individual
chips are marked as: 48LC16M8A2, which suggests 16Mx8 chips, 16 of them,
for a total of 256MB/MiB.

Searching 48LC16M8A2 at google turns up a number of hits that suggest a
module made from these chips will be 128MB/MiB if it's single-sided, or
256MB/MiB if double-sided.

Assuming all the speculation above is correct, how do I match it up with
my mainboard's vocabulary?  ie, Is it 64, 128 or 256 Mbit technology?

## Re: You say megabyte, I say mebibyte

Grinder wrote:

The 16M8 is a total of 128 megabits. Dividing the number by 8, that is a
16Megabyte memory. It has a x8 interface (so 8 of them are needed to
make a 64 bit wide array, forming one "rank" of memory).

So the question remains as to why the SPD information doesn't seem
to match. When using CPUZ, you select the slot to display from a tiny
menu. Does it still say the stick is only 128MB, when in fact it
physically contains 256MB of chips ?

In years past, before SPD existed, the BIOS determined the memory
size by "probing". We used to do that on some of the proprietary
computers we used to build. If you wrote past the end, during
the BIOS test, then the read back value would fail to verify.

The BIOS today may consult the SPD, for initial setup, but may still use
probing to determine the "end of memory" on each chip select. Otherwise,
for those computers where there is a density mismatch (440BX), how does the
computer actually set up the chip select for half the normal memory
space ? It doesn't seem to rely on the SPD alone. I would consider it
a bit strange though, if it actually searched past the SPD declared
value, because that would imply the SPD was really worthless :-)
Well, at least it still has those timing values written in it.

By the way, the SPD on your stick of RAM is one of those "slightly
bogus" ones. The memory I have in this computer, has real serial
numbers written into the SPD chips. Some manufacturers prepare ROM
chips in advance, and each one of those is identical. They set the
serial number to FFFFFFFF. It is more trouble to use an EEPROM
for the SPD, and spend a few seconds writing a unique set of values
in the stick, but some follow the rules and do it right. The idea
is, it is to the manufacturer's advantage to put a unique number in
it, for easier tracking and verification during a warranty claim.
It would make it easier to identify a counterfeit, if the serial
numbers were unique and recorded at the plant.

Paul

## Re: You say megabyte, I say mebibyte

Paul wrote:

Absolutely.

I installed the module back into the computer that can only see it as
128MB, and ran CPU-Z.  The SPD tab was identical, unsurprisingly, the
system tab was a bit different:

1st Machine     2nd Machine
GENERAL FRAME
Type:              SDRAM           SDRAM
Size:              256 MBytes      128 MBytes
Bank Interleave:   none            (blank)

TIMINGS FRAME
Frequency:         100.0 MHz       132.9 MHz
FSB:DRAM:          (blank)         3:4
tCL:               3.0 clocks      3.0 clocks
tRCD:              3 clocks        3 clocks
tRP:               3 clocks        3 clocks
tRAS:              6 clocks        6 clocks