Talk:History of supercomputing

(untitled)
For the lay reader, it would be helpful to benchmark supercomputer performance to PC performance, e.g., compare the Cray 1 to PCs of, say, 2000, 2005, and 2010 as well as at the time the Cray was built.202.179.19.24 (talk) 07:29, 16 February 2013 (UTC)

CDC 1604 world's fastest computer claim
The article makes the claim that the CDC 1604, designed by Seymour Cray while at Control Data Corporation, was "the fastest computer in the world at a time when vacuum tubes were found in most large computers." It is true that the 1604 was one of the first fully-transistorized computers, but it was not the only one (the first IBM 7090 was delivered in November 1959). According to Gordon Bell at a talk he gave at the University of Minnesota in 1997, the CDC 1604 was "...somewhat slower than an IBM 7090 with its 2.2 microsecond memory" but made up for it in other ways with "...the longer word length [which] helped by avoiding double precision arithmetic and having two instructions per word.". Also note that the CDC 1604 article itself doesn't make the claim of it being the world's fastest computer (before 1964). Bumm13 (talk) 03:52, 16 January 2014 (UTC)


 * Yes, well it probably depends on how you measure it. According to that, the 1604 clock speed was slower, but it might have actually been faster.  The longer word length is a factor.  I remember benchmarking a CDC Cyber 70/74 versus and IBM 370/158.  The CDC was about 6x faster, using single precision on both.  But the CDC had a 60-bit word versus 32-bits for the IBM.  If you had a floating-point intensive task, and the IBM single-precision wasn't enough, the IBM double precision was about the same as the CDC single precision.  And when you compared IBM double precision to CDC single precision, the CDC was close to 20x as fast.  So the word length does matter. But the 1604 had 48-bit words compared to 36 for the 7090, so that is somewhat of a factor. (And the 1604 can get two instructions out of memory at the time.) Bubba73 You talkin' to me? 00:01, 24 February 2019 (UTC)


 * The Control Data 1604 had a 48-bit word length. That meant that its single-precision floating-point numbers would have had at least ten significant digits of precision, sufficient for most scientific computations. So, if one compared 48-bit floating-point operations on a 1604 to 72-bit double precision floating-point operations on a 7090, one probably could make the case for the 1604 being faster. Whether or not that is a fair comparison is another matter. 2001:56A:F384:4D00:5433:527E:AF83:CC7C (talk) 20:50, 12 October 2020 (UTC)

First use of the term "Super Computing"
The article says: The term "Super Computing" was first used in the New York World in 1929 to refer to large custom-built tabulators that IBM had made for Columbia University.

I believe that date should be 1931. I also believe that a 1930 use of the term "super-computing" by Silas Bent is earlier than the NYW use.

While references like http://books.google.com/books?id=wX_9EEEuPUsC&pg=PA169&lpg=PA169&dq=super+computing+%22new+york+world%22&source=bl&ots=4x-Sl3SnwD&sig=4ALrGob_NDNUikd51QWy36U8Qm4&hl=sv&sa=X&ei=TLTXUsORDoiEyQPXtIHoAQ&ved=0CHIQ6AEwCQ#v=onepage&q=super%20computing%20%22new%20york%20world%22&f=false give a date of 1920 (more specifically, of Mar. 1 1920), this appears to be erroneous.

There's a copy of the clipping at http://www.columbia.edu/cu/computinghistory/packard.html, which comes from Eames, Charles; Eames, Ray (1973). A Computer Perspective. Cambridge, Mass: Harvard University Press. p. 95. The 1973 reference asserts a 1920 origin, but the columbia.edu site points out that it "is obviously a mistake, since it refers to the Statistical Bureau in Hamilton Hall, which did not exist until 1929".

According to the same columbia.edu page, funding for the Columbia Difference Tabulator started in 1930, making a demonstration date of 1929 impossible.

This is likely why http://en.wikipedia.org/wiki/Tabulating_machine which attributes a 1931 date to that same clipping. (Though it doesn't have an actual date.)

In any case, my own investigations found a 1930 reference to the term "super-computing" in "Machine Made Man", by Silas Bent, 1930. See http://en.wikipedia.org/wiki/Talk:Tabulating_machine#First_use_of_the_term_.22Super_Computing.22 for details. This means it's unlikely that the New York World headline is the earliest known use of the term "super computing machines"/"super-computing machine". -- Andrew Dalke

Internally inconsistent article?
We have:


 * Around 1960 Cray decided to design a computer that would be the fastest in the world by a large margin. After four years of experimentation along with Jim :Thornton, and Dean Roush and about 30 other engineers Cray completed the CDC 6600 in 1964.

And:


 * In 1956, a team at Manchester University in the United Kingdom, began development of MUSE — a name derived from microsecond engine — with the aim of eventually :building a computer that could operate at processing speeds approaching one microsecond per instruction

...


 * At the end of 1958 Ferranti agreed to begin to collaborate with Manchester University on the project, and the computer was shortly afterwards renamed Atlas, :with the joint venture under the control of Tom Kilburn. The first Atlas was officially commissioned on 7 December 1962

So Atlas became operational before the CDC 6600, not "nearly three years after". HughesJohn (talk) 12:21, 29 September 2015 (UTC)

External links modified
Hello fellow Wikipedians,

I have just modified 3 external links on History of supercomputing. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive https://web.archive.org/web/20120728105352/http://www.computer50.org/kgill/atlas/atlas.html to http://www.computer50.org/kgill/atlas/atlas.html
 * Added archive https://web.archive.org/web/20110815102821/http://www.cc.gatech.edu/classes/AY2008/cs8803hpc_spring/papers/bgLtorusnetwork.pdf to http://www.cc.gatech.edu/classes/AY2008/cs8803hpc_spring/papers/bgLtorusnetwork.pdf
 * Added archive https://web.archive.org/web/20120330182549/http://sss.lanl.gov/pubs/031001-acmq.pdf to http://sss.lanl.gov/pubs/031001-acmq.pdf

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Cheers.— InternetArchiveBot  (Report bug) 05:53, 5 November 2017 (UTC)

ILLIAC IV?
Shouldn't the ILLIAC IV be in the article? It was said to have been the fastest computer at the time. Bubba73 You talkin' to me? 18:42, 20 February 2018 (UTC)


 * Yes, at least as the predecessor of the Connection Machine (SIMD architecture) Poil (talk) 01:26, 6 July 2024 (UTC)

ATLAS/Stretch discrepancy
So, it's not just this article. Check the leads on Manchester computers and IBM 7030 Stretch. Stretch claims to be the fastest from their release until 6600, Stretch came first in May 1961 and ATLAS in December 1962. The latter only makes the claim, as does this article, that ATLAS was fastest "at the time of its release." The Stretch article implies that it remained the fastest. So where are the comparisons? And I wonder if IBM updated its numbers after "customizing" the Stretch for the NSA. That would have been released around the same time as the Atlas - if I know IBM they declared it installed 12/31 for a contracted Jan 1 delivery. So it's possible the Atlas was only fastest for a moment.

An in-depth look at this discrepancy might be a great addition to the article. I've taken out the citation request, b/c there are two citations at the end of the sentence and the cites for the Atlas only verify the status in 1962 and don't talk about the situation in 1964, so there is no conflict. Here's where you add 'however, so and so said the Atlas was faster'. I'll add a clarify tag instead of a citation tag. Maybe somebody will read this. Skyerise (talk) 00:39, 12 January 2022 (UTC)

> 60 000 processing units is not a 21th century feature, no dramatic progress either
Progress in the first decade of the 21st century was dramatic and supercomputers with over 60,000 processors appeared, reaching petaFLOPS performance levels.

This needs to be corrected since the leader super computers of the ~1986-1990 years were the SIMD Connection Machines (Thinking Machines) with models CM-1, CM-2, CM-200 which had 64K processing units. Also I can't see anything which can be described as a dramatic progress since 2000 : https://ourworldindata.org/grapher/supercomputer-power-flops Poil (talk) 01:21, 6 July 2024 (UTC)