China Dominates the World TOP500 Supercomputers

[cirr]

@cirr @AndrewJin

What is the future Chinese development track regarding HPC?

As far as the US is concerned, it has financed three 100-200 Petaflop machines by 2018, and then start work on Exascale, to be commissioned by 2023.

The only knowledge that I have is that China will upgrade Tianhe 2 to Tianhe 2A, anywhere between 100-200 Petaflops.

Of course there is the plan to reach Exascale by 2020, but I think US would lead in performance count, and may be even in top machine if China doesn't introduce one more machine above 100 Petaflops by 2018.

Briefly
(1) 100 Pflop Tianhe-2A later in the year(NUDT and partners)
(2) 200-300 Pflop Tianhe-2B in 2018(NUDT and partners)
(3) an upgraded Sunway Taihulight in 2 years(Jiangnan Inst. and partners)?
(4) exascale prototype in 2018(NUDT and partners)
(5) commerical exascale in 2020(NUDT and partners)
(6) CAS has more or less finalized its roadmap for an exascale(don't know if CAS will build the machine on its own hook or with partners, but it can certainly do with a better machine for its supercomputing centre)
(7) a bunch of other HPCs by the likes of Sugon(the Dawning series), Inspur, Lenovo etc.
(8) quantum computer prototype in or around 2020(（USTC and partners. Real quantum computer, not D-Wave the pretender)
(9) a bunch of cities vying for supercomputing centres, state approval being sought
In short, this is not a turtle hare race. China is in this game for the long term.

 

[TaiShang]

The Sunway TaihuLight has already provided computational support to research universities and institutes in China, such as Tsinghua University, Beijing Normal University, Chinese Academy of Sciences, and the First Institute of Oceanography, and achieved significant performance results in key research domains.

Looks like the TaihuLight is going to enjoy high demand from universities across the country.

 

[AndrewJin]

If every Indian was the same with these India guys in this thread, it would bring much fun to us.

Mostly are.
Some not so insane ones have left.

Briefly
(1) 100 Pflop Tianhe-2A later in the year(NUDT and partners)
(2) 200-300 Pflop Tianhe-2B in 2018(NUDT and partners)
(3) an upgraded Sunway Taihulight in 2 years(Jiangnan Inst. and partners)?
(4) exascale prototype in 2018(NUDT and partners)
(5) commerical exascale in 2020(NUDT and partners)
(6) CAS has more or less finalized its roadmap for an exascale(don't know if CAS will build the machine on its own hook or with partners, but it can certainly do with a better machine for its supercomputing centre)
(7) a bunch of other HPCs by the likes of Sugon(the Dawning series), Inspur, Lenovo etc.
(8) quantum computer prototype in or around 2020(（USTC and partners. Real quantum computer, not D-Wave the pretender)
(9) a bunch of cities vying for supercomputing centres, state approval being sought
In short, this is not a turtle hare race. China is in this game for the long term.

I dunno what u are talking about, look like some cool stuff

India has a better chance getting 10 gold medals in 2020 Olympics than achieving 132 exa flops.

lol
Let's bet this year's game.
If India can get 5 gold medals, I will donate one dollar for their supa powa computa.

 

[TaiShang]

The US banned the use of Intel chips in China's supercomputers.

This is great for China and it turned out to be a blessing in disguise. This forces China to develop and use its own indigenous chips. Also, other major countries are now thinking the same, it's better not to rely on US tech/chips. The US will ban the use of it when it suit them.

I wonder what else the US can do to China. The answer - not much.

Maybe, the US should ban the export of Boeing planes to China. Let's see if this will give a boost to the C919.

I think sanctions and other ways of punitive measures tend to backfire on certain nations; it is not like you sanction Iraq for two decades and keep children from key medicine, weaken the country and invade it in a second attempt.

Sanctions always misfire on countries like Russia. It solidifies the leadership, invigorates industry and put unbridgeable distances between peoples.

US won't be a partner for Russia in the coming five centuries, for sure, and same goes for China. You show so much hostility and expect us to forget? That's impossible on great nations although it might work on weak ones.

This supercomputer saga involves lots of lessons.

 

[AndrewJin]

I think sanctions and other ways of punitive measures tend to backfire on certain nations; it is not like you sanction Iraq for two decades and keep children from key medicine, weaken the country and invade it in a second attempt.

Sanctions always misfire on countries like Russia. It solidifies the leadership, invigorates industry and put unbridgeable distances between peoples.

US won't be a partner for Russia in the coming five centuries, for sure, and same goes for China. You show so much hostility and expect us to forget? That's impossible on great nations although it might work on weak ones.

This supercomputer saga involves lots of lessons.

There is one thing in which China and US won't compete, that's building a wall which will be paid by Mexicans.

 

[TaiShang]

There is one thing in which China and US won't compete, that's building a wall which will be paid by Mexicans.

Maybe Mr. Trump will outsource part of the engineering (the parts that involve difficult terrain and other technological challenges) to China's companies whose track record shows that they can build anything anywhere.

Do you think Trump would be bribed?

 

[AndrewJin]

Maybe Mr. Trump will outsource part of the engineering (the parts that involve difficult terrain and other technological challenges) to China's companies whose track record shows that they can build anything anywhere.

Do you think Trump would be bribed?

Bribery is legal in a demoncrazy.
Just sign a deal with the regime in the Trump Tower

 

[TaiShang]

Bribery is legal in a demoncrazy.
Just sign a deal with the regime in the Trump Tower

Exactly. These days they technically call "bribe" a "campaign donation."

 

[rott]

I saw from many western news report this and many of western losers cannot accept this reality. Some has the typical claim of Chinese stole their technology. Ya!! We stole a technology you yourself can't even build

I can bet many of the are killing themselves over this.

Exactly, "how can we steal something that you don't have?" begs the question.

 

[hexagonsnow]

Sunway-TaihuLight outperforms Tianhe-2 as world's fastest supercomputer
Source: Xinhua | 2016-06-20 15:56:14 | Editor: huaxia

WUXI, June 20, 2016 (Xinhua) -- Photo taken on June 16, 2016 shows interchanger cables of Sunway TaihuLight, a new Chinese supercomputer, in Wuxi, east China's Jiangsu Province. Performing 93 quadrillion calculations per second, Sunway TaihuLight dethroned China's Tianhe-2 from the top in a list of the 500 most powerful supercomputers in the world. Sunway TaihuLight, with 10,649,600 computing cores comprising 40,960 nodes, is twice as fast and three times as efficient as Tianhe-2, which has a performance of 33.86 quadrillion calculations per second, or petaflop/s. The new system was developed by the Chinese National Research Center of Parallel Computer Engineering & Technology and installed at the National Supercomputing Center in Wuxi. (Xinhua/Li Xiang)​

NANJING, June 20 (Xinhua) -- China's new supercomputing system, Sunway-TaihuLight, was named the world's fastest computer at the International Supercomputing Conference in Germany on Monday.

The National Supercomputing Center was also unveiled Monday in Wuxi, east China's Jiangsu Province, where the new-generation supercomputer is installed.

With processing capacity of 125.436 petaflops (PFlops) per second, which means it can perform quadrillions of calculations per second at peak performance, Sunway-TaihuLight is the first supercomputer to achieve speeds in excess of 100 PFlops.

The computing power of the supercomputer is provided by a China-developed many-core CPU chip, which is just 25 square cm.

 

[HariPrasad]

Dear elite member, it is OK to admit you mistakenly boasted your goal by 1000 folds, and you would have earned the respect, at least from me. By listing up those "me-too" achievements proudly, you actually further embarrassed yourself in front of learned members of PDF.

We are happy for Indian's development, but none of those really amount to any degree of "frustration" to us. By thinking your petty success would somehow impact others in any significant way, you are showing a classic trait of certain people, overestimating your significance a little too much.

Telling people you are making $100 per day is a cute way to make argument. I do hope other elites in Indian society have better grasp of today's world, especially in science and tech front.

You can rant whatever suites you. In last year or so, It is undisputadely proven that who is the boss in space technology and rocketry. It is a matter of time when it will be proven that who is the boss in super computing. We are heading for THAAD like air defense and nano missiles. Ultra long range, Air breathing , composite motor etc. In the same way, we are putting up a nation wide super computing network and our 132 exa flop computer shall be ready by 2020. just 600 time faster than what you have. so enjoy your superiority for few years untill we surpass you with unmatchable lead in super computing speed.

 

[cirr]

I love it when the Indians boasts, "we'll train the Vietnamese" or "we'll train the Philippines". With an average IQ lower than the Vietnamese or the Pinoys, I wonder what the training is all about.

Guess who will come out the winner when the Indian army is pitted against the Vietnamese army?

 

[ahojunk]

There is one thing in which China and US won't compete, that's building a wall which will be paid by Mexicans.

.
China already has the Great Wall. There is no need for another one.

What is needed is the OBOR.

 

[HariPrasad]

I love it when the Indians boasts, "we'll train the Vietnamese" or "we'll train the Philippines". With an average IQ lower than the Vietnamese or the Pinoys, I wonder what the training is all about.

It is proved in number of times whose IQ is high. You can not reach Mangal even with the help of russian rocket. Failed to carry out orbital correction maneuver and even could not leave even earth orbit. Here you are talking of high IQ. The fact of the matter is that CHinese is not known for their high IQ ever anywhere.

 

[cirr]

Totally incorrect. Trust me I have met many people who actually work on these things. Our national supercomputing mission aims to put three 5-10 Petaflops machines by 2020.

Volume production of cutting-edge ultrahigh modulus carbon fibre starts

http://news.ifeng.com/a/20160621/49207693_0.shtml

Modulus of elasticity > 700GPa
tensile strength > 3239 MPa
Load: 26.24CN

can be optimised for further huge increase in performances