News on China's scientific and technological development.

SamuraiBlue · Nov 30, 2016

Quickie said:
Haven't you heard yet. There has been an embargo on high end chips for use in top Chinese supercomputers not only from Intel but Nvidia and AMD as well. Tianhe-2 was then about to be boosted to 110 petaflops which would have left the K-Computer in the smoke in whatever benchmarks they can come out with.....

Please, Log in or Register to view URLs content!

In an aggressive move to curb’s China’s advancements in supercomputing the US government has proceeded to ban sales of high end Intel, Nvidia and AMD CPUs and GPUs for use in Chinese supercomputers.

This development came to light as Intel was preparing to ship high-end Xeon E5 Haswell based processors to refresh the Tianhe-2 supercomputer belonging to the Chinese government. The plan for the Tianhe-2 supercomputer was to upgrade its computational capability from 33 petaflops to 110 petaflops reports the
Please, Log in or Register to view URLs content!
. However the US Department of Commerce barred Intel from shipping the processors required for the upgrade citing worries over nuclear weapons related research.

Aughhh, so?????
I believe you are barking up the wrong zip code.
What are you asking of me, pity ? Want me to think you Chinese will do better if you have had better CPUs?????
To me that is irrelevant. To my understanding the present CPUs that is powering the Sunway TaihuLight is based on ARM architecture.
The Tianhe-2 that is ranked as #2 on November 2016 HPCG is powered by Xeon CPUs. By the way TaihuLight which is supposedly the fasted super computer ranks #4 on the same rating slower than the Tianhe-2 and Tianhe-2 has not changed it's out put from June 2014.
No optimization from then and why is a supposedly faster computer TaihuLight designed and made years after the lower ranking Tianhe-2 that had been developed before the more advanced TaihuLight be ranking higher on different tests??
Makes me wonder how much intelligence had been acquired through the pain staking years spent on R&D?

Quickie · Nov 30, 2016

SamuraiBlue said:
Aughhh, so?????
I believe you are barking up the wrong zip code.
What are you asking of me, pity ? Want me to think you Chinese will do better if you have had better CPUs?????

Butt hurt to know the truth?

Why should anyone ask pity from you who come in here to poo poo any kind of achievement of another nation other than your own?

To me that is irrelevant. To my understanding the present CPUs that is powering the Sunway TaihuLight is based on ARM architecture.
The Tianhe-2 that is ranked as #2 on November 2016 HPCG is powered by Xeon CPUs. By the way TaihuLight which is supposedly the fasted super computer ranks #4 on the same rating slower than the Tianhe-2 and Tianhe-2 has not changed it's out put from June 2014.
No optimization from then and why is a supposedly faster computer TaihuLight designed and made years after the lower ranking Tianhe-2 that had been developed before the more advanced TaihuLight be ranking higher on different tests??
Makes me wonder how much intelligence had been acquired through the pain staking years spent on R&D?

So, going back to all the old arguments which have already been discussed to death?

Look, the different benchmarks are different kind of test of the supercomputer system with some of them more specific to the different hardwares of the system.

All these benchmarks are just an indication of how the supercomputer would perform in the different kind of task.

The ultimate test is to run on the actual kind of application that the supercomputer was designed to run on.

Different supercomputers may have different performances in the various task and it's useless to harp on how good the computer is at only a certain task since in the end the only thing that matters is the final outcome of the performance in the overall running of the application.

broadsword · Nov 30, 2016

SamuraiBlue said:
Makes me wonder how much intelligence had been acquired through the pain staking years spent on R&D?

With far more years of experience in developing supercomputers and with the backing of Fujitsu, Hitachi and NEC and the government, and a state of the art chip designing and fabrication industry, I would expect the K to be at least 4 times the speed of the Chinese systems.

But, it was only slightly faster.

The Chinese chips, in terms of nm, are being two generations behind and yet they are ranked top in Top500 and second and fourth on the HPCG.

So who has learned more to bridge the gap?

delft · Nov 30, 2016

Just a wild story I read a few months ago:
A result was calculated on a supercomputer in Bochum in three days and the same result was achieved with an Nvidia card in Moscow in a quarter of an hour. No doubt the problem was chosen to be difficult for the supercomputer and particularly easy for the modern video card. It shows that a general purpose supercomputer might not be the right choice for your particular problem.

solarz · Nov 30, 2016

delft said:
Just a wild story I read a few months ago:
A result was calculated on a supercomputer in Bochum in three days and the same result was achieved with an Nvidia card in Moscow in a quarter of an hour. No doubt the problem was chosen to be difficult for the supercomputer and particularly easy for the modern video card. It shows that a general purpose supercomputer might not be the right choice for your particular problem.

Quite possible, GPUs and CPUs are built for different kinds of operations. Building on this difference, it is quite possible to create a problem that specifically targets the weaknesses of a particular supercomputer architecture, while being trivial to a GPU.

Quickie · Nov 30, 2016

t2contra said:
With far more years of experience in developing supercomputers and with the backing of Fujitsu, Hitachi and NEC and the government, and a state of the art chip designing and fabrication industry, I would expect the K to be at least 4 times the speed of the Chinese systems.

But, it was only slightly faster.

The Chinese chips, in terms of nm, are being two generations behind and yet they are ranked top in Top500 and second and fourth on the HPCG.

So who has learned more to bridge the gap?

delft said:
Just a wild story I read a few months ago:
A result was calculated on a supercomputer in Bochum in three days and the same result was achieved with an Nvidia card in Moscow in a quarter of an hour. No doubt the problem was chosen to be difficult for the supercomputer and particularly easy for the modern video card. It shows that a general purpose supercomputer might not be the right choice for your particular problem.

solarz said:
Quite possible, GPUs and CPUs are built for different kinds of operations. Building on this difference, it is quite possible to create a problem that specifically targets the weaknesses of a particular supercomputer architecture, while being trivial to a GPU.

That is exactly the things that's at issue here. It has nothing to do with the delusion that somehow just because a certain supercomputer top a certain benchmark that's skewed toward a certain kind of problem, it must necessary be of a superior design that is beyond the grasp of other supercomputer designers.

manqiangrexue · Dec 1, 2016

SamuraiBlue said:
Aughhh, so?????
I believe you are barking up the wrong zip code.
What are you asking of me, pity ? Want me to think you Chinese will do better if you have had better CPUs?????
To me that is irrelevant. To my understanding the present CPUs that is powering the Sunway TaihuLight is based on ARM architecture.
The Tianhe-2 that is ranked as #2 on November 2016 HPCG is powered by Xeon CPUs. By the way TaihuLight which is supposedly the fasted super computer ranks #4 on the same rating slower than the Tianhe-2 and Tianhe-2 has not changed it's out put from June 2014.
No optimization from then and why is a supposedly faster computer TaihuLight designed and made years after the lower ranking Tianhe-2 that had been developed before the more advanced TaihuLight be ranking higher on different tests??
Makes me wonder how much intelligence had been acquired through the pain staking years spent on R&D?

The answer is simple. It's because you're completely fixated on a ranking system that the mainstream doesn't use as the true standard and you're doing that to make yourself feel better because it's the standard that says something close to what you want to hear. You're picking the only thing that fits into your narrative and ignoring everything else. You could just as easily say, "Well, when it comes to speed, MiG-25 is the fastest. So many years and development of aircraft, J-10, Su-35, F-35, F-22, J-20; they all can't match the speed of the foxbat. Makes me wonder if people have made any progress at all on fighter jets..."

Franklin · Dec 1, 2016

We had Super computing, we had cloud computing, we had quantum computing and now we have edge computing.

Edge Computing Consortium is Established to Deepen Digital Transformation

BEIJING, Nov. 30, 2016 /PRNewswire/ -- Today, the Edge Computing Consortium (ECC) was officially established in Beijing, China. This initiative was jointly created by Huawei Technologies Co., Ltd., Shenyang Institute of Automation of Chinese Academy of Sciences, China Academy of Information and Communications Technology (CAICT), Intel Corporation, ARM, and iSoftStone.

The ECC aims to build a cooperative platform for the edge computing industry that will give impetus to openness and collaboration in the Operational Technology (OT) and Information and Communications Technology (ICT) industries, nurtures industrial best practices, and stimulates the healthy and sustainable development of edge computing.

Today's global digital revolution is driving a new round of industrial restructuring. Through the digital transformation of industries, products are incorporated into intelligent interconnection. In-depth coordination and convergence of OT and ICT help improve industrial automation, meet the customized requirements of products and services, promote full-lifecycle transformation from products to service operations, and trigger the innovation of products, services, and business models. This will have a lasting impact on the value chain, supply chain, and ecosystem.

Yu Haibin, Chairman of the ECC and Director of Shenyang Institute of Automation, the Chinese Academy of Sciences, said, "In the 13th Five Year Plan, China launched two national strategies, integration of digitization and industrialization, as well as 'Made in China 2025'. This requires much on ICT and OT convergence. Edge Computing is key to supporting and enabling this convergence. Meanwhile, industrial development is also facing a turning point.

"Industrial automation technology systems will evolve from layered architecture and information silos to IoT, cloud computing, and Big Data analytics architecture. Amidst the evolution, edge computing will bolster distributed industrial automatic self-control architecture. Therefore, the ECC will keep an eye on the design of the architecture and the choice of technical roadmap, as well as promoting industrial development through standardization. In addition, building an ecosystem will also be focused," continued Yu Haibin.

The ECC is in pursuit of the OICT concept that OT, information technology (IT), and communications technology (CT) resources should integrate and coordinate with each other, and stick to the spirit of consensus, unity, and win-win cooperation, to drive forward the ECC's healthy development. The ECC strives to advance cooperation among industry resources from government, vendors, academics, research, and customer sectors.

The Edge Computing Consortium's White Paper was also released at the 2016 Edge Computing Industrial Summit, during the ECC's launch ceremony. It puts emphasis on the edge computing industry's trends and major challenges, elaborates on the definition and content of edge computing, displays the ECC's top-level design and operational model, and formulates the reference architecture and technological framework of edge computing, guiding the ECC's future development.

Please, Log in or Register to view URLs content!

Iron Man · Dec 1, 2016

manqiangrexue said:
The answer is simple. It's because you're completely fixated on a ranking system that the mainstream doesn't use as the true standard and you're doing that to make yourself feel better because it's the standard that says something close to what you want to hear. You're picking the only thing that fits into your narrative and ignoring everything else. You could just as easily say, "Well, when it comes to speed, MiG-25 is the fastest. So many years and development of aircraft, J-10, Su-35, F-35, F-22, J-20; they all can't match the speed of the foxbat. Makes me wonder if people have made any progress at all on fighter jets..."

Exactly. I have no doubt that if the K Computer was at the top of the LINPACK benchmark list and TianhuLight was at the top of Graph500, this SadBlue character would be screaming that LINPACK was the only appropriate benchmark to use and that Graph500 was totally irrelevant. He has been completely beaten down on so many topics in so many threads he must be a masochist to still be trolling SDF.

taxiya · Dec 1, 2016

Iron Man said:
Exactly. I have no doubt that if the K Computer was at the top of the LINPACK benchmark list and TianhuLight was at the top of Graph500, this SadBlue character would be screaming that LINPACK was the only appropriate benchmark to use and that Graph500 was totally irrelevant. He has been completely beaten down on so many topics in so many threads he must be a masochist to still be trolling SDF.

In your post, I see a characteristic word to describe people of the same origin of him.

News on China's scientific and technological development.

SamuraiBlue

Captain

Quickie

Colonel

broadsword

Brigadier

delft

Brigadier

solarz

Brigadier

Quickie

Colonel

manqiangrexue

Brigadier

Franklin

Captain

Iron Man

Major

taxiya

Brigadier