China's GPU Supercomputer Begins Paying Off Big Time

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
Scientists at the Institute of Process Engineering (IPE) at the Chinese Academy of Sciences (CAS) today announced a record-breaking scientific simulation on the Tianhe-1A GPU supercomputer that furthers their research in solar energy. CAS-IPE scientists ran a complex molecular dynamics (MD) simulation on all 7,168 NVIDIA Tesla GPUs to achieve a performance of 1.87 petaflops per second – that’s about the same performance as 130,000 laptops. These scientists are simulating the structure of crystalline silicon which is used in solar panels and also in the semiconductor industry. And they were able to accomplish this world’s fastest MD simulation by writing just 2,000 lines of CUDA code.
 
2000 lines of just CUDA code? Sounds like this was rather complex!
 
that’s about the same performance as 130,000 laptops.

In the name of clarity, I have a few questions.

Would those laptops have Intel processors, AMD processors, MIPS processors, or some other type? What clock speed/generation are they? How much memory do they have? What kind of video card do they have and, if a GPU, is it being used in the computations? What kind of interconnect are these laptops using (USB, Ethernet, Wi-Fi, Firewire, External PCI-E, custom, etc.)? Lastly, is there any chance you could convert this value into a more standard unit type such as football fields or Library of Congresses?
 
In the name of clarity, I have a few questions.

Would those laptops have Intel processors, AMD processors, MIPS processors, or some other type? What clock speed/generation are they? How much memory do they have? What kind of video card do they have and, if a GPU, is it being used in the computations? What kind of interconnect are these laptops using (USB, Ethernet, Wi-Fi, Firewire, External PCI-E, custom, etc.)? Lastly, is there any chance you could convert this value into a more standard unit type such as football fields or Library of Congresses?

And why compare it to laptops at all? Seems a strange comparison.
 
sweet so now China will be able to actually come up with their own ideas rather than just steal them from other companies
 
2000 lines of just CUDA code? Sounds like this was rather complex!

2,000 lines is complex? That's nothing. I've written c++ programs 10 times that long in a month that barely do anything.
 
But can it run Crysis?

No, but has enough power to develop a rudimentary consciousness, and that consciousness is not amused by your tired cliche of a comment.

sweet so now China will be able to actually come up with their own ideas rather than just steal them from other companies

Yes, but because IP laws are non-existent in China, some guy living in his parent's basement down the street will steal the data, get a patent on it, apply and receive government loans to create a business. Said business will destroy the original scientists attempts at founding their own firm, and said scientists will either emigrate out of country, or do what most PhD's in China do, which is plagiarize white papers and western studies. When Basement Billionaire is no longer cutting edge, he will turn his firm into a low-wage factory for western tech co's. Things will return to the present.
 
Last edited:
2,000 lines is complex? That's nothing. I've written c++ programs 10 times that long in a month that barely do anything.

lol I was thinking the same. I wrote a mini database system (never actually ended up finishing it, so it's incomplete) that is 1786 lines of code. That's not counting returns and other "white space". All the other apps I checked were well over that.

I manage a game server that is near 1mil lines of code. I did not write all of it mind you but I modified/added/deleted tons of stuff and touched pretty much every part of it at some point.

So if I can do it, a bunch of Chinese monkeys can easily do it too. :D

I wonder how many farming bots this can run at once.
 
Good thing to know you don't know CUDA or the simplest development precepts then. Code complexity is NOT measured by lines.

That was supposed to be my point. I was simply pointing out that I've written much longer programs that do almost nothing, so being 2,000 lines doesn't make it complex by itself. I was making the point that programs can be much longer and still be simple. Perhaps I worded my post poorly and it was unclear?

just make sure your employer doesn't read this forum

Not a problem, programming is a hobby for me, not a profession. And I'm not very good at it....yet.
 
That was supposed to be my point. I was simply pointing out that I've written much longer programs that do almost nothing, so being 2,000 lines doesn't make it complex by itself. I was making the point that programs can be much longer and still be simple. Perhaps I worded my post poorly and it was unclear?

Yes, it was confusing because you didn't stated the argument first.
They are just probably talking about the CUDA section of the code as well. Let's not forget the rest of the program.
 
10 to 1 odds says China will use this thing to try and crack Lockheeds RSA and other defense servers.
 
In the name of clarity, I have a few questions.

Would those laptops have Intel processors, AMD processors, MIPS processors, or some other type? What clock speed/generation are they? How much memory do they have? What kind of video card do they have and, if a GPU, is it being used in the computations? What kind of interconnect are these laptops using (USB, Ethernet, Wi-Fi, Firewire, External PCI-E, custom, etc.)? Lastly, is there any chance you could convert this value into a more standard unit type such as football fields or Library of Congresses?

Rofl, your satire is not lost on me. I find it incredibly frustrating when somebody insults my intelligence with these shitty "comparisons". I can understand standard units of measurement just fine. I don't need someone to put it in "perspective".
 
Reading the title "begins paying off big time" I thought they actually made a useful discovery...
So this is just about bragging?

As for the "2000 lines of CUDA", for those who know it is equivalent* to 1 line of Perl :D

*disregarding performance, readability, sanity and all goodwill towards peers
 
funny how china is always bragging... biggest this and fastest that!
Its like that saying "nice truck, sorry about your dick!" :)
 
This article has some interesting points about how western countries are starting to fall behind in HPC, and what it could mean to future competitiveness (IMO, slow adoption of GPU computing isn't the main problem... emphasis on short term gains is, but meh):

Virtual design, forecasting and simulations are now essential for smarter science, faster innovation and better product development. Which means that High Performance Computing (HPC) is critical to U.S. competitiveness and standards of living.
But the traditional CPU-based technology that once put America in the lead is now the anchor holding us back. Our legacy computing is no longer scaling cost-effectively and power-efficiently enough. The effects of this lost leadership will soon be severely felt in every aspect of American business and economic life unless we decide to do something about it.

10 to 1 odds says China will use this thing to try and crack Lockheeds RSA and other defense servers.
China won't use that one, but could easily build a similar or better supercomputer in secret. LOL, the thing was almost certainly funded by the government anyways.
 
Warning, Nvidia wants to take over the world. tell them to make Crysis or games of similar nature run just as fast the first year of its relase. for them being able to accomplish such a fet, they were very rude about there 8800GTX and Crysis. ATI no better. humm....8800gtx,9800,280,480,580. ahhhh now we can play in all its glory....ill have to send them a Milk man with a candy painted truck.
 
Back
Top