Slashdot Mirror
China Makes World's Fastest Supercomputer
shmG writes "China has replaced the United States as the maker of the world's fastest supercomputer. A Chinese research center has made the world's faster super computer — named as Tianhe-1A, which was released at a national conference on high-performance computers (HPC) in China. Made at a cost of over $88 million, Tianhe-1A is theoretically able to do more than 1 quadrillion calculations per second (one petaflop) at peak speed. Tianhe-1A 's peak performance reaches 1.206 petaflops, and it runs at 563.1 teraflops (1,000 teraflops is equal to one petaflop) on the Linpack benchmark."
But does it run (Red Flag) Linux?
And the fastest social and economic downturn is in America...coincidence?
He who knows best knows how little he knows. - Thomas Jefferson
Yeah, lets borrow more money from the Chinese Government so we can build a useless supercomputer to outdo them - just to say we did it! Thanks, grandkids!
Stolen? I don't know. Purchased? From the article:
So unless Nvidia and Intel have reported 20,000 or so stolen processors lately, I wouldn't worry too much.
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
Sadly, China needed to build this computer simply to calculate the interest on the US debt in realtime.
Serious research still needs much faster supercomputers than we have now. All kinds of science - from artificial intelligence to weather modeling to astrophysics to genetic research to nuclear simulations. Access to a powerful supercomputer is a major boon for academia in the country.
As someone whose works depends on HPC I disagree with you. A lot of people in life sciences, materials science, nuclear physics, geophysics and other knowledge areas needs clusters and super computers.
English is not my first language. Corrections and suggestions are welcome.
What are you suggesting?
Growing population by a factor of 5?
Decrease salaries?
Spend lots of the money on small high image projects while most of the rest of the country remain poor?
I do understand that they will eventually catch up, but in the mean time you Americans are way ahead of the average Chinese.
Oak Ridge (Jaguar):
Cores Rmax(GFlops) Rpeak(GFlops) Nmax Nhalf
224162 1759000 2331000 5474272 0
Seems faster by a good margin.
we went to the moon during a period of nationalist chest thumping, and when the nationalist chest thumping subsided, we haven't been back. countries that are interested in nationalist chest thumping: china, india, etc, are still pumping up their space programs
what i am saying is, for all the evils of nationalism, scientific advancement in the realm of large projects seems to be a positive byproduct
for example, if we were still in a cold war with the ussr in the 1990s, i will bet you anything that this would have been completed and would be producing amazing science at this point in time:
http://en.wikipedia.org/wiki/Superconducting_Super_Collider
intellectual property law is philosophically incoherent. it is your moral duty to ignore it or sabotage it
Hint: cynics don't require facts to be cynical.
Remember to maintain your supply of
The slashdot summary has the wrong numbers. The actual article which slashdot quotes is contradictory. Its starts by saying:
"Tianhe-1A has set a new performance record of 2.507 petaflops, as measured by the Linpack benchmark, making it the fastest system in China and in the world today."
and then one paragraph later it gives the same numbers as the slashdot summary.
Other articles (from other sites) are claiming theoretical peak performance of 4 Petaflops (from an Nvidia source) and sustained petaflops of 2.5.
As a researcher using (US&European) supercomputers for high-performance numbercrunching, I strongly disagree. Some of the calculations take so much time that if I had to run it on one single (powerfull) CPU, I could probably do one big calculation in my entire PhD... Nevermind that by the time it was finished, the hardware would be long obsolete, and its not possible to fit the data in a single computer's memory anyway. So for some tasks, large/powerfull & shared computer resources really makes sense.
But of course, computers, electronic calculators, abacus'es, and pen&paper isn't strictly neccessary. Get off my stone-garden so I can carve my calculations on it!
Competitive in what, exactly? We have many supercomputers in the USA; we have no idea what to do with them, though, and many of them are spending a lot of time idle. Some supercomputers are now being rented out to investors, because the people the computers were built for -- the scientists -- are not using enough computer time.
What we really need to do is look at the state of research in this country. Also, maybe if we had a more solid economic base, one in which we solve the trade imbalance by exporting real goods rather than copyrights, we could spend more money on science and supercomputing. Oh well, in your words, "who am I kidding?"
Palm trees and 8
There is absolutely no consistency in numbers in this story. Some measurements show this computer to be about 45% slower than the Cray XT5 and some show it to be faster. Given China's history of arbitrarily throwing out numbers to try to prop themselves up in the international community I cannot accept this as fact without some sort of independent verification. If China has in fact created the worlds fastest supercomputer, I congratulate them on a job well done. But I am still skeptical about this story. Sounds like my government (US) is just looking for an excuse to spend billions more on a new supercomputer.
The article starts by claiming 2.507 petaflops, but gives no mention if that is Rmax or Rpeak. We have to assume that it is Rmax, since 2.5 petaflops is no big deal in terms of Rpeak.
Unfortunately, then the article lists both Rpeak and Rmax. But the numbers quoted seem to be for Tianhe-I (#7 on the top 500 list), not Tianhe-IA (not currently listed). Wikipedia table of the top 10
Oh, and it gets better. The article claims that Tianhe-IA has 7,168 GPUs and 14,336 CPUs. Very strange, since the Tianhe-I has 71,680 CPU/GPU pairs.
My guess is that China doubled up their Tianhe-I computer and swapped out for newer GPUs, then named the new thing Tianhe-IA (this is pretty normal when competing for top500 spots). I'm going to go with 143,360 Xeon/M2050 pairs. Either that, or the Chinese found a way to overclock 10% of their chips into the 20+ GHz range and threw out the rest.
See that "Preview" button?
INTEL, IBM, and other high tech firms have been sending their R&D, engineering and other high up on the job food chain jobs over there and to India. They have been building up expertise in other countries. Of course this happened.
We the US will become a technological backwater. Of course the pundits will say shit like "American kids just aren't studying science and engineering" or "It's our education system."
The answer is: why should a bright kid go into science or engineering when he won't be able to get a job? Whereas, if he goes into medical, he's pretty much guaranteed a very nice living.
It's not the education system; it's the market. The market here in the US is saying that engineering and science careers just aren't worth as much as others and it's saying that there are plenty of qualified and cheaper engineers overseas - all thanks to US companies moving there.
As we are seeing NOW, the Chinese and Indians no longer need American companies - they don't need IBM or whoever to come in a spend the millions setting up shop. They can do that themselves now thank you very much. End result: US based companies will be sidelined.
So kids, apply to foreign firms because US based companies have made themselves irrelevant.
And business owners, bypass the middlemen (IBM and whatnot) and buy direct from their suppliers in India and China - you'll save the costs of over paid American management and sales people.
RIP America
July 4, 1776 - September 11, 2001
Does research still need supercomputers, though? If you're writing a program parallelized enough to split across tens of thousands of processing units, why not go with a full cluster, like the vaunted Hadoop or EC2?
I think the time for a single powerful machine is long past. Maintaining the level of interconnection between the nodes is expensive, and we can do better. In the words of Dr. Ken Batcher, "A supercomputer is a device for turning compute-bound problems into I/O-bound problems." With distributed storage (like HDFS) coupled to the distributed processing (like Hadoop), we can turn those same problems into merely cost-bound problems.
You do not have a moral or legal right to do absolutely anything you want.
Supercomputing is largely a solved problem...do a regression analysis on the variables on the Top 500 sometime. Primarily it's a function of the number of cores. I've told people in my own workplace - you want a machine on the Top 500 then write me a cheque - making a supercomputer isn't the feat of skill or engineering that it was in the days of Cray. This doesn't even touch on what the hell you use these things for the problem space for parallel processing is clearly smaller than that of serial processing. Add to that the assertion that the number of useful applications drops off steeply as you add more cores (at some point you are left with only the "embarrassingly parallel" ones) and creating the largest supercomputer in the world is akin to saying you are creating the least useful computer in the world. Not to mention probably one of the least power efficient, highest maintenance costs, etc..
Oh, like this doesn't go both ways. At least they didn't start bribing/kidnapping our scientists to make "Dooms Day" devices in a hidden lair in the desert, like we did with the Germans in the Manhattan project.
Also that Washington Times article is way out of date. The proceedings of the Wen Hoo Lee trial (accused of stealing the W88 warhead for China) embarrassed public officials to the point where Bill Clinton had to issue an apology and the US government had to pay an undisclosed settlement to Lee.
The fact was US military research bases had slack handling protocols for classified documents. This was glaringly obvious as the cause of all the leaked secrets. However, to save their careers, the managers made Lee a scapegoat. Turns out it was common practice to take classified documents to work on at home. The only reason Lee was singled out for it was, because he was Asian.
Nearly all of these supercomputers are just that - VERY large clusters.
Although in many cases they have specialized communications backplanes for communications between nodes with capabilities (such as low latency) that can't be achieved with geographically distributed clusters. (Note the mention of parts from Intel and Nvidia, combined with undefined "domestic" communications silicon.)
Also note that geographic distribution leads to all sorts of information assurance nightmares when you're simulating nukes...
retrorocket.o not found, launch anyway?
The only reason it's so fast is because a half hour after you feed it data, it's hungry again...
There exists no way of exchanging information without making judgments. --Bene Gesserit Axiom
There is nothing useless about a supercomputer. Oak Ridge National Lab has over a billion dollar budget each year and huge portions of that budget relies on the availability of high performance computing resources. (Not to mention all the other national labs) HPC supports research in areas like energy conservation, new power sources, bioinformatics, material science, weapons simulations, engineering, and computer science. Applications range from freeing ourselves of fossil fuel reliance to designing materials to be used in [insert next big product]. HPC is the reason we don't need to do nuclear weapon testing anymore. HPC is the reason our grandkids will have a longer average lifespan. I can guarantee that these machines wouldn't be built for tens or hundreds of millions of dollars if they weren't being used. And I can guarantee that when the US regains #1, it won't be for the sake of being #1... it will be for the necessity of furthering science that benefits us all.
China: "We made the fastest super-computer!!!"
Intel and NVidia: "Uh - no you didn't, We are own all your processors!!!"
Is this faster than Blue Waters at NCSA is going to be in 2011?
Nope, Blue Waters is supposed to be significantly faster. According to NCSA's page about Blue Waters, Blue Waters is supposed to have peak performance of 10 petaflops, and sustained performance at 1 petaflop. Tianhe-1A, according to the summary, peaks at about 1.2 petaflops.
all of european advancement over the last 10 centuries can be traced to tribal and then nationalist competition
french spanish and british frigates would not have been sailing around india, china and the south pacific, making military inroads, if french spanish and british galleons were not first doing their best to better shoot holes into one another
in fact, you can say china and india stagnated behind european scientific advancement precisely because there was no fever pitch nationalist rivalries in those areas
european history is exhibit number one of scientific advancement propelled forward by nationalist rivalry, hardly an example of a contrast to what i am saying
intellectual property law is philosophically incoherent. it is your moral duty to ignore it or sabotage it
China: "We made the fastest super-computer!!!"
Intel and NVidia: "Uh - no you didn't, We are own all your processors!!!"
The magic behind supercomputing isn't CPUs.
The real trick has always been the interconnects & the software that gets those thousands of C/GPUs talking to each other.
And don't ever forget that China is developing its own CPU.
It still sucks right now, but like with rare earth minerals, China is in this for the long haul.
[Fuck Beta]
o0t!
Yes, this is spot on for massively parallel systems. The interesting thing is that China does actually make their own interconnects, but they aren't so great. The Tianhe-1 actually runs at 47% of the theoretical capacity. In contrast, the previous number 1 (Jaguar) runs at about 76%. In fact, China's previous big HPC was Nebulae, which had a higher theoretical peak than Jaguar, but didn't actually perform faster because of interconnects problem.
Gentlemen! You can't fight in here, this is the war room!