What's new

Next Frontier of Supercomputer: China's Exascale in the Making

Chinese Processors?

Are any Chinese processor available in the market that can compete with Intel/AMD? Any plans?

And how about Chinese processors against Taiwanese processors etc., for mobile computing? There should be threads on these things.
 
I am interested in HPC, and follow the scene. So, in the community of HPC, it is not a big secret. Rather, Jack Dongarra, the person behind TOP 500 even told it in open that China will have a new top supercomputer at ISC 2016.
I think u should spend more time on college and chemistry!
 
Chinese Processors?

uncle sam banning intel from selling hpc chips to china. what do you think?

Are any Chinese processor available in the market that can compete with Intel/AMD? Any plans?

only for hpc, space and military stuff. i believe they can make supa aweshome cpu for consumers like intel, but nobody will use 'em. it's too late to get in the game now. it's like windows phone os with very few app developers and manufacturers supporting the platform. so no plan...

And how about Chinese processors against Taiwanese processors etc., for mobile computing? There should be threads on these things.

huawei uses their own chips in their phones and networking equipments. xiaomi will use their own too and many more fabless that just do designing for others like spreadtrum, rockchip, allwinner, actions..

this company here is developing their own architecture from scratch. i think they won't go anywhere. like i said above, it's too late now.
Code:
http://icubecorp.com/

this startup here might become successful, because ai chip is still very new. they have a lot big backers behind them. i can see their chips will be in many chinese drones. maybe dji phantom 7 :D?
Code:
http://www.horizon-robotics.com/index_en.html
Code:
http://www.scmp.com/tech/article/1921396/chinese-start-track-deliver-artificial-intelligence-chip

LhLZAZR.jpg

.
 
Last edited:
China is predicted to announce the world's first 100 petaFLOPS computer tomorrow

The supercomputer race might have just been won.

FIONA MACDONALD

20 JUN 2016

Experts are reporting that China will announce tomorrow that it's built the world's first 100 petaFLOPS computer - an incredibly powerful supercomputer capable of processing a quadrillion floating point operations per second (FLOPS).

To put that into perspective, Apple's latest Mac Pro tower has up to 7 teraFLOPS of performance. This rumoured new computer would be able to process a mind-blowing 1,000 teraFLOPS each second. And it's a computing milestone the US government has reportedly worked hard to stop China from reaching.

If you haven't heard much about the supercomputer race so far, here's what you need to know. Back in April last year, the US State Department set up an embargo that blocked the sales of Intel Xeon and Xeon Phi processors to Chinese institutions - most notably the National University for Defence Technologies (NUDT), which is home to the Tianhe-2 supercomputer, the most powerful supercomputer in the world since 2013, with a sustained performance of 33.86 petaFLOPS.

Impressively, Tianhe-2 isn't even fully built yet, and experts speculated that the US's sales block was to prevent the Chinese government from improving the supercomputer further.

But instead of slowing down performance, China has since spent all the money it had budgeted for foreign superprocessors on the development of its own Alpha and ARM superprocessors.

"In terms of funds, NUDT planned to buy 32,000 more Xeon processors ... and 48,000 more Xeon Phi co-processors," reports Theo Valich for VR World. "We’ve been hearing that over US$500 million was invested in bringing the Chinese silicon from a prototype phase to production-grade level.":partay:

All that investment has allowed Chinese engineers to upgrade Tianhe-2 and they'll be announcing the new performance tomorrow on June 20 at the 2016 International Supercomputing Conference in Germany.

Right now, the idea that they'll announce the 100 petaFLOPS milestone is purely speculative, so let's not all get too excited just yet. But there are some good indications that it's about to happen.

For starters, Tianhe-2 was always intended to reach 100 petaFLOPS at full capacity - and at the Supercomputing Frontiers Conference in Singapore back in March, NUDT engineers announced that it's now fully developed (that's a pretty big clue right there).

Secondly, based on the amount of investment into local processors, it makes sense that they should have now hit their goal. Tianhe-2 already had 32,000 Intel Xeon processors and 48,000 Intel Xeon Phi co-processors, giving it peak performance of 54.9 petaFLOPS and a sustained performance of 33.86 as of 2013, according to the internationally recognised Linpack benchmark.

So what's changed now? Valich explains:

"The new Tianhe-2 represents a hybrid design, featuring two new additions, as the old Xeon Phi cards are being phased out. Phytium Technologies recently delivered their 'Mars' processors in the form of PCI Express cards that replaced the Xeon Phi cards, and motherboards to upgrade the system.

Given that there are 48,000 add-in boards installed, the new 64-core design enables the system to reach its original performance targets. With the three million new ARM cores inside the Tianhe-2, its estimated [peak] performance in the Linpack benchmark should exceed 100 petaFLOPS."

Tl;dr: Unless something's gone really wrong, China should have hit the 100 petaFLOPS milestone - all without the help of any new Intel chips from the US.

What does that mean for those of us who don't work in supercomputing? Quite a lot, actually, seeing as supercomputers provide such enormous benefits in so many fields. Basically, anything that our regular computers can do, supercomputers can do thousands of times faster.

And that will give China a competitive edge in fields such as weather forecasting, oil and gas exploration, drug development, physical simulations, engineering... pretty much anything you need a computer for, China will now be able to do better than anyone else in the world.

And 100 petaFLOPS isn't the end. "Should Tianhe-2 reach its full deployment of 32,000 Xeons, 32,000 ShenWei processor, and 96,000 Phytium accelerator cards, we might see an upgrade in the range of 200-300 PFLOPS - if the building can withstand the thermal and power challenges associated with it," reports Valich.

In fact, there are rumours that China will actually be announcing not one, but two 100 petaFLOPS computers this year.

Like we mentioned earlier, it's not over until the fat lady sings, and a milestone hasn't been reached until it's been independently verified and peer-reviewed. But it's looking highly likely that tomorrow might be a pretty big win for China, and we'll be watching closely so we can keep you updated with announcements.

http://www.sciencealert.com/expert-...-world-s-first-100-petaflop-computer-tomorrow
 
June 2016 is a very good month for China. More good news coming up. Hater just wait and cry later :lol:
 
I'll believe it when I see it :cheesy:

still waiting on the updated Tianhe-2 300 PFlop coming this year or next :pop:
 
I'll believe it when I see it :cheesy:

still waiting on the updated Tianhe-2 300 PFlop coming this year or next :pop:

Don't worry, you need not wait too long and you can cry on your bed later. :lol:
 
Don't worry, you need not wait too long and you can cry on your bed later. :lol:


why would I cry buddy :what:

I am happy China is advancing on it own and not copying like it has in the past.

I just have doubts is all :wave:

like I said I'll wait for this 300 Pflop with Chinese co-processors and Tianhe-3

U.S has it's own 100+ Pflop and 1 exa in the works as well, it's all good :oops:


just got done reading about Stampede 2......

http://www.top500.org/news/stampede-2-the-18-petaflop-sequel/

beast of a machine

 
why would I cry buddy :what:

I am happy China is advancing on it own and not copying like it has in the past.

I just have doubts is all :wave:

like I said I'll wait for this 300 Pflop with Chinese co-processors and Tianhe-3

U.S has it's own 100+ Pflop and 1 exa in the works, it's all good :oops:


just got done reading about Stampede 2

http://www.top500.org/news/stampede-2-the-18-petaflop-sequel/

impressive machine.
While you switch tone so quickly? I thought few months ago. You still brag about your mighty US and so full of yourself. Even post garbage anti China link smearing like tian-he 2A inefficient and all those nonsense.

It's good that you eat your humble pie! :lol:
 
While you switch tone so quickly? I thought few months ago. You still brag about your mighty US and so full of yourself. Even post garbage anti China link smearing like tian-he 2A inefficient and all those nonsense.

It's good that you eat your humble pie! :lol:


Tianhe-2A is inefficent and that's why I have doubts about this 300 PFlop and 1 exaflop by 2020


biggest challenge for reaching EXA is power/heat.


China just builds inefficent systems that turn into white elephants. seems this is more about PRIDE than actually getting work done, which is what happens in the U.S,EU, and Japan.
 
why would I cry buddy :what:

I am happy China is advancing on it own and not copying like it has in the past.

I just have doubts is all :wave:

like I said I'll wait for this 300 Pflop with Chinese co-processors and Tianhe-3

U.S has it's own 100+ Pflop and 1 exa in the works as well, it's all good :oops:


just got done reading about Stampede 2......

http://www.top500.org/news/stampede-2-the-18-petaflop-sequel/

beast of a machine


The main problem I can see for the HPC development between US and China is that, US is going on the other direction, while China continue on with ARM design.

Technically, we can make a ARM into exascale supercomputer now, basically it is sheer core and computing power, by bunching a bunch (I mean A LOT) of CPUs in a supercomputer, You are bounded to reach that level of operation, but in turn, how much power it drawn and how you can maintain it is whole other question together.

US on the other hand, uses a New 3D-Node design and use a hybrid CPU/GPU which utilise the core parallel processing power of a GPU coupled with the decision making CPU to form the backbone of a exascale computer.

Unless there are way to make the processor core consume less power, I don't think ARM architecture is the way to go.
 
The main problem I can see for the HPC development between US and China is that, US is going on the other direction, while China continue on with ARM design.

Technically, we can make a ARM into exascale supercomputer now, basically it is sheer core and computing power, by bunching a bunch (I mean A LOT) of CPUs in a supercomputer, You are bounded to reach that level of operation, but in turn, how much power it drawn and how you can maintain it is whole other question together.

US on the other hand, uses a New 3D-Node design and use a hybrid CPU/GPU which utilise the core parallel processing power of a GPU coupled with the decision making CPU to form the backbone of a exascale computer.

Unless there are way to make the processor core consume less power, I don't think ARM architecture is the way to go.
I know China is developing heavily on ARM/MIPS and it's possible to build a high Pflop and EXA with that, just wondering if it's inefficent enough to do.

you will still be looking at like a 50+MW system
 
The main problem I can see for the HPC development between US and China is that, US is going on the other direction, while China continue on with ARM design.

Technically, we can make a ARM into exascale supercomputer now, basically it is sheer core and computing power, by bunching a bunch (I mean A LOT) of CPUs in a supercomputer, You are bounded to reach that level of operation, but in turn, how much power it drawn and how you can maintain it is whole other question together.

US on the other hand, uses a New 3D-Node design and use a hybrid CPU/GPU which utilise the core parallel processing power of a GPU coupled with the decision making CPU to form the backbone of a exascale computer.

Unless there are way to make the processor core consume less power, I don't think ARM architecture is the way to go.
Another loser and Ah Q comment. America can't never beat China with mentality like you. Can't even be a graceful loser and so thick skinned even face with total defeat. :lol:
 
I know China is developing heavily on ARM/MIPS and it's possible to build a high Pflop and EXA with that, just wondering if it's inefficent enough to do.

you will still be looking at like a 50+MW system

More than 50+ MW, It will be in GW........If I remember correctly, I think someone done some calculation before, the only way you can power an ARM Exascale Supercomputer is to have a Nuclear reactor right next to it. Don't forget, you need uninterrupted power supply for any supercomputer, and that plus the nuclear reactor is the reason why nobody is building ARM exascale computer at this time.

Another loser and Ah Q comment. America can't never beat China with mentality like you. Can't even be a graceful loser and so thick skinned even face with total defeat. :lol:

Firstly, it's not about winning or losing, it's about technological advancement. You defeat someone when the race Is OVER not before, unless you claim the Chinese build "THE MOTHER OF SUPERCOMPUTER" and nobody can ever beats Tian-he 1,2 or 3, you defeat nobody.

US is light year ahead than China in CPU technology, you can try to denied it but people will just laugh at you for your own ignorant. If US wanted they could have build a Tianhe-2 like machine using ARM. Or you honestly think US cannot build a TianHe Like machine which just focus on core processing power? Given the ARM Architecture is a British Technology and both CPU and Interconnect Card are US Technology. What makes you think US cannot bunch 60,000 Intel E7 CPU and get to 100 Pflop? But if the American do this, then what would be the fun of it?

Plus, you also don't understand that you will need the 3-D node to break the current supercomputer barrier, you can stay with ARM all you want, and claim this and that and be happy about it, but the American is simply moving on to the new technology.
 

Back
Top Bottom