China's supercomputers race past US to world dominance
Source: CNET
China's supercomputers race past US to world dominance
China doesn't just have the single fastest supercomputer in the world. It now dominates the list of the 500 fastest.
BY STEPHEN SHANKLAND NOVEMBER 13, 2017 6:00 AM PST
For years, China has claimed the top spot on a list of the 500 fastest supercomputers. Now it dominates the overall list, too, pushing the United States into second place.
For the first time, China has the most systems on the Top500 list, 202, up from 159 six months ago. The US dropped from 169 to 144. And in terms of the total performance of those machines, China also overtook the US, the Top500 supercomputer list organizers said. ... The news underscores the relentless ascent of China's supercomputing trajectory in recent years. It also marks a notable shift in the international balance of high-end computing power that's closely tied to industrial, academic and military abilities.
....
Supercomputers, mammoth machines that can occupy entire buildings and use thousands of processors, are useful for tasks like simulating nuclear weapons explosions, forecasting weather, designing aircraft and investigating the cosmos by reconstructing billions of year of the universe's history. ... At the SC 17 supercomputing show starting Monday in Denver, NASA will show off supercomputing work into precise simulations of climate change on Earth, the aerodynamics of drones that fly using multiple propellers and detailed forecasts of shock-wave damage from meteors.
The Top500 list, released twice a year in conjunction with the annual SC conference, is compiled by researchers at Lawrence Berkeley National Laboratory, the University of Tennessee and cloud-computing company Prometeus. It ranks supercomputers by how fast they can perform mathematical calculations on an imperfect but still useful speed test called Linpack. Results are measured in floating-point operations per second, or FLOPS.
Read more: https://www.cnet.com/news/china-surpasses-us-in-supercomputer-usage-on-top-500-list/
Full disclosure: I know diddly squat about supercomputers. I hope someone at DU who does know something about them chimes in.
I don't know why, if this an annual survey, the news was released in June in 2016 and not until November in 2017.
ETA: it's not an annual survey. It's a semi-annual survey.
* * * * *
Previously at DU, like a year and a half ago:
China Tops Global Supercomputer Speed List for 7th year
https://www.democraticunderground.com/10027934258
http://www.sacbee.com/news/business/article84780327.html
BEIJING
A Chinese supercomputer has topped a list of the world's fastest computers for the seventh straight year and for the first time the winner uses only Chinese-designed processors instead of U.S. technology.
The announcement Monday is a new milestone for Chinese supercomputer development and a further erosion of past U.S. dominance of the field.
Last year's Chinese winner in the TOP500 ranking maintained by researchers in the United States and Germany slipped to No. 2, followed by a computer at the U.S. government's Oak Ridge National Laboratory in Tennessee.
Also this year, China displaced the United States for the first time as the country with the most supercomputers in the top 500. China had 167 systems and the United States had 165. Japan was a distant No. 3 with 29 systems.
...
Irish_Dem
(47,440 posts)mahatmakanejeeves
(57,618 posts)Last year's Chinese supercomputers used Chinese processors.
Irish_Dem
(47,440 posts)It will achieve its goal and soon.
LanternWaste
(37,748 posts)China is already an existing superpower as it is able to exert a global influence via economic, military, technological and cultural strength, as well as its diplomatic and soft power influence.
Irish_Dem
(47,440 posts)dlk
(11,578 posts)Priorities...
Achilleaze
(15,543 posts)and other people, the USA would have a chance? Too bad the republicans are shutting down learning and facts and science and so forth. Really way too freaking bad.
lapfog_1
(29,226 posts)I've spent my entire career (minus a few years doing a couple of small startups) in the field of Super Computers starting with serial number 7 of the Cray-1 back in 1980. For a decade I was Chief Scientist at the NASA facility mentioned in the article (specifically the NAS program located at NASA Ames Research Center in Mountain View). For the last 7 years I was Chief Architect at a large corporation where I designed high-performance storage systems for Supercomputers.
What do you want me to say?
It's been well known that China was investing more into their supercomputers than we have... and that most of theirs disappear into their defense industry whereas only around 60% of ours go into classified projects, most of which are in the DOE (Dept of Energy).
Why DOE? Because DOE is where we design nuke weapons. And test the aging stockpile by simulating the aging warheads and how they should perform if needed (instead of underground testing).
Why are we falling behind? Because we haven't decided to invest ( a very high percentage of Supercomputers are bought by the federal government ). I doubt that the current administration is going to change this... in fact, I highly doubt that until very recently, the current administration even knew what the Dept of Energy actually did... I bet they thought it was related to oil and gas and renewables (it does some things with that stuff too, but not very much).
Recently I have been shopping around a business plan to raise private funds to build a next-generation climate modeling supercomputer ($100M+) not to figure out if we have climate change... but rather to determine the possible outcomes of geoengineering projects to reduce climate change impacts. Why? Because it is my firm belief we cannot stop climate change today by simply halting the use of fossil fuels. That means that IF we want the human race to continue at something even approaching the current population levels we will need to control the global climate. Before we do that, we should be able to predict if the multi-trillion dollar investments in such geoengineering projects would be effective and what the unintended consequences might be.
It is time to quit treating our one and only planet like a test tube.
My design for this climate control modeling supercomputer would not be the "fastest in the world" based on current benchmarks that determine such things, but rather it would build a system that is far more effective at certain simulations. Current systems (like the Tihane-2 in China) are very good at peak FLOPS (floating point operations per second), but the effective usage is much more like 3 or 4 PERCENT of that top peak FLOP rate. My system (co-designed by someone at NCAR) is more like 12 percent effective rate (which is very good).
Getting the government to fund such a system is simply not going to happen.
I was hoping to attract the attention of liberal or liberal-leaning billionaires like Tom Steyer, Paul Allen, Elon Musk, and others to fund this system (and a new facility powered by solar and cooled by ground or ocean water). So far, no takers. I can't even get a meeting with people that can take this project to such people.
I don't think we have the time to wait out a change in US government (back to sanity) to get such a program off the drawing board. That means that individuals are going to have to pony up the money.
And, yes, such a system could be used to further cancer research or other items if Climate Change isn't a big enough reason to fund this.
Blue_Tires
(55,445 posts)and that quantum computer?
lapfog_1
(29,226 posts)NASA bought a quantum computer after I left the agency... but I have heard reports about it.
A supercomputer is more or less a collection of regular computers (cloud servers, laptops, desksides). Most of them these days have specialized hardware to help them do basic math functions like FFTs... mostly these are GPUs from companies like Nvidia (thanks to all you gamers out there and your need for realistic animated action!). The difference between a "supercomputer" and, say, a collection of gaming computers on the internet is simply one of the speeds of communication between the processors.
They pack these processors (and their associated GPUs) into tight spaces and use very high-speed networks between them so that they can share the data that they either are getting from the outside world or the results of a specific calculation. These high-speed networks cost hundreds to thousands of dollars per processor (or node which typically has a collection of processors and their GPUs). They measure speed in terms of Gigabits/sec and the overall speed of the system in Terabits (or even Petabits) per second. They pack these together into a single room to reduce the speed of light issue (actually speed of signals down a copper wire which is a little be slower than the speed of light). That creates another problem... heat... it was often said that the father of supercomputing (Seymour Cray) was not so much a brilliant chip designer or even computer engineer... but he was the past master of packaging and cooling.
A computer pioneer named Grace Murray Hopper used to give out a little length of wire (about 11 centimeters long) that she called a "nanosecond" (the amount of time it takes to propagate a signal down that wire). I met Grace Murray Hopper (Rear Admiral) and I have a nanosecond from her in my collection of things not to lose.
A quantum computer is something else entirely.
https://en.wikipedia.org/wiki/Quantum_computing
I'm not a physics person... but my understanding is that qubit (as opposed to a "bit" - 0 or 1 - in normal computers) can hold many states at once, and therefore it can calculate many results simultaneously.
Experts agree that traditional encryption becomes almost useless with the advent of useful Quantum computers (because they are so good at factoring large integers).
Today we only have the beginnings of Quantum Computing... and we will likely see hybrids or "quantum GPUs" added into traditional Supercomputers once they figure out some more physics stuff (like how to do this at "nearer normal" temperatures).
I have heard privately from the NASA people that use the Quantum system that they bought (in partnership with Google I believe) that it wasn't that effective. However that wasn't the intent, the intent was to invest in the future and start adapting codes to work with Quantum Supercomputers.
----
By the way, there is a class of problems in supercomputing known as "embarrassing parallel". These are problems that can be split up and sent to processors where other than the initial data set, there isn't much communication (if any) that needs to happen until a result is known... factoring prime numbers, protein folding, even SETI signal analytics... can all be done by using thousands or millions of home computers with only marginal internet connections (know as "XXX-at-home" projects). I urge all DUers to participate in such science projects if they have the resource. I think there was even an interest subgroup promoting those projects.
A side note, some websites have become infected with a virus called Coin-Hive. This is attempting to use your computer to mine bitcoins or other virtual currency by running code on your computer (without your permission) to do blockchain encryption functions simply by you visiting the website.
Grins
(7,231 posts)Got it from her after I met her in Atlanta's airport after a conference there. She had a whole bunch of them.
And my wire - like yours - is framed!!!
And she was very gracious to talk to. I happened to mentioned to her an Army General and his fiercely adamant views on a certain computer language. Her response: "He's an idiot!"
Nope. She wasn't shy either!
lapfog_1
(29,226 posts)I met her at the dedication of our Cray-1... I was one of about 3 people on the "Cray" team so I was instrumental in making the arrangements for her talk.
Her pronouncement on the brand new $20M (1980 dollars) machine??? "It's a dinosaur!"
She was right but about 10 years ahead of her time (maybe 15). But she knew parallel computing was the future (and not vector).
I've only personally met a few people as impressive than her... one was Richard Feynman. Which gets back to Quantum Computing.
MSNBC was running a thing about statues to people ( especially women ) who have contributed to our nation. I think we should nominate her.
Of course, having a U.S. Naval Ship named after you (in addition to other things) is pretty awesome.
pwb
(11,291 posts)People that could help us stay on top don't have a chance.
Irish_Dem
(47,440 posts)Edited to add: They enjoy all the benefits of living in the US, but refuse to give back any of their wealth to support the country. They are like Putin who has drained Russia of many of its assets.
mahatmakanejeeves
(57,618 posts)Would you like to share with us the rationale for your remarkable pronouncement?
Thanks.
pwb
(11,291 posts)The Saudis, Kim Jung Un, most dictators, papa john whatever his name is. It wasn't meant to be a broad brush statement. Most people would pick up on that, I thought. All the money in so few hands lessens our chance for success as a country. Having money does not necessarily go along with being bright. Think Trumps. Anyone can be picky about anything can't they? But thank you for your remarkable imagination and thought.
Hassin Bin Sober
(26,343 posts)lunasun
(21,646 posts)NCjack
(10,279 posts)Snotcicles
(9,089 posts)Calista241
(5,586 posts)The US, both the government and commercial sector, is investing hundreds of billions annually into data center and fiber communication infrastructure.
This infrastructure investment is more relevant to our daily lives and our future economy than investing in a single computer network designed to model hyper complex problems. For example, If you wanted to track a red blood cell through a human body 10 minutes or even 10 hours faster than a similar computer in the US. Or, as another poster mentioned, If you wanted to model climate change effects faster than other computers. Or if you want to model the expansion of the latest supernova Hubble photographed. Thats the kind of stuff super computers are used for.
The DOE estimates There are over 3 million data centers here in the US. The US has 7 of the top 10 largest data centers in the world. Other than those 7, theres 1 in China, 1 in India and 1 in the UK.
jmowreader
(50,562 posts)they're also useful for manipulating the securities markets.
Calista241
(5,586 posts)You just need one thats fast enough to get the job done.
DoctorJoJo
(1,134 posts)paleotn
(17,989 posts)at the beginning of the 20th century. The US is Great Britain. Within 50 years or less, we'll be a paper tiger.
harun
(11,348 posts)them in hundreds of thousands of dollars of debt, that is if they can even get in to a College if they aren't rich.
U.S. cannot make STEM college for US citizens free fast enough.
mahatmakanejeeves
(57,618 posts)their student loans.
harun
(11,348 posts)Engineering graduate numbers:
(China and India are much higher but hard to get real numbers)
Russia 454,000
US 237,826
Iran 233,695
Indonesia 140,000
Vietnam 100,000
https://www.weforum.org/agenda/2017/04/higher-education-in-china-has-boomed-in-the-last-decade
https://www.quora.com/How-many-engineering-students-graduate-every-year-in-India-and-how-can-one-aspire-to-get-it
https://timesofindia.indiatimes.com/india/India-set-to-produce-worlds-largest-number-of-engineers/articleshow/49532113.cms
sandensea
(21,672 posts)this will be the new normal.
xor
(1,204 posts)The US still has a massive amount of computing power. Far more than the rest of the world (aside from China) It's just that China is actually competition here, and they are really going at it. It's not like the US is sitting on its ass doing nothing while China takes it all. President Obama signed this a couple years ago https://en.wikipedia.org/wiki/National_Strategic_Computing_Initiative
Now, China could also finish their exascale supercomputers before that, but I guess we'll have to see how that plays out as apparently things get really hard to do at that point (based on my limited understanding of what's going on) Maybe these results will spark something to get more funding or whatever is needed.