Talk:TOP500

From Wikipedia, the free encyclopedia

Contents

[edit] Introduction

The discussion of the list of the world's most powerful computing sites as compared to the Top 500 list is POV and lacks cites for its claims and should be altered to NPOV or cut. Tdewey 06:29, 30 October 2006 (UTC)

It is not a discussion, it is fact, since i was the author of the list, i am stating the facts first hand. Gunter 11:23, 30 October 2006 (UTC)

Provide a cite, even if it is to your own blog or drop it. Tdewey 02:26, 31 October 2006 (UTC)

[edit] POWER to Power

I will change the references to POWER to Power since this term encompasses both POWER and PowerPC, and Cell (when Cell based supercomputers show up in this list). IBM POWER is "just" POWER3/4/5. PowerPC 970 and the BlueGene/L-processors ain't POWER, but all three are Power Architecture. -- Henriok 20:17, 3 August 2006 (UTC)

I think IBM POWER-based supercomputers should stay linekd to IBM POWER, and Cell-based supercomputers should just link to Cell. Why generalize at the cost of usefulness, especially since Cell makes a huge difference in the field? -- intgr 20:33, 3 August 2006 (UTC)
I think it would become to diverse. POWER5, PowerPC 970 and BG/L belong to the same family, the Power Architecture.. not POWER. It certainly was wrong before. BG/L is not POWER, it's PowerPC. I think the usefulness is enhanced by the generalization. TOP500.org itself makes this caregorization (link). Perhaps we could do a combination, write the exact kind of processor, and the corresponding family (Power, Itanium, Xeon, Opteron, SPARC and so forth)? -- Henriok 21:20, 3 August 2006 (UTC)
Well, if it was incorrect before and if the official TOP500 list classifies them all into the Power architecture then I guess this is the right way to go. If my two cents is worth anything, though, I would certainly like to see Cell-based supercomputers clearly distinguished from usual Power-based ones. -- intgr 21:49, 3 August 2006 (UTC)
Yeah, I can live with that. Cell is special. I'm really looking forward to seeing them on the list. -- Henriok 22:34, 3 August 2006 (UTC)
This article states: "For the first time in the history of the TOP500 project (since 1993), the top Japanese system is not manufactured in Japan itself."

This sentence really ought to be referenced. I'd be curious to know where the manufacturing is done, if not in Japan (which is by far the world's leader in high-tech manufacturing). Just because the brand name of a product might be American doesn't mean that the product is manufactured in the U.S. these days. A lot of high-tech products' key enabling components are made in Japan these days, such as the Apple iPod. Other products, such as Boeing's upcoming Dreamliner are actually more of a Japanese product than American these days (some 70 percent of the Dreamliner will be made in Japan, including the ultra-sophisticated wings).

[edit] Removing the "more sources" tag

I am removing the "more sources" tag. We only need one source: the external link to the "top 500" site itself. This site is universally cited in the supercomputer community, and has been maintained by two separate major universities in two countries for more than a decade. If that is not good enough, then please place a comment here to explain what would be good enough. I do not wish to appear contentious. Instead, I truly want an explanation of what we need to cite to ensure that any ressonable editor would accept as a reference. I realize that not all editors are familiar with the supercomputer community (loosely defined) and may not realize that TOP500 is now the defacto definition for "supercomputer." Thanks. -Arch dude 01:26, 9 December 2006 (UTC)

I agree with you 100%. Nice initiative. -- Henriok 21:44, 22 December 2006 (UTC)

[edit] Flops/s?

Since Flops are (correct me if I'm wrong!) "FLoating point Operations Per Second", isn't it redundantly redundant to say TFlops/s? Middlenamefrank 21:31, 22 January 2007 (UTC)

[edit] 500?

 Please help improve this article or section by expanding it.
Further information might be found on the talk page or at requests for expansion. (June 2007)

if this is about top 500 supercomputers, where's the list of the other 490 ones? Jernejl 12:03, 26 March 2007 (UTC)

The top500.org website, was, about 10-15years ago a good site. there actually were 500 machines listed. web archive has some lists...

http://web.archive.org/web/20001202071700/www.top500.org/lists/TOP500List.php3?Y=1999&M=11

no idea if a list of 500 is possible now, it is a terrible site to navigate these days. 88.96.235.230

[edit] Dire need for Criticisms section

In my opinion, this article is in dire need of a Criticisms section. In fact, someone ought to write an article titled "How the Top 500 List is Destroying Supercomputing". There is less and less relationship between Linpack performance and achievement of actual science. Many supercomputers are bought and configured simply to get a high Linpack number, tremendously wasting the taxpayer money spent on them. This problem has risen to scandalous proportions. Westwind273 17:08, 22 May 2007 (UTC)

A plug for MDGRAPE-3  ? Find someone notable who says what you say because I question your claim. Given HPL benchmarks floating point operations which are kind of useful across a wide variety of tasks, any effort to get high floating-point results is a good thing that is broadly appealing. Much better to spend taxpayers money on complex simulations of say bombs than trying destructive testing of the same. Ttiotsw 07:53, 9 June 2007 (UTC)

See the article http://news.zdnet.com/2100-9584_22-5238405.html Also, NPOV warnings are meant to placed on the article itself, not on the talk page. I have removed the one you placed on this section. Westwind273 18:04, 27 June 2007 (UTC)

Umm...I have not placed any tags here. See here. On the other hand the news article you link to is dated from 2004 and Seymour Cray (if still alive) would say why the NEC got the top prize - because they were using vector processing whilst the US Gov. focused on scalar processing. It even says ... "and for another, a system that can't get a high Linpack won't do well on other tests,". So basically there is absolutely NO information content in that link: it is a 3 year old ZDNet 'troll' article. Ttiotsw 21:41, 16 July 2007 (UTC)

If you would like a more recent article, read the first three paragraphs of http://www.washingtontechnology.com/print/22_16/31358-1.html People that need to get real science done with their supercomputers don't evaluate them with Linpack. Westwind273 21:18, 4 September 2007 (UTC)

Check out the article at http://www.hpcwire.com/topic/systems/HPC_Innovation_in_the_Era_of_Good_Enough.html "In practice, the success of an installation is rarely measured by the performance of applications, but by the Linpack performance of the solution. Customers are often eager to be "on the list" or, lacking better tools and a vocabulary to talk about the effectiveness of a deployment in any other terms, simply focus on the one measure they can specify. Vendors who don't respond to this customer focus lose procurements, so prices get pushed down and designs are increasingly based on commodity technologies." --Westwind273 (talk) 23:36, 28 April 2008 (UTC)

[edit] Terminology

"Rmax" and "Rpeak" are used as table column headers, but are not explained. -- Beland 05:23, 4 June 2007 (UTC)

[edit] November 2007 list

When will be updating the Nov 2007 list. Since there would be a large number of hits for this page, now that the latest list has been released.

http://www.top500.org/blog/2007/11/09/30th_edition_top500_list_world_s_fastest_supercomputers_released_big_turnover_among_top_10_systems


Kunalpathak13 09:04, 14 November 2007 (UTC)kunalpathak13

[edit] June 2007 list

When will the June 2007 list be released?

During ISC07 which is this week.Ttiotsw 10:53, 25 June 2007 (UTC)

[edit] New Number One at Utexas

It is called Ranger with it's 504 teraflops.[1]

They only rank the Top500 list twice a year. There probably are several systems that will battle for the #1 spot but are yet to be announced. -- Henriok (talk) 21:39, 21 February 2008 (UTC)

[edit] Temporary Page to Edit

Given that most of this page is very specific to each year i have decided to create a new page at TOP500/2007 to make edits to this years list. I think this will be best until someone can rewrite the speicifc sections to reperesent this year instead of giving two halfs of different stories. -- Jimmi Hugh 15:17, 27 June 2007 (UTC)

TOP500/2007 Now contains the updated list of Top 10 Super Computers. Before Being copied here though the information around it needs to be changed in order to prevent misleading information. I won't move it now because i am unsure how long it will be sat here before someone else updates it. Please check that page before rushing ahead with edits here. -- Jimmi Hugh 15:48, 27 June 2007 (UTC)
TOP500/2007 has Highlights from the Top 10 and General highlights from the Top 500 since the last edition sections updated from official TOP500 site by copying all text on source site. Last Wikipedia version was a subset of source. Please edit as appropriate. ConradPino 16:24, 27 June 2007 (UTC)

Backed myself into a bit of a hole here. 2007 page seems pretty much about only 2007 list, if in serious need of some rewrites. I don't know how to copy it without losing Contribution history though. -- Jimmi Hugh 20:44, 27 June 2007 (UTC)

I see only 2 contributors so far, you and I. I'm not concerned with Contribution history loss. Conrad T. Pino 08:15, 28 June 2007 (UTC)

[edit] June '07 list

It has been released.