Talk:Tianhe-I/Archive 1

Latest comment: 9 years ago by 2602:252:D10:2340:EC7A:2700:1ADF:842 in topic Criticism
Archive 1

Cost?

How much did this thing cost to build, and was it funded solely by the government or do some corporations have an interest in it as well? Ham Pastrami (talk) 04:44, 29 October 2010 (UTC)

Who did it replace as the world's fastest supercomputer?

The article doesn't say which computer it replaced as the world's fastest supercomputer. Does anyone know? Could it be added in the article or in a succession box below, like we have on, for example, the worlds tallest building? Cheers, Calistemon (talk) 23:23, 29 October 2010 (UTC)

I became proactive for once, researched my own question and added the information plus a succession box to the article. Calistemon (talk) 00:00, 30 October 2010 (UTC)

No information on the interconnect

The interconnect, the infrastructure that connects all the processors together is a proprietary one of Chinese design. The interconnect has approximately twice the bandwidth of InfiniBand, a US interconnect used in HPC.[1][2] There's absolutely no mention of this in the Wikipedia article, I think there should be, to show that it isn't all off the shelf parts. 87.113.181.217 (talk) 09:47, 30 October 2010 (UTC)

Then add it yourself. It is clearly important. There is a lot that is missing...I think we should split Tianhe-1A into a separate article since is it a completely different supercomputer. 76.69.62.155 (talk) 18:05, 30 October 2010 (UTC)

OS?

I'm just curious to know what operating system is installed on Tianhe-I, if anybody knows, that would be lovely to be added to the article, 谢谢!Comment added by User:141.24.172.1

Perhaps a variant of Linux developed by PRC to suit their needs.[3][4] Please don't forget to sign your comments with four ~~~~ :) --Zvn (talk) 20:04, 16 November 2009 (UTC)

windows 95 —Preceding unsigned comment added by 58.7.221.189 (talk) 04:23, 31 October 2010 (UTC)

How many desk-tops or home computers would be needed to perform the same tasks as this computer ?

A simple way to communicate the power of this computer might be to say how much more powerful it is than an office computer. Any ideas ?89.194.193.190 (talk) 23:25, 29 October 2010 (UTC)

That wouldn't provide a definite value at all - what defines a "home computer"? Such an example is vague because home computers themselves vary. My custom-built gaming PC would be more powerful than that of an employee working at a software programming firm, an office PC designed for multimedia development would be more powerful than a receptionist's, and the home PCs belonging to the elderly would be less reliable than those used at Blizzcon 2010. -- 李博杰  | Talk contribs email 09:27, 31 October 2010 (UTC)

Peak Speed

It lists that 2.5 petaflops is its peak performance. From what I understand this is its nominal performance, its peak is much higher. —Preceding unsigned comment added by 99.30.226.124 (talk) 02:11, 30 October 2010 (UTC)

But can it run Crysis?! —Preceding unsigned comment added by 58.7.221.189 (talk) 04:25, 31 October 2010 (UTC)

It runs more than just Crysis. It clocks more than the Jaguar (computer) which is used to simulate weather patterns, while most of you casuals can't even run Crysis on 1680x1050 with high lighting filters. inb4 "Linux has no gaems". </tongue-in-cheek> -- 李博杰  | Talk contribs email 09:31, 31 October 2010 (UTC)

Title

Should title of the article be renamed "Tianhe-1" instead of "Tianhe-I"?---Now wiki (talk) 02:30, 31 October 2010 (UTC)

probably, where did Tianhe-I even come from? —Preceding unsigned comment added by 76.69.62.155 (talk) 03:11, 31 October 2010 (UTC)
Roman numerals, e.g. I, II, III, IV, V, VI, VII... It's not rare to give supercomputers names such as SomeCompany Mark IV or Harvard Mark I. -- 李博杰  | Talk contribs email 09:32, 31 October 2010 (UTC)

Large Font?

Anyone else getting very large font on this specific article and not others? Tried it with multiple browsers and it seems to persist.

207.112.39.86 (talk) 15:59, 31 October 2010 (UTC)

I decided to fix this, so fixed. - 93.97.255.48 (talk) 16:32, 31 October 2010 (UTC)

Number of Processors?

The quoted number of processors in the Tianhe-I paragraph seems incorrect:

"It is now equipped with 500,854 Xeon X5670 processors and 7,168 Nvidia Tesla M2050 general purpose GPUs"

500,000 Xeon CPUs? —Preceding unsigned comment added by 93.97.246.99 (talk) 20:02, 31 October 2010 (UTC)

Storage Space?

How many libraries of congress does it hold? —Preceding unsigned comment added by 96.28.89.27 (talk) 23:44, 29 October 2010 (UTC)

That would be trivial and irrelevant to the article. Supercomputers are for processing calculations, not for storing information. -- 李博杰  | Talk contribs email 09:39, 3 November 2010 (UTC)

Sky river = Milky Way

In the reference cited, Tianhe is given an alternative translation as "Milky Way". I recognize that there are other ways of rendering "Milky Way" in Chinese, not only "Sky river" and "Silver river", but also "Star river" 星河 xīng hé - see, for example, Oxford Chinese Dictionary, Oxford University Press, 2010 ISBN 978-0-19-920761-9. Do you have any reference which backs up your claim that Tianhe does not mean Milky Way? What could it possibly refer to other than the Milky Way? TomS TDotO (talk) 12:25, 3 November 2010 (UTC)

Another alternative names of milk way, may be poetry or obsolete, is Han/汉, and 清汉, 银汉, 天汉, 云汉 also. Note that Han means Chinese people as well.--刻意(Kèyì) 04:05, 18 November 2010 (UTC)

Thanks, and a suggestion: Add these names to the article List of names for the Milky Way. TomS TDotO (talk) 11:25, 18 November 2010 (UTC)

Quadrillions and quadrillions

I'm not going to revert the recent change from "thousand billion" to "million billion" (in the long scale). But it is confusing and we ought to be sure to get this right. When we say "quadrillion" (in the short scale), we mean "thousand trillion" (short scale) or "million billion" (short scale) or "thousand million million" or "thousand billion" (long scale). I think. I'm not so confident as to change this (again) without a bit of discussion. I thought that I was making it clearer, but I seem not to have accomplished the hoped-for clarification. We should be able to get everybody to agree on this, shouldn't we? TomS TDotO (talk) 18:23, 4 January 2011 (UTC)

Couldn't we write 2.5 1015 and then explain how many millions it is? 1exec1 (talk) 00:10, 8 January 2011 (UTC)
My personal opinion is no. But I'm not insisting. What I do insist is that we don't explain it wrong.
Why not explain? Because we can't go through Wikipedia and explain every time the words "billion", "trillion", ... are used. I think that our article as it is now written is adequate. "Thousand million million" should be enough.
But "million billion" is confusing, at best, and wrong (unless I've made a mistake, which is possible). TomS TDotO (talk) 13:09, 8 January 2011 (UTC)
Depends on which scale you're referring to 1exec1 (talk) 17:16, 8 January 2011 (UTC)
The figure we're talking about here is of the order of 1015, I think we all agree. A million is 106, so a thousand million million is 103+6+6. If a "billion" is a "million million", then 1015 is a "thousand million million", which is a "thousand billion". If a billion is a "thousand million", 109, on the same scale a "trillion" is a "million million", 106+6=1012; and on the same scale, a "quadrillion" is a "thousand trillion", 1012+3. I don't know of any scale on which a "million billion" is 1015.
In my own opinion, it would be confusing to get into this in an article on the Tianhe I. It is best discussed in an article on the different numbering systems. Let's just leave things as is.
However, if you disagree, if you have some numbering system in which a "million billion" is 1015, let's hear about it.
Or if you think that this is not confusing, or if you think that it's fine to go off on a discussion of numbering systems, I'm willing to listen. Reasonable people can disagree. I certainly know that I can be wrong. TomS TDotO (talk) 18:11, 8 January 2011 (UTC)
million=106 , billion=109, then "million billion" = 109+5 = 1015 so I don't see any problems here. Did I misunderstood you? 1exec1 (talk) 15:27, 9 January 2011 (UTC)
This is why I am so careful to hedge my statements with "if I am not mistaken". It happens so often. Thank you for the correction. (I won't be so snide as to comment on your 9+5 = 15.) My excuse is that explanation of "quadrillion" as "million billion" had the appearance of explaining the "American" quadrillion in "British" terms. Indeed, it arose as a change from the "thousand billion" which was such an explanation. So, what is the least confusing way of explaining "quadrillion"? I'm suggesting "thousand million million" (which "Americans" can group as "million" x "billion", and "British" can treat as "thousand" x "billion") TomS TDotO (talk) 15:53, 9 January 2011 (UTC)
Actually I also never want to think about these billions since they are so error prone. That's why I suggest using 10x. PFLOPS are for computer scientists. However there are a lot people with technical background who are accustomed to 10x, but don't know or remember what PFLOPS is and have the same attitude towards the billions. So the current version doesn't convey any information to them. Of course there are people who don't understand neither PFLOPS nor 10xbut they probably won't bother with the billions either. 1exec1 (talk) 22:14, 10 January 2011 (UTC)

Criticism

I have serious concerns about Wikipedia swallowing whole the "Top500" ranking of the world's fastest supercomputers. Linpack is a highly simplistic way of measuring how powerful a supercomputer is. The real measure of a supercomputer's power is how much breakthrough science it is achieving. The Chinese have succeeded in building the world's fastest Linpack machine, which anyone can do with enough money. Whether the Tianhe-I is capable of the most break-through science is a completely different matter. --Westwind273 (talk) 22:33, 15 November 2010 (UTC)

That is highly opinionated, and technically incorrect for a couple of reasons. The "value" of any computer is subjective. Define "break-through science", for starters, and name one supercomputer that can be precisely credited for such a thing. Linpack is an objective test, whether or not you agree with its metrics, and that's why it's used as a benchmark. If you can find an objective test for "value" benchmarks, that would be fine too, but chances are that their criteria would be even more debatable. The TOP500 does not claim to measure value; it only claims to measure speed, and that's what the article is about -- being the fastest, not the most valuable. If you read more into that than you should, that is your problem, not the article's. Ham Pastrami (talk) 04:20, 25 November 2010 (UTC)
How accurately Top500 measures speed is debatable. It measures FLOPS fine, but FLOPS is not considered the most important metric any more. Graph500 is an up-and-coming metric. Top500 is still the standard, though, so it is probably the best metric for not.2602:252:D10:2340:EC7A:2700:1ADF:842 (talk) 23:54, 12 August 2015 (UTC)