Talk:Binary prefix/Archive 8

From Wikipedia, the free encyclopedia

 This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.
← Archive 7 Archive 8 Archive 9 →

Contents

The 'as early as 1974' claim

the link show few things. The text use, int he actual specification of the different model, unit like K and M for data rate (see page 6, bottom: "Rata rate: 1.2M bytes per second" (note that they do NOT use MB, but still use M), Yet the Capacity is always given as n x 106 bytes, never has n MB or even n M bytes, which surely would have been more convinient. That shows that 1/ They expected the reader to be familiar with M, and bytes since they use them for data rate, and 2/ they chose not to use such notation to describe the capacity of their product, using a n x 10p bytes notation instead. In fact the only place where 'MB' is present is on the first graphic—presented as the first page in this pdf, although the next page is page #6— not even on the scale which read 'CAPACITY (million bytes)', but on top of each bar of the bar graph. That typography not being re-use anywhere in the document hint that is more an aesthetic choice made by the graphic design people rather than a policy, engineering choice or a customary usage. If anything this document actually show that the hard drive industry, at that time, did not have the habit of referring to the capacity of their device with MB, metric or otherwise. -- Shmget 10:21, 14 June 2007 (UTC)

Thanks for acknowledging that in 1974 CDC used MB in a decimal sense to characterize their product line. The Disk/Trend report establishes that it was in general use by 1977. What is the earliest use you can find of an OS using MB in binary sense?Tom94022 17:56, 14 June 2007 (UTC)
No need to thank me for something I haven't done.
"What is the earliest use you can find of an OS using MB in binary sense?" What is the earliest you can find of an OS using any kind of metric prefix, for memory size ? for file size ? for file system size ? , for device size ? What does Posix says about, for instance, du and df, and the meaning of -k in their command line ?
How was block device size represented in Unix in 70's,80's,90's.... (hint that involve a 'block', that happen to be a power of 2 in size, typically 512 bytes). Btw, who did those 'confused' consumers sued exactly ? How has to put a footnote to explain what the heck they mean on their package ? And the cherry on the cake. How does Hard drive manufacturer indicate the amoung of Cache Memory in theire drive ? you guessed it in MB, the real ones, not the 'metric' ones. When they were selling disk in the megabyte ranges then MB should be read as metric MB, but as soon as they shift to the gigabyte range, then they switch and start using MB in it's common sens, like everybody else. No wonder they got sued. -- 20:41, 14 June 2007 (UTC)
Posix is 1988. If that is your earliest date I rest my case.
Posix is just the setting in stone of what has been common use. Posix didn't 'invent' any things in that regard, just codified it. -- Shmget 06:26, 15 June 2007 (UTC)
BTW, MS/PC DOS FDISK at some point added (not later than DOS 6 in 1993) a line "1 Mbyte = 1084576 bytes" but that is well after the phrase Mbyte appeared in the Memorex annual reports and elsewhere in the decimal sense. BTW, I did look into some UNIX's, MVS, IBM DOS, CPM, Apple DOS and others OS's and concluded they displayed digits without prefixes.
For one thing, these 'digits' are never 'prefixed'. when did you see "MB256" ? But, sure Unix display the number of 512 bytes block without 'prefix' nor suffix for that matter. and df -k display a number of KB without suffixing... and your point is ? How exactly does it change the fact that file size are expressed using power of two multiple unit ? How does it change the fact that, from an operating system point of view, every structures on disk or in memory are power of two in size. a block is 512 bytes, most system use a paging mechanism based on 4K pages, The minimal unit of allocation is a power of two (depend on the file system, but goes from 1K to 32, 64K...). The only time in this world that a normal human being is confronted with a size in metric MB/GB is when they read the box and the marketing fliers of a hard drive... thankfully, as soon as the said hard drive in mounted and formated, that is just a bad memory..... and what the OS is telling you is : How many 1GB file can I store in this disk! that is the only thing a user care about, hard drive are used to store files in them and file size are NOT in metric MB/GB.
Don Rumsfeld got to answer a question with a series of rhetorical questions and you see where that got us. So if you don't have any serious information, I suggest u stop wasting our time. Tom94022 21:58, 14 June 2007 (UTC)
You refer to yourself with a plural now ? Or you have been mandated to speak for a 'group' of some kind ? These questions are not rhetorical, except to the extent that you already know the answer to them, just conveniently ignore them (that also fit well with the notorious character mentioned above), but we should probably stay away from political references, or someone will have to invoke Godwin's Law pretty soon -- Shmget 06:26, 15 June 2007 (UTC)
Again we agree that UNIX did not display any prefixes (suffix, if you prefer) until very late with regard to the HDD industry's general use of MB in the decimal sense. BTW, it is not just the HDD companies being sued, please check out the Dan Sep 2003 litigattion.
you mean :"In September of 2003, Lanchau Dan, Adam Selkowitz, Tim Swan and John Zahabian filed a lawsuit against Dell, Inc., Apple Computer Inc., Gateway, Inc., Hewlett-Packard Co., IBM Corp., Sharp Corporation, Sony Corp. and Toshiba Corp claiming their advertising deceptively exaggerates the real capacity of their hard drives." , yes point in case. -- Shmget 23:33, 15 June 2007 (UTC)
We agree again, Dell, Apple et al, at least within some portion of their company think the prefixes M and G have a decimal sense. The fact that they are sued doesn't make the plaintiff correct; it might be interesting to see how they answered the complaint.Tom94022 17:43, 16 June 2007 (UTC)
Certainly, 'some portion of their company' i.e. the marketing dept, tend to use whatever look better. I would certainly agree that Apple et al. should have marketed the hardrive capacity with a unit consistant with their O.S. i.e. having a 152.5 MB dirve not a 160 metric MB.
"it might be interesting to see how they answered the complaint." Yes, I agree. I've looked for it, but I was not able to find readily accessible info in this. There are many press article about the complaint, but nothing about how it was disposed (settlement, summary judgment, trial ?... I could not even find the case number. If somebody has a subscription of FindLaw or something like that, it would be interesting to get the details... well, interesting for the sake of curiosity) -- Shmget 23:24, 16 June 2007 (UTC)
If you have any facts, instead of guesses, POV and rants about the display by OS's and utilities of MB (and then GB)
You are the one ranting about how OS's developer, the IC industry,.. did not comply with the hard drive industry wishes.
I, and I suspect any other reader of this thread, would appreciate it; otherwise you are wasting our time.Tom94022 18:51, 15 June 2007 (UTC)
I just went thru the timeline above. There is not one mention of an OS's or utility's usage of MB. I believe somewheres on this page u state, "Of course it matter, that is the ONLY thing that matter here. the significance of 'MB', ..." I'll give u Windows 3.11 in 1993. Can u find any such usage prior to 1990?Tom94022 19:05, 15 June 2007 (UTC)
Since you insist, screen capture of an Apple IIgs, Prodos16, 1987. Just because you selectively looked for illustrations of one side of it, doesn't mean that there is no evidence of the other side.... And as far as 'guesses' goes. I actually owned and operated computers during the whole period. This is not 'history' to me, I lived through it.
also: In "ProDOS Technical Reference Manual (c) 1985 Apple", page 5: "1.1.2 Volume and File Characteristics ... Maximum capacity 32 megabytes on a volume ... afile can hold up to 16 megabytes of data" page 19 : "The largest possible standard file has a master index block that point to 128 index block. Each index block point to 256 data block and each data block can hold 512 byte in data" (just in case you try to argue that the 16 MB above are metric megabyte....), p163 "A tree file contains more than 128K bytes, and less than 16M butes ($20000 < EOF < $1000000)", p171 " B.4.2.1 The sstorage-type attribute: ... $3 indicate a tree file entry (127K < EOF < 16 M bytes).
Thanks for moving the binary MB timeline to 1986. I had guessed that Apple was the first OS to use MB but didn't find any examples. That still leaves 10 year gap between the HDD industry's use and the advent of binary MB. Tom94022 17:38, 16 June 2007 (UTC)
Actually, If you want to move the timeline, you might want to concentrate of the IC industry. IIRC the first megabyte-memory chip came out in 1984. The first Megabit memory chips had to come earlier than that -- Shmget 23:24, 16 June 2007 (UTC)
That being said, the crux of the argument is that it would have been non-sens, and for that matter extremely confusing, for an OS to start using metric MB.. that would have resulted in the generalization of the notorious 1.44MB floppy hybrid (1440 K). Again memory is in KB/MB/GB. Files use the same unit for obvious technical and practical reason. The size of a physical disk, one way or the other, only matter when you buy it, after what that matter is how much of a file you can put on it. From the day computer were based on binary state, the unit of size was bound to be using power of two units... and that is exactly what happened. The so called 'consumer confused', in all the example given on this page, is limited to consumer being confused about the hard drive manufacturers notations, not about the rest of the computing industry -- Shmget
You make your POV clear. Your whole argument about things fitting is specious. By the 1980's, internally everything was Hex without prefixes (suffixes, if you prefer - please don't get semantical, we are talking about the M as in both 160M and 160 MB).
I have the souce of the Aplle II Monitor, and I beg to differ. Hexadecimal constant were all prefixed by '$', and decimal constant were used.
Decimal was rarely if ever used internally except in anticipation of display; that's because consumers (as opposed to programmers) understand decimal.
You are playing on the sens of 'decimal'. using it to qualifiy a sequence of number or to qualify a metric unit as 'decimal' unit. These are two different concept. Just because I count eggs using the unit 'dozen', does not imply that I should display the number of dozen in base twelve. -- Shmget 23:24, 16 June 2007 (UTC)
If a programmer worried about file space, he/she converted file size to number of blocks
That is he take the size and shift it by 9 bits.. or he take the size in KB and multiply by two.... -- Shmget 23:24, 16 June 2007 (UTC)
and figured out if you had enough blocks all in Hex (Again, please don't get semantical on me; I use blocks generically to account for differences in file systems between OS's).
No, not necessarily in Hex -- Shmget 23:24, 16 June 2007 (UTC)
Calculations using prefixed sizes may not fit because of truncation and blocking, e.g., a 499K file not to fit on a 500K space. BTW, the mixed usage of decimal digits with binary prefixes is equally as likely to cause consumer confusion since, e.g., a 0.5M file will not fit into 500K remaining space (both in binary sense).
Of course it won't because 1/2 MB is 512 KB. but what's you point? that if you take a MB in a 'binary sens' and do math on it as if it was a metric MB, you get a wrong result ? duh!. The example above is the same as saying that 1MB will not fit on 1000KB (both in binary sense). Of course it won't Just like 1024 kB will not fit in 1MB (both decimal sense)!!!
As the illustrations in the article show, the consumer confusion occurs when the consumer buys a 160GB HDD but system then reports it both as 149.05GB and 152655MB.
I believe we all agree on that. (except to the extent that 160 metric GB is actually 152.58 GB, not 149.05 GB, so the delta is not entirely attributable to a unit conversion, hence the 'confusion' would remain due to people confusing 'raw' space nd 'formateed, usable space). The argument is which of the two information is 'confusing', what size is the consumers most likely to care about: the size reported on the box or the size he can actually use.-- Shmget 23:24, 16 June 2007 (UTC)
Based upon the current evidence, it is clear that the OS's folks adopted the term MB in a binary sense many years after the HDD industry had had standardized on MB in a decimal sense.
'standardized' I have plenty of magazine and magazine ads for harddrive, that illustrate the the use of these unit was all but standard, even in the industry. and to top that the spec of the CD-ROM are a clear counter example to that claim. -- Shmget 23:24, 16 June 2007 (UTC)
Beside, consumers (as in the general public - the segment were you can find some confusion - did not commonly use hard-drive until at least 1983. (with the IBM PC/XT), and by then teh size was a few MB at best, which could and was easely refered as 1000's of KB. of 1000's of 'blocks'.
Whether it was a good idea or a bad idea is irrelevant, it is the source of consumer confusion BTW, I suspect careful research would show the same sequence occurred with k/K to K in FDDs and G to G in HDD.Tom94022 17:38, 16 June 2007 (UTC)
That is very doubtful of FDD, since its invention post-date the use of K=1024 by half a decade, add another 5-10 years for it to really 'hit the public', beside there is no serious argument that KB=1024, especially sinc KB is not a 'prefixed B unit as some what to describe these unit,, since K is not a SI prefix....
That is irrelevant for GB, TB and all subsequent unit. It is a consistency issue. Regardless of any one's preference. the one thing we all agree - I think - is that mixing these unit is completely dumb, like the notorious 1.44 MB disk... Within a given environment it is either entirely binary or entirely decimal. Of course OSs, when they eventually needed to represent GB sized object, used binary GB not metric GB. It's an obvious extension of the use of the units KB, then MB, then... -- Shmget 23:24, 16 June 2007 (UTC)


History of binary K and M

Tom, I don't follow your line of thought.

K and k were used in the mid 1960s as 1024. By the late 1960 computers came with over a meg of RAM. Are you saying the M for 1024*1024 did not come into use until the 1980s?

Exactly my point with regard to M and in particular with regard to MB, the initial source of "consumer confusion". Please see the timeline above; the earliest use of MB right now is Apple IIgs in 1986 and the earliest use of MByte might be 8086 in 1981.Tom94022 19:51, 17 June 2007 (UTC)

The terms like 64 K words, or 64 Kbytes are more common than 64 kilobytes in the 1970s. The ads in the first 2 years BYTE magazine use something like "4K words (4096 bytes)". Kilobyte and Megabyte became common in the very late 1970s.

No question about the dual meanings of kilobyte and all its abbreviations becoming common in the 1970s. There is no evidence of MByte and MB in a binary sense before the 1980s. I am less certain about binary megabyte, but the article is talking about M not mega. Tom94022 19:51, 17 June 2007 (UTC)

In the 1960s disk had over a million bytes of storage. Most of the disk references from the 1960s are in Mbits. The disks were formatted differently for programs vs. data. They only used 6 bits for characters. (No lower case.) --

I suggest most of the 1960's references come from IBM and they are all millions of bytes. Check out the IBM literature on the 1311, 2314 and 2305 at the [| IBM Storage product profiles] Tom94022 19:51, 17 June 2007 (UTC)

SWTPC6800 05:36, 17 June 2007 (UTC)


Tom, Here is an early 16 Mbyte example, it used 64 Kbit MOS dynamic RAM. The prototype was only 2 Mbytes total. I bought 8 Motorola 64K D-RAMs in 1981 for $100. A year later they were $10 for 8.

MIP: A flexible, microprogrammable image processor

Gonzalez, Manuel Gonzalez, Jorge IBM-Madrid Scientific Center, Madrid, Spain

This paper appears in: Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP '82.

Publication Date: May 1982

Volume: 7 On page(s): 1211 - 1214

Abstract This paper presents an experimental, micro-programmable processor intended for the efficient processing of multiband images. The processor is attached to a commercially available image display terminal and operates under the control of a host computer. Flexibility and programming easiness have been the most important design objectives. Main characteristics of the processor are the following: Four-stage pipe-line architecture, dedicated arithmetic and logic unit for address computation, fixed-point and floating-point instructions, microprogram memory separated from image store, look-up tables for fast implementation of some functions and 16-Mbyte image storage addressing capability. The processor performs about 3.6 millions of fixed-point instructions per second. -- SWTPC6800 21:41, 17 June 2007 (UTC)


Here is a Mbyte from 1972.

Lin, Yeong; Mattson, Richard (September 1972). "Cost-performance evaluation of memory hierarchies". Magnetics, IEEE Transactions on 8 (3): pg 390-392. IEEE. 
IBM Thomas J. Watson Research Center, Yorktown Heights, NY
"The stack processing method is used to compare the cost-performance ($/access) between the following: 1) two-level and three-level hierarchies; and 2) hierarchies using random access and serial access memories as a backing store. It was found that busswidth between levels in the hierarchy strongly affects the system cost-performance, and the access time ratio between the backing store and buffer somewhat less. With the technology cost assumptions used in this study, the three-level hierarchy does not have as good a figure of merit ($/access) as the two-level hierarchy unless the access time ratio exceeds about fifty. Also, random access devices are advantageous over serial access devices for backing store applications only when the memory capacity is less than 1 Mbyte. For capacities of 4 Mbyte and 16 Mbyte serial access stores with shift register lengths of 256 bit and 1024 bit, respectively, look favorable."

Gordon Bell used megabyte in this 1976 paper.(Written November 1975) "memory size (8k bytes to 4 megabytes)."

DOI= http://doi.acm.org/10.1145/800110.803541
Bell, Gordon; Strecker, William (November 1975). "Computer structures: What have we learned from the PDP-11?". ISCA '76: Proceedings of the 3rd annual symposium on Computer architecture: pg 1-14. ACM Press. “memory size (8k bytes to 4 megabytes).” 

-- SWTPC6800 02:49, 18 June 2007 (UTC)

Tom, Could you look at this proposed update to the History section? User:Swtpc6800/Binary

-- SWTPC6800 03:38, 18 June 2007 (UTC)


There are three styles mentioned in the History section: binary, decimal, and truncation. An example is given ("The HP 21MX real-time computer (1974) denoted 196,608 as 196K and 1,048,576 as 1 M"), but the 1M could be any of the three styles. What evidence is there that this was not just another instance of truncation? — Omegatron 01:39, 22 June 2007 (UTC)

You can't tell which method was used for values like 32K or 1M. You just have to look at the other values an assume they used one method. The HP 21MX also used 131K. Look on page 18 in the HP Journal link.
Thanks for the earlier comment on truncation. It turned out to be a common style. -- SWTPC6800 04:13, 22 June 2007 (UTC)

Binary Prefix Confusion from 1968

Morrison, Donald (March 1968). "Letters to the editor: Abbreviations for computer and memory sizes". Communications of the ACM 11 (3): pg 150. ACM Press. 

Free copy here. (Page 150) [1]


Editor:

The fact that 210 and 103 are almost but not quite equal creates a lot of trivial confusion in the computing world and around its periphery. One hears, for example, of doubling the size of a 32K memory and getting 65K (not 64K) memories. Doubling again yields a 131K (not 130K) memory. People who use powers of two all the time know that these are approximations to a number they could compute exactly if they wanted to, but they seldom tale the trouble. In conversions with outsiders, much time is wasted explaining that we really can do simple arithmetic and we didn't mean exactly what we said.

The confusion arises because we use K, which traditionally means 1000, as an approximation for 1024. If we had a handy name for 1024, we wouldn’t have to approximate. I suggest that κ (kappa) be used for this purpose. Thus a 32κ memory means one of exactly 32,768 words. Doubling it produces a 64κ memory which is exactly 65,536 words. As memories get larger and go into the millions of words, one can speak of a 32κ2(33,554,432-word) memory and doubling it will yield a 64κ2 (67,108,864-word) memory. Users of the language will need to have at there fingertips only the first nine powers of 2 and will not need to explain the discrepancies between what they said and what they meant.

Donald R. Morrison
Computer Science, Division 5256
Sandia Corporation, Sandia Base
Albuquerque, N, Mex.
If only...  :-) — Omegatron 01:36, 22 June 2007 (UTC)

Third External Link Broken

{{editprotected}}

I've been a Wikipedia user, and never a contributor (i.e. talk newbie), but I thought I'd mention that the third external link is broken.

In the external links section, the following line...

* [http://meta.ath0.com/articles/2005/02/23/a-plea-for-sanity A plea for sanity]

...should probably be changed to...

* [http://meta.ath0.com/2005/02/23/ A plea for sanity]

BLLuten 06:39, 23 June 2007 (UTC)

Y Done Adambro 11:17, 23 June 2007 (UTC)
This should not have been protected indefinitely. Sorry if it's my fault for not including the right template on the page, but someone should have unprotected it after a few days. — Omegatron 17:37, 23 June 2007 (UTC)

IEC binary prefix adoption

I have been trying to measure the adoption of the IEC binary prefixes in the computer world. There is no legal obligation to use this IEC and ANSI/IEEE standard. Every IEEE standard has these disclaimers. "Use of an IEEE Standard is wholly voluntary." "The existence of an IEEE Standard does not imply that there are no other ways to produce, test, measure, purchase, market, or provide other goods and services related to the scope of the IEEE Standard." After the ASME vs Hydrolevel antitrust case [2] was upheld by the Supreme Court[3] standards organizations are abundantly cautious about pushing their standards.

The use of MB and megabyte for binary values is more than common usage, there is 50 years industry practice codified in ANSI/IEEE and other standards. The previous standards formally defined what the industry was already using. Coining new terms like mebibyte is an attempt to change industry practice. In 1984 the ANSI/IEEE Std 91-1984 and IEC 60617-12 standards recommended that everyone start drawing schematic symbols of AND gates as a square box with an ampersand in it. Changing something that the industry thinks is working is very difficult.

The only significant usage is in elite standards groups. I would have the say the adoption of the IEC binary prefixes is minuscule and static. One of the major points on the consumer confusion argument was the difference between RAM, floppy disk and hard disk measurements. Floppy disks are gone, and all hard disk now come with a disclaimer stating that a GB is a billion bytes. The rest of the computer industry is staying with previous ANSI/IEEE standards that define KB, MB and GB as binary units.

Survey Method

I looked at the IEC binary prefix use by news, technical and reference publications; manufactures of computer components, semiconductors, and computer systems; operating system and application software; retail sales of computers and software; and other standards organizations.

The measurement methodology should give weight to the importance of a source. An established technical publication with a monthly circulation of 400,000 readers and an online edition is more influential than a small circulation newsletter or a personal blog. Intel has more clout than the 87th largest maker of novelty USB flash memory drives.

One search method used was Google's Advanced Search with the site limited to the company domain. For example: Mbyte site:forbes.com . The results have to be analyzed to eliminate blog content and un-edited press releases. Forbes has over 1600 hits on MB, and 19 on Mbyte. Searching MIB is a problem because it is a popular acronym (Management Information Base). Add another term such as DRAM, 64 or 512 to MIB to get better results. Forbes uses MB and GB.

Rather than just making up a list in each category, I used rankings from various sources. The lists are not perfect, but the top companies in each category make the lists.

Computer Industry Press

The IntelliQuest CIMS Business Study is a survey publishers and advertiser to determine pecking order of computer magazines. It claims the three largest publishers are International Data Group (IDG), CMP and Ziff Davis Media.

Here are the circulation numbers for the North American magazines of these publishers. This is not a definitive list of computer industry magazines but it is a good cross section. These magazines also have online versions.

None of these magazines use the IEC prefixes. They all use MB and GB with the exception of two CMP magazines; InformationWeek and EE Times also use Mbyte and GByte.

International Data Group (IDG): CIO (twice per month - 140,000}, Computerworld (weekly - 180,038), Macworld (monthly - 350,000), PC World (monthly - 853,952), InfoWorld (was around 300,000 - now web only)

CMP: InformationWeek (weekly - 440,000), EE Times (weekly 150,000) ,Dr. Dobb's Journal (monthly 120,000)

Ziff Davis Media: PC Magazine (bi-weekly 700,000), eWEEK (weekly - 400,100)

Business Press

The Wall Street Journal, Business Week, Forbes and Fortune all use the traditional MB and GB. All of the mainstream newspapers and magazines also use MB and GB.

Technical Journals

The IEEE and ACM have their journals on-line to members and subscribers. The contents can be searched by Google but the articles can not be viewed. Some universities and businesses have access in their libraries. The most effective searching required access to the IEEE and ACM search tools.

The terms Kbyte and Mbyte were the most common units in the 1970s and 1980s and are still very popular. The IEEE Computer Society magazine, Computer, still uses them. Current technical journals allow the authors to select the unit. There are some articles using the IEC prefixes but Mbyte and MB are still more common. The April and June 2007 issues of the IEEE Computer Society magazine Computer do not have any IEC binary prefixes, only Kbyte, Mbyte and Gbyte. These units are recommended by the IEEE Computer Society Style Guide.[4]

"MB: megabyte; use Mbyte (40-Mbyte hard disk, 12 Mbytes of memory)"

Here is a sample of current (2006-2007) technical articles from IEE and ACM publications.

Rusu, Stefan; Tam, Simon.; Muljono, Harry.; Ayers, David.; Chang, Jonathan.; Cherkauer, Brian.; Stinson, Jason.; Benoit, John.; Varada, Raj.; Leung, Justin.; Limaye, Rahul Dilip.; Vora, Sujal. (Janurary, 2007). "A 65-nm Dual-Core Multithreaded Xeon Processor With 16-MB L3 Cache". Solid-State Circuits, IEEE Journal of 42 (1): pg 17-25. doi:10.1109/JSSC.2006.885041.  The authors are with Intel Corp.

Hamamoto, Takeshi; Minami, Yoshihiro; Shino, Tomoaki; Kusunoki, Naoki; Nakajima, Hiroomi; Morikado, Mutsuo; Yamada, Takashi; Inoh, Kazumi; Sakamoto, Atsushi; Higashi, Tomoki; Fujita, Katsuyuki; Hatsuda, Kosuke; Ohsawa, Takashi; Nitayama, Akihiro (Match 2007). "A Floating-Body Cell Fully Compatible With 90-nm CMOS Technology Node for a 128-Mb SOI DRAM and Its Scalability". Electron Devices, IEEE Transactions on 54 (3): pg 563 - 571. doi:10.1109/TED.2006.890597.  The authors are with Toshiba Corp.

Pilo, Harold; Barwin, Charlie; Braceras, Geordie; Browning, Chris; Lamphier, Steve; Towler, Fred (April 2007). "An SRAM Design in 65-nm Technology Node Featuring Read and Write-Assist Circuits to Expand Operating Voltage". Solid-State Circuits, IEEE Journal of 42 (4): pg 813 - 819. doi:10.1109/JSSC.2007.892153. “This paper describes a 32-Mb SRAM that has been designed and fabricated in a 65-nm low-power CMOS Technology.”  The authors are with IBM Corp.

Birrell, Andrew; Isard, Michael; Thacker, Chuck; Wobber, Ted (April 2007). "A design for high-performance flash disks". ACM SIGOPS Operating Systems Review 41 (2): pg 88-93. doi:10.1145/1243418.1243429. “A typical 1 GByte device, the Samsung K9W8G08U1M, consists of two 512 MByte dies in the same package.”  The authors are with Microsoft Research.

Gebis, Joe (April 2007). "Embracing and Extending 20th-Century Instruction Set Architectures". Computer 40 (4): pg 68-75. IEEE Computer Society Press. doi:10.1109/MC.2007.124. “For example, Cray’s most recent vector processor, the X1, has 2 Mbytes of L2 cache that include vector references on each node.”  The authors are with the University of California, Berkeley.

Goldstein, Harry (January 2006). "Too little, too soon - solid-state flash memories". IEEE Spectrum 43 (1): pg 30-31. doi:10.1109/MSPEC.2006.1572348. “This paper discusses the disadvantages of Samsung's new NAND flash-based solid-state disks (SSD), which range in capacity from 4 to 32 GB aimed at notebook, subnotebook, and tablet computers.”  Harry Goldstein is a Senior Associate Editor for the IEEE Spectrum.

Moreira, Jose; Brutman, M., Castaños, J., Engelsiepen, T., Giampapa, M., Gooding, T., Haskin, R., Inglett, T., Lieber, D., McCarthy, P., Mundy, M., Parker, J., and Wallenfelt, B. (November 11 - 17, 2006). "Designing a highly-scalable operating system: the Blue Gene/L story". Proceedings of the 2006 ACM/IEEE conference on Supercomputing, Tampa, Florida: ACM Press. doi:10.1145/1188455.1188578. ISBN 0-7695-2700-0.  The authors are with IBM Corp. "completing on-chip memory hierarchy is 4 MiB of embedded DRAM" The IEC binary prefixes are explained in footnotes like this: " MiB = 1,048,576 (mebibyte - http:/en.wikipedia.org/wiki/Binary_prefixes) "

Ellsworth, David; Patrick J. Moran (October 2003). "Accelerating Large Data Analysis By Exploiting Regularities". IEEE Visualization: pp. 561 - 568. NASA Ames Research Center. “The file server had two 1 GHz Pentium III processors, 2 GiB of memory, and eight 120 GB 5400 RPM IDE disks combined into one logical volume using software RAID-5.” 

Ingalls, R.G.; M. D. Rossetti, et al.. "Integration of the FreeBSD/kb TCP/IP-stack into the discrete event simulator OMNET++". Proceedings of the 2004 Winter Simulation Conference, Institute of Telematics, University of Karlsruhe.  "All tests for the evaluation were performed on a Xeon dual processor system running at 2.2 GHz with 4 GiB RAM (1 GiB = 1024 MiB, 1 MiB = 1024 KiB, 1 Kib = 1024 bytes). ... The average memory consumption per host was determined to be around 20 KiB. Additional memory of 150–170 KiB is required per (bi-directional) TCP connection"

Dave Chinner, Jeremy Higdon. "Exploring High Bandwidth Filesystems on Large Systems". . Silicon Graphics, Inc. "JFS was unable to exceed an average of 80MiB/s write speed in all but two of the many test points executed, and Ext3 did not score above 250MiB/s and decreased to less than 100MiB/s at sixteen or more threads."

Lemire, Daniel; Owen Kaser (2006-10-20). "Hierarchical Bin Buffering: Online Local Moments for Dynamic External Memory Arrays". cs/0610128. “Throughout, we use the preferred IEC units[5] of KiB, MiB, GiB and TiB, which respectively measure storage in groups of 210, 220, 230 and 240 bytes.” 

J. Andy Harriman. "A Reconfigurable Four-Channel Transceiver Testbed with Signalling-Wavelength-Spaced Antennas". . University of New Brunswick (Master's thesis) "a Texas Instruments (TI) TMS320C6701 32-bit DSP which operates at 167 MHz, and up to 128 MiB of RAM to name but a few of its features."

Pramod Korathota. "Investigation of Swarming Content Delivery Systems". . University of Technology, Sydney "The distributor of some content runs an application which splits the content into small blocks, usually 256 KiB to 1 MiB in size."

Martin, P. (2002-04-03). "A pipelined hardware implementation of genetic programming using FPGAs and Handel-C". Genetic Programming: 5th European Conference Proceedings 2278: 115-121.  "The Celoxica RC1000 board has 8 MiB of SRAM arranged as 4 banks of 2 MiB that can be directly addressed by the FPGA, and each bank is configured as 512 Ki 32bit words."

Thorvald Natvig (2005-06-09). "Blue Gene/L". . Norwegian University of Science and Technology Retrieved on 2007-07-14. "The L2 cache is only 2 KiB large, and is more accurately referred to as a large hardware prefetch buffer. The onboard 4 MiB eDRAM component can be split as desired between L3 cache and addressable memory."

Leopold, M.; M. B. Dydensborg, P. Bonnet (2003). "Bluetooth and sensor networks: a reality check". Proceedings of the 1st international conference on Embedded networked sensor systems: 103-113.  "They are based on the Atmel ATmega128L microcontroller - an 8 bit microcontroller (MCU) clocked at 7.4 MHz, with 4 KiB on chip memory and an external memory chip of up to 64 KiB."

M. Leopold (2003). "Tiny Bluetooth stack for TinyOS". . Department of Computer Science, University of Copenhagen "It has 4 KiB on chip memory and an external memory chip of 512 Kibit (64 KiB). ... (Lower case “k” refers to 1000, “Ki” refers to 1024, “b” to bits, “B” to bytes.)"

Menon, C.J. (2005). "A single network solution for safety-related applications using CANopen" in Industrial Electronics Society, 2005. IECON 2005. 31st Annual Conference of IEEE. Industrial Electronics Society, 2005. IECON 2005. 31st Annual Conference of IEEE: 6 pp..  "The CSC01 is clocked at 16 MHz. It provides 10 KiB of SRAM and 256 KiB of Flash memory."

Worringen, J.; A. Gaer, F. Reker (2002). "Exploiting transparent remote memory access for non-contiguous- and one-sided-communication" in Parallel and Distributed Processing Symposium., Proceedings International, IPDPS 2002, Abstracts and CD-ROM. Parallel and Distributed Processing Symposium., Proceedings International, IPDPS 2002, Abstracts and CD-ROM: 163-172.  "The Cray T3E reaches an efficiency of about 1 for blocksizes between 8 and 32 kiB, but has a very low efficiency for very small (< 4 kiB) and big (> 32 kiB) blocksizes."

Heirman, W.; J. Dambre, J. Van Campenhout, C. Debaes, H. Thienpont (2005). "Traffic Temporal Analysis for Reconfigurable Interconnects in Shared-Memory Systems" in Parallel and Distributed Processing Symposium, 2005. Proceedings. 19th IEEE International. Parallel and Distributed Processing Symposium, 2005. Proceedings. 19th IEEE International: 150a.  "Since this influences the working set, and thus the cache hit rate, the level 2 cache was resized from an actual 8 MiB to 512 KiB, resulting in a realistic 80 % hit rate. The simulation slowdown (simulated time versus simulation time) was a factor of 50,000 resulting in execution times of roughly 2 hours per benchmark on a Pentium 4 running at 2.6 GHz with 2 GiB RAM."

G. Gerhardsson. "Cacheprobe: An open source multiplatform library implementation for extraction of cache memory parameters". . Master’s Thesis, Department of Computing Science, Umea University, 2004 "For example, lets take a cache of size 1 MiB. If the stride is 0.5 MiB, or a multiple thereof, the memory positions will compete for the same cache positions."

Dictionary and Encyclopedia

byte . (2007). In Encyclopædia Britannica. Retrieved 28, 2007, from Encyclopædia Britannica Online Library Edition: 6:18 PM 6/28/2007

"Because a byte contains so little information, the processing and storage capacities of computer hardware are usually given in kilobytes (1,024 bytes) or megabytes (1,048,576 bytes). Still larger capacities are expressed in gigabytes (about one billion bytes) and terabytes (one trillion bytes)."


"megabyte n." The Concise Oxford English Dictionary, Eleventh edition revised . Ed. Catherine Soanes and Angus Stevenson. Oxford University Press, 2006. Oxford Reference Online. Oxford University Press. 29 June 2007

"megabyte ( abbrev.: Mb or MB ) n. (Computing) a unit of information equal to one million or (strictly) 1,048,576 bytes."

Computer Component

The research company, iSuppli Corp, measures various segments of the electronics industry and publishes the rankings.[5] They are frequently quoted in the technical and mainstream press.

Top ten Chip Suppliers, December 2006 [6]

Intel, Samsung Electronics, Texas Instruments, Toshiba, STMicroelectronics, Renesas Technology, Advanced Micro Devices, Hynix, NXP, Freescale Semiconductor

DRAM suppliers, April 2007 [7]

Samsung , Hynix, Qimonda, Elpida, Micron, Powerchip, Nanya, ProMos, Etro, Winbond (All use MB and Mb)

Top-10 Third Party DRAM Module Supplier, October 2006 [8]

Kingston Technology, Smart Modular Technology, A-Data, Crucial Technology, TwinMOS, MA Lab, Ramaxel Technology, Corsair Memory Apacer Technology, Trandscend (All use MB)

Six major HDD suppliers, June 2007 [9]

Seagate Technology, Western Digital Corp., Hitachi Global Storage Technologies (GST), Samsung Electronics Co. Ltd., Fujitsu Ltd., Toshiba Corp.
FWIW, my recently purchased Samsung HDD comes with an installation guide that uses GB throughout to mean 109, but in one place it does use GiB: "FAQ 3: Although my drive is larger than 137GB, it is only shown as 137GB. — There are two reasons of the 137GB (128GiB) capacity barrier...". The rest of that section then keeps using 137GB. (And yes, the English is very ropey). --KJBracey 18:34, 7 August 2007 (UTC)

Computer Equipment Companies

Forbes Technology Hardware Equipment Ranking [10]

IBM, Hewlett-Packard, Cisco, Dell, Apple, Toshiba

Software Companies

Forbes has a list of top software companies but over half do consulting or other behind the scene work. [11] Here are the companies that sell software applications.

Microsoft, Oracle, SAP, Symantec, CA, Adobe Systems, Intuit, Autodesk,

Here are some other significant software companies from the Linux, FSF world.

Red Hat, Novell, Mozilla, Apache

Retailers

InternetRetailer.com published a survey to the top e-retail businesses. [12]

Amazon, Staples, Office Depot, Dell, HP Home & Home Office Store, OfficeMax, Sears Holding Corp, CDW, SonyStyle, Newegg were the top ten. Most sell computer equipment. There could be a better selection but this is a workable list.

None of these retailers use the IEC binary prefixes.

-- SWTPC6800 22:03, 1 July 2007 (UTC)

How many use SI prefixes? — Omegatron 06:10, 4 July 2007 (UTC)
The point being all of them use KB/MB/GB kilobyte/megabyte/gigabyte etc. Fnagaton 09:28, 4 July 2007 (UTC)
Perhaps it is more accurate to say all of them use K/M/G and/or KB/MB/GB and/or kilobyte/megabyte/gigabyte, etc., in both binary and decimal senses, frequently in the same advertisement but none of them use IEC binary prefixes at all. Tom94022 23:48, 4 July 2007 (UTC)
Exactly. — Omegatron 12:58, 14 July 2007 (UTC)