China successfully developed "Milky Way One" Supercomputer

Printable View

10-29-2009, 09:28 AM
zalbard

China successfully developed "Milky Way One" Supercomputer

Source: http://english.cctv.com/

Quote:

China unveils its fastest supercomputer. The powerful giant server makes China the second country after the US, to build supercomputers capable of making a quadrillion calculations per second.

http://p5.img.cctvpic.com/program/cc...07688389_r.jpg

The National University of Defense Technology unveiled the supercomputer in Hunan Province. With 7 key innovative technologies, the computer can do more than one quadrillion calculations per second at its peak speed. Experts say the standard of technology in the giant server ranks it the highest of China's 100 supercomputers.

Professor Zhou Xingming from National University of Defense Technology, said, "Simply put, if everyone of the 1.3 billion people that live in China took a digital photograph, this supercomputer could hold all of them."

The high performance server will provide services to users at home and abroad. Scientists say it will be used in various fields, such as biotechnology, aviation, resource survey, remote sensing satellites and forecasting earthquakes.

There is also a video available on the page. No clue how to link it, sorry. :p:
Have a look at this image, though... Took a screenshot from the video.

http://i301.photobucket.com/albums/n...ard/radeon.jpg

I spot something familiar! :D

Quote:

China became the world's second nation after the United States can develop a petaflop supercomputer.

@ http://english.people.com.cn/

Good job guys! :up: This had to happen sooner or later.

Edit: UPDATE!
Source: http://news.xinhuanet.com/
Edit2: UPDATE! :D
Source: http://en.expreview.com/

Quote:

Equipped with 6,144 Intel CPUs (3072 Quad-Core Xeon E5540s and 3072 Quad-Core Xeon E5450) and 5120 AMD GPUs (2560 Radeon HD 4870 X2s), Milky Way One’s peak performance reaches 1.206 petaflops, and it runs at 563.1 teraflops on the Linpack benchmark.

The 155-ton system, with 103 refrigerator-like cabinets lined up on an area of about 1,000 square meters, which cost at least 600 million yuan (88.24 million U.S. dollars) is expected to process seismic data for oil exploration, conduct bio-medical computing and help design aerospace vehicles, according to NUDT president Zhang Yulin.

A single-day task for Milky Way One might take a mainstream dual-core personal computer 160 years to complete, working non-top - if it can last that long.

Zhang said the technical data of Milky Way One had been submitted to the world Top 500 list, compiled by the University of Mannheim, in Germany, the National Energy Research Scientific Computing Center, the Lawrence Berkeley National Laboratory, and the University of Tennessee in the United States.
10-29-2009, 09:33 AM
IvanAndreevich

That's a dual GPU Radeon, judging by the card on top (it's flipped). How do they have access to unreleased hardware on a massive scale?
10-29-2009, 09:35 AM
zalbard

Quote:

Originally Posted by IvanAndreevich

That's a dual GPU Radeon, judging by the card on top (it's flipped). How do they have access to unreleased hardware on a massive scale?

Well, looks like 4870x2 to me.
10-29-2009, 09:38 AM
surfinhicdude

Is it filled with caramel and delicious?
10-29-2009, 09:38 AM
Chumbucket843

Quote:

Originally Posted by zalbard

Well, looks like 4870x2 to me.

why the hell would they use radeons? i think its fake. no one would deploy a massive scale of gpu's and not use the professional series. their is just too much risk.
10-29-2009, 09:39 AM
ajaidev

lolll, i am running MilkyWay@home on 4x5850 with GD70 and X4 955, more power than that blade :P
10-29-2009, 09:39 AM
zalbard

Wonder why they used 4870x2 if GT200 is really better for GPGPU.

Quote:

Originally Posted by Chumbucket843

why the hell would they use radeons? i think its fake. no one would deploy a massive scale of gpu's and not use the professional series. their is just too much risk.

Fake? LOL. It's a live video made by CCTV! :D
10-29-2009, 09:52 AM
zalbard

Updated, for those not believing that it uses AMD GPUs. Also specs in the OP.
10-29-2009, 09:59 AM
annihilat0r

"Equipped with 6,144 Intel CPUs and 5,120 AMD GPUs, Tianhe was able to store all 27 million books in the National Library of China four times over"

the professor who said this thinks that CPUs and GPUs are there for storage? :D
10-29-2009, 10:04 AM
rogueagent6

Quote:

Originally Posted by annihilat0r

"Equipped with 6,144 Intel CPUs and 5,120 AMD GPUs, Tianhe was able to store all 27 million books in the National Library of China four times over"

the professor who said this thinks that CPUs and GPUs are there for storage? :D

This is China we are talking about, they're probably only going to use it to infiltrate more gold farmers into WoW.
10-29-2009, 10:05 AM
Z3R0-CooL

Quote:

Originally Posted by annihilat0r

"Equipped with 6,144 Intel CPUs and 5,120 AMD GPUs, Tianhe was able to store all 27 million books in the National Library of China four times over"

the professor who said this thinks that CPUs and GPUs are there for storage? :D

those where my exact thoughts...
10-29-2009, 10:05 AM
eric66

cool so this is why we have shortage on radeons damn you chinese
10-29-2009, 10:06 AM
annihilat0r

in communist China, supercomputer uses YOU!!!!
10-29-2009, 10:09 AM
Chumbucket843

Quote:

Originally Posted by zalbard

Updated, for those not believing that it uses AMD GPUs. Also specs in the OP.

i believe its using AMD gpu's but not their desktop line. they already lied about this computer saying it is on the petascale. on linpack it only gets 563 teraflops which is the super pi benchie of supercomputers. maybe it has different firmware. its only ranked 5th in energy efficiency. it looks like the cell is still the king of efficiency.
10-29-2009, 10:09 AM
kamieldehond

Quote:

Originally Posted by annihilat0r

"
the professor who said this thinks that CPUs and GPUs are there for storage? :D

:clap:
10-29-2009, 10:09 AM
Sam_oslo

Quote:

Originally Posted by annihilat0r

"Equipped with 6,144 Intel CPUs and 5,120 AMD GPUs, Tianhe was able to store all 27 million books in the National Library of China four times over"

the professor who said this thinks that CPUs and GPUs are there for storage? :D

Well even my 30GB SSD can keep that many books (ok ok almost). I guess the translation is a bit off, maybe he mean keep i "memory".
10-29-2009, 10:11 AM
turtletrax

Quote:

Originally Posted by annihilat0r

in communist China, supercomputer uses YOU!!!!

:ROTF: Classic.
10-29-2009, 10:15 AM
AbelJemka

I also think the traduction is off!
10-29-2009, 10:20 AM
Helloworld_98

And now the Chinese wage war with the US because the 5970 and Intel's Gulftown is about to be released and they've already paid out a load for the current setup.
10-29-2009, 10:24 AM
Sam_oslo

Those 4870x2 wouldn't make any sense for computing power (it would be really crappy compared to NVIDIA CUDA GPGPU), but those "crappy" GPUs have 2GB memory each.

Are they using the memory of GPU for caching?
10-29-2009, 10:26 AM
annihilat0r

more like for storing books from the national chinese library
10-29-2009, 10:27 AM
Helloworld_98

Quote:

Originally Posted by Sam_oslo

Those 4870x2 wouldn't make any sense for computing power (it would be really crappy compared to NVIDIA CUDA GPGPU), but those "crappy" GPUs have 2GB memory each.

Are they using the memory of GPU for caching?

what are you on about?

the programs these guys use would be looking for more teraflops, not ease of programming, the 4870X2 has over 2Tflops, even the GTX 295 has nowhere near that number.
10-29-2009, 10:31 AM
AbelJemka

Quote:

Originally Posted by Sam_oslo

Those 4870x2 wouldn't make any sense for computing power (it would be really crappy compared to NVIDIA CUDA GPGPU), but those "crappy" GPUs have 2GB memory each.

Are they using the memory of GPU for caching?

You know that theorically RV770 has more computing power than G200?
10-29-2009, 10:32 AM
HamidFULL

Quote:

Originally Posted by annihilat0r

in communist China, supercomputer uses YOU!!!!

+1 :up:
10-29-2009, 10:33 AM
zalbard

Quote:

Originally Posted by Chumbucket843

i believe its using AMD gpu's but not their desktop line. they already lied about this computer saying it is on the petascale. on linpack it only gets 563 teraflops which is the super pi benchie of supercomputers. maybe it has different firmware. its only ranked 5th in energy efficiency. it looks like the cell is still the king of efficiency.

Maybe cause GPUs don't work in Linpack? I don't know, honestly.
Or maybe they listed peak performance, which is always higher than what Linpack shows us.
10-29-2009, 10:36 AM
Chumbucket843

Quote:

Originally Posted by AbelJemka

You know that theorically RV770 has more computing power than G200?

you know that in the real world a 600 gigaflop gtx280 beats a 1.2teraflop 4870. comparing flops across architectures is almost as bad as comparing clock speed. ATi has a "markitecture" just like intel did.:up:
10-29-2009, 10:39 AM
annihilat0r

"in the real world" doesn't mean "in Crysis" for the Chinese supercomputer users, though
10-29-2009, 10:40 AM
Jamesrt2004

Quote:

Originally Posted by zalbard

Wonder why they used 4870x2 if GT200 is really better for GPGPU.

Fake? LOL. It's a live video made by CCTV! :D

mm ati have better gpgpu perf actually check dp benchmarks not nooby stuff like f@h
10-29-2009, 10:46 AM
Sam_oslo

Quote:

Originally Posted by Helloworld_98

what are you on about?

the programs these guys use would be looking for more teraflops, not ease of programming, the 4870X2 has over 2Tflops, even the GTX 295 has nowhere near that number.

Quote:

Originally Posted by AbelJemka

You know that theorically RV770 has more computing power than G200?

You are both right, but the API for converting that power to computing power (GPGPU) is still "crappy". CUDA would do a better job, for sure. Those 2Tflops is good for games, but can't get converted to computing power easily, is my point.

But 4870x2 got 2GB of really fast memory, so i was wondering if they are using the memory as some kind of caching, and maybe that's why he said ""Equipped with 6,144 Intel CPUs and 5,120 AMD GPUs, Tianhe was able to store all 27 million books in the National Library of China four times over", meaning that they are keeping this much info in cache ?
10-29-2009, 10:46 AM
AbelJemka

Quote:

Originally Posted by Chumbucket843

you know that in the real world a 600 gigaflop gtx280 beats a 1.2teraflop 4870. comparing flops across architectures is almost as bad as comparing clock speed. ATi has a "markitecture" just like intel did.:up:

I know that a GTX280 has 933 gigaflop of computing power (yours numbers us "markitecture"? :rolleyes:)
I also know that without good software the best hardware is pointless.
Nvidia has a good software (CUDA).
Apart from this I know nothing :shrug:

Quote:

Originally Posted by Sam_oslo

You are both right, but the API for converting that power to computing power (GPGPU) is still "crappy". CUDA would do a better job, for sure. Those 2Tflops is good for games, but can't get converted to computing power easily, is my point.

+1 you posted before me.
10-29-2009, 11:06 AM
rogueagent6

Quote:

Originally Posted by Chumbucket843

you know that in the real world a 600 gigaflop gtx280 beats a 1.2teraflop 4870. comparing flops across architectures is almost as bad as comparing clock speed. ATi has a "markitecture" just like intel did.:up:

I'm sure China took ample time evaluating the offerings from both vendors. If Nvidia's solution is so much better than ATI's, then why didn't China go with Nvidia?
10-29-2009, 11:14 AM
hurleybird

Quote:

Originally Posted by annihilat0r

"in the real world" doesn't mean "in Crysis" for the Chinese supercomputer users, though

Just to add to that, GTX 280 blows at double precision, while the 4870 doesn't. Obviously both are worse than 5870 and especially Fermi whenever it comes out.
10-29-2009, 12:05 PM
eric66

Quote:

Originally Posted by Chumbucket843

you know that in the real world a 600 gigaflop gtx280 beats a 1.2teraflop 4870. comparing flops across architectures is almost as bad as comparing clock speed. ATi has a "markitecture" just like intel did.:up:

lol real world do you mean games or something :ROTF: there are no nvidia sponsored titles in real world :rofl:
10-29-2009, 12:11 PM
Splave

yeah but can it play crysis

to quote the pic
"wow look at all those harddrives guyz" ;)
10-29-2009, 12:14 PM
annihilat0r

Five thousand 4870x2's... I think Crysis was created with this computer in mind.
10-29-2009, 12:52 PM
Chumbucket843

Quote:

Originally Posted by Jamesrt2004

mm ati have better gpgpu perf actually check dp benchmarks not nooby stuff like f@h

i would say its dependent on the application. video encoding works on ATi hardware much better but good software isnt really there yet. ATi certainly has a unique advantage in double precision in that it can pair 2 single precision units into one DP unit. they dont waste any die space doing that. nvidia cant do that. they have to dedicate die space for DP.

the current f@h ATi client has been tainted by r600. the entire ATi implementation is a mess. the two main theings holding it back is the fact there is no cache in r600 and rv770 and up have cache but its not used. the speed would easily double if they utilized it. the other bottleneck is the cpu. it has to be used for random number generation. this is why the 5870 has the same ppd ad the 4870. but dont worry gpu3 will fix all of those problems. it uses opencl too. i would expect improvements in the nvidia client too. i would not blame all of the problems on stanford molecular dynamics is really complicated and nvidia is taking a smart approach and helping programmers with libraries and optimizations. ATi has not really done this well. the reason why companies in the past with massively parallel architectures failed is usually because of software.
10-29-2009, 01:14 PM
Jamesrt2004

Quote:

Originally Posted by Chumbucket843

i would say its dependent on the application. video encoding works on ATi hardware much better but good software isnt really there yet. ATi certainly has a unique advantage in double precision in that it can pair 2 single precision units into one DP unit. they dont waste any die space doing that. nvidia cant do that. they have to dedicate die space for DP.

the current f@h ATi client has been tainted by r600. the entire ATi implementation is a mess. the two main theings holding it back is the fact there is no cache in r600 and rv770 and up have cache but its not used. the speed would easily double if they utilized it. the other bottleneck is the cpu. it has to be used for random number generation. this is why the 5870 has the same ppd ad the 4870. but dont worry gpu3 will fix all of those problems. it uses opencl too. i would expect improvements in the nvidia client too. i would not blame all of the problems on stanford molecular dynamics is really complicated and nvidia is taking a smart approach and helping programmers with libraries and optimizations. ATi has not really done this well. the reason why companies in the past with massively parallel architectures failed is usually because of software.

yeah it's stupid, they both went the wrong way, at i could do with good software, and nvidia the hardware :p: (atleast till fermi's out ;)..)

gpu3 sounds interesting, and actually usefull :up: im sure a lot of people nvidia and ati folders a like are waiting for it :)
10-29-2009, 01:16 PM
blob

Quote:

Originally Posted by Chumbucket843

i would say its dependent on the application. video encoding works on ATi hardware much better but good software isnt really there yet. ATi certainly has a unique advantage in double precision in that it can pair 2 single precision units into one DP unit. they dont waste any die space doing that. nvidia cant do that. they have to dedicate die space for DP.

the current f@h ATi client has been tainted by r600. the entire ATi implementation is a mess. the two main theings holding it back is the fact there is no cache in r600 and rv770 and up have cache but its not used. the speed would easily double if they utilized it. the other bottleneck is the cpu. it has to be used for random number generation. this is why the 5870 has the same ppd ad the 4870. but dont worry gpu3 will fix all of those problems. it uses opencl too. i would expect improvements in the nvidia client too. i would not blame all of the problems on stanford molecular dynamics is really complicated and nvidia is taking a smart approach and helping programmers with libraries and optimizations. ATi has not really done this well. the reason why companies in the past with massively parallel architectures failed is usually because of software.

Maybe you should checkout Milkyway@home and Collatz Conjecture

Quote by Andreas Przystawik, programmer who optimized the Milkyway@Home client for GPU Computing:
I have no idea what Folding is doing in its code. But you may want to look at two other distributed computing projects using ATI cards: Milkyway@home and Collatz Conjecture.

Both scale virtually perfectly with the higher shader count. That a HD5870 achieves this with the Collatz Conjecture project is also showing that ATI has done a tremendeous job with the memory controller efficiency as it is quite bandwidth hungry.

And I guess I don't have to tell you how pathetic a GTX285 looks compared to a HD5870 at the Milkyway project which uses double precision calculations (it barely matches the performance of an old and rusty HD3850). Nvidia badly needs a factor 8 improvement here to even approach a Cypress, let alone to pass it. And by the way, an nvidia engineer helped the project with its CUDA application, I guess one can assume it's near the maximum performance the hardware is capable of.
I have high hopes for Fermi, but frankly I still doubt a bit it will pass Cypress on pure number crunching.

Oh, by the way, both projects use the Brook+ layer as a base for the ATI applications. So that has nothing to do with the possible scaling.
10-29-2009, 01:59 PM
WhiteFireDragon

Quote:

Professor Zhou Xingming from National University of Defense Technology, said, "Simply put, if everyone of the 1.3 billion people that live in China took a digital photograph, this supercomputer could hold all of them."

i'm questioning if this guy is really qualified to be a professor. my netbook can do this as long as i have enough external storage drives, you don't need a super computer. he doesn't even know the difference between processing power and storage space :shakes:
10-29-2009, 02:01 PM
Sam_oslo

LOL, i hope this is not becoming a ATI and Nvidia fight (again), LOL.

Anyways this is a really strange supercomputer with that many CPU and GPUs, but not a word about the amount of RAM.
10-29-2009, 03:20 PM
haylui

TianHe (天河), I don't understand why the official translation to be Milky Way.
Milky Way should be called 银河 rather than 天河....a huge difference...lolz
10-29-2009, 03:20 PM
Spawne32

and everyone within 100 miles has lead poisoning :rofl:
10-29-2009, 03:27 PM
annihilat0r

Quote:

Originally Posted by Spawne32

and everyone within 100 miles has lead poisoning :rofl:

this totally cracked me up :rofl::rofl::rofl::rofl:
10-29-2009, 03:28 PM
Serra

Aside from apparently storing a lot of data on the processors (?) does anyone know what they plan on using it for?
10-29-2009, 03:41 PM
570091D

Quote:

Originally Posted by Serra

Aside from apparently storing a lot of data on the processors (?) does anyone know what they plan on using it for?

filtering pr0n
10-29-2009, 03:54 PM
eric66

Quote:

Originally Posted by 570091D

filtering pr0n

thats what japanese do :p:
10-29-2009, 03:55 PM
ownage

Quote:

Originally Posted by AbelJemka

You know that theorically RV770 has more computing power than G200?

Yes and thats what allot of nubs don't know. People don't blind yourself with words like GPGPU, Cuda, OpenCL or anything like that. When the software is right, RV770 can be faster, and thats already proven by expensive decryption software that cant be compared with games or Cuda crap atm.
10-29-2009, 04:20 PM
apt403

Quote:

Originally Posted by WhiteFireDragon

i'm questioning if this guy is really qualified to be a professor. my netbook can do this as long as i have enough external storage drives, you don't need a super computer. he doesn't even know the difference between processing power and storage space :shakes:

I'm not sure he can even do basic math...

Working off the assumption that he's slightly mentally deficient and talking about total cache of the system.

Best case senario, the CPUs have 8MB L3, 1MB of L2, and 256KB of L1 (Standard quad-nehalem specs), for a total of 9472KB of cache per CPU. Assuming there are 6144 of these in the system, that's 58,195,968KB of total CPU cache, or about 55.5GB. With 5210 GPUs at 2GB of cache a piece, that's 10420GB of GPU cache, and a total cache of 10475.5GB/~10.2TB for the entire system.

With 10.2TB of space, if you wanted to store 1.3 billion pictures, they'd need to be on the order of 8KB each. I'm not aware of the image quality of cameras in China, but they've either got some damn fine compression, or have a max res of about 100x100 to take pictures that only take up 8KB of space...
10-29-2009, 04:28 PM
Utnorris

Quote:

Originally Posted by Serra

Aside from apparently storing a lot of data on the processors (?) does anyone know what they plan on using it for?

Hacking the U.S. Super Computer.
10-29-2009, 04:30 PM
Hu1kamania

Quote:

Originally Posted by Utnorris

Hacking the U.S. Super Computer.

lol yeah, but I think the US one will win!
10-29-2009, 04:43 PM
annihilat0r

Quote:

Originally Posted by Hu1kamania

lol yeah, but I think the US one will win!

If Fermi powers that one:

- Sir, we have a problem. the 5000 Fermi videocards we bought for our supercomputer...
- ...... yes? What's the problem?
- Wooden scre-
- DAMN YOU JENSENNNNNNNNNNNNNNnnnnnnnnnnnnnnnnnnn:::::::.:::: :...:..... .... . . (slowly disappearing scream)
10-29-2009, 04:44 PM
Hu1kamania

Quote:

Originally Posted by annihilat0r

If Fermi powers that one:

- Sir, we have a problem. the 5000 Fermi videocards we bought for our supercomputer...
- ...... yes? What's the problem?
- Wooden scre-
- DAMN YOU JENSENNNNNNNNNNNNNNnnnnnnnnnnnnnnnnnnn:::::::.:::: :...:..... .... . . (slowly disappearing scream)

What do you have against wooden screws? Thats just nvidias way of going green.
10-29-2009, 04:45 PM
Speederlander

Quote:

Originally Posted by Spawne32

and everyone within 100 miles has lead poisoning :rofl:

Exactly.
10-29-2009, 04:47 PM
Speederlander

Quote:

Originally Posted by Utnorris

Hacking the U.S. Super Computer.

So U.S. hardware hacking U.S. hardware. Kind of ironic.
10-29-2009, 04:59 PM
twwen2

Quote:

Originally Posted by Utnorris

Hacking the U.S. Super Computer.

Maybe not the US computer, but US encryption perhaps. Digital Fortress anyone?
10-29-2009, 05:11 PM
Buckeye

Quote:

Originally Posted by rogueagent6

This is China we are talking about, they're probably only going to use it to infiltrate more gold farmers into WoW.

The new rage is Aion, man they are every where. Login with a new character and get slamed with gold seller spam. Bots and sellers non stop.
10-29-2009, 06:17 PM
Serpentarius

what the point for such server farms to have massive GPUs running for?

BIG Bro observation?
10-29-2009, 06:27 PM
Spawne32

i was talking to a friend last night about enermax's chinese website, its hilarious how every product they have has to be an "anti-us" product :rofl:
10-29-2009, 07:03 PM
eva2000

woah crazy.. imagine if they changed the 4870x2 to 5970 gpus... but imagine having to physically swap all 5,120 gpus out :eek:
10-29-2009, 08:22 PM
duckbillclinton

Quote:

Originally Posted by Serra

Aside from apparently storing a lot of data on the processors (?) does anyone know what they plan on using it for?

It will be used for national weather forecast center. The setup uses Intel Xeon processor, 4870x2, and a specially designed OS. According to some reports, when using NV's GPU during early stage testing, 4870x2 only trailed behind for 2% in efficiency in certain test, and others pull ahead. Since 4870x2 is cheaper, we all know what the outcome would be.
10-29-2009, 08:27 PM
Splave

Quote:

Originally Posted by eva2000

woah crazy.. imagine if they changed the 4870x2 to 5970 gpus... but imagine having to physically swap all 5,120 gpus out :eek:

Im sure someone there is willing to do it for .05 an hour ;)

I think bill gates' desktop is of similar performance no? :rofl:
10-30-2009, 12:43 AM
Sam_oslo

Quote:

Originally Posted by duckbillclinton

It will be used for national weather forecast center. The setup uses Intel Xeon processor, 4870x2, and a specially designed OS. According to some reports, when using NV's GPU during early stage testing, 4870x2 only trailed behind for 2% in efficiency in certain test, and others pull ahead. Since 4870x2 is cheaper, we all know what the outcome would be.

Do you have a link, or more details about this "specially designed OS"?

Do I see it right?, is it 2 CPU + 2 GPU on each blade in this picture?
http://i301.photobucket.com/albums/n...ard/radeon.jpg
10-30-2009, 01:07 AM
eva2000

looks like quad cpu + 2x hd4870x2
10-30-2009, 01:54 AM
mursaat

And 16 DIMMs per plate. So they are using Brook+, that makes it almost impossible to switch to Nvidia hardware. Why didn't they use OpenCL and program the whole thing from scratch? I assumed that should be the case for mainframes like this. Two years in development and they use the standard API? With so many people they only had one guy assembling the cards? :rofl:

Quote:

Originally Posted by the article

"As far as I know, a combination of CPU and GPU is something new used to make a petaflop computer. A GPU, or graphic processing unit, plays a role as an accelerator to make the computer run faster, but reduces its power consumption and cost," Zhou explained.

I'm not sure on the figures about processing power against power consumption and price. Is a GPU better for crunching in power consumption too?
10-30-2009, 03:41 AM
Sam_oslo

Quote:

Originally Posted by mursaat

And 16 DIMMs per plate. So they are using Brook+, that makes it almost impossible to switch to Nvidia hardware. Why didn't they use OpenCL and program the whole thing from scratch? I assumed that should be the case for mainframes like this. Two years in development and they use the standard API? With so many people they only had one guy assembling the cards? :rofl:
I'm not sure on the figures about processing power against power consumption and price. Is a GPU better for crunching in power consumption too?

The idea of using the GPU as a "accelerator to make the computer run faster" is not new. But then as you mentioned, it would be more natural to use OpenCL (or other proven and standard API), but they have chosen to use a "specially designed OS".

This approach is totally "off" (from the "normal" approach for using the GPU as accelerator). If I'm not mistaking, it seams they are not using those GPUs for delivering processing power (GPGPU), at least not as their main function. And it makes me wonder if they are using those 2GB fast GPU-memory as some kind of cache?. I guess the answer (which is really interesting) lies in what is going on with that "specially designed OS".
10-30-2009, 04:29 AM
mk_dir

Quote:

Originally Posted by twwen2

Maybe not the US computer, but US encryption perhaps. Digital Fortress anyone?

Yeah, AES be dare xd
11-02-2009, 11:27 AM
zalbard

I updated the OP with some info about processors and graphics cards used, and also performance, cooling used, and its computing purposes.
It is confirmed that it does use 4870x2.
11-02-2009, 01:40 PM
Chumbucket843

why are they saying it is a petascale supercomputer? it only gets 560 teraflops in linpack.
11-02-2009, 01:42 PM
Solus Corvus

Quote:

Originally Posted by Chumbucket843

why are they saying it is a petascale supercomputer? it only gets 560 teraflops in linpack.

Can you run linpack on GPUs?

Didn't think so.
11-02-2009, 02:02 PM
Chumbucket843

Quote:

Originally Posted by Solus Corvus

Can you run linpack on GPUs?

Didn't think so.

as far as i am concerned it doesnt matter that the gpu's cant be used. you either can run the benchmark or you cant.
11-02-2009, 02:03 PM
Solus Corvus

They didn't build a supercomputer to benchmark linpack all day. So even if it couldn't run linpack at all, your point is still moot.
11-02-2009, 02:25 PM
Chumbucket843

Quote:

Originally Posted by Solus Corvus

They didn't build a supercomputer to benchmark linpack all day. So even if it couldn't run linpack at all, your point is still moot.

no one builds a supercomputer to run linpack. linpack is blas so its going to reach peak flops better than anything else and they arent getting in the top500 with out it. you dont know if it will be faster running the software for its intended purpose so how can you claim it will reach over a petaflop?
11-02-2009, 02:40 PM
Solus Corvus

I'm not the one claiming a petaflop, the designers are. And they obviously aren't claiming over a petaflop in linpack. They designed it to be over a petaflop for a specific usage. Does it actually reach over a petaflop in that application? I don't know and I'm willing to bet that you don't either. But if it does then it's definitely a petaflop supercomputer, regardless of whatever it scores in linpack.

You seem like a smart person so I'm kind of curious why you are so hung up on the linpack thing. Linpack isn't the only measure of flops in the entire universe. Sure, linpack is standard and the score won't get them into the top500. But they didn't build it to benchmark linpack or to get into the top500 - they built it to perform a certain type of computation and how good it is at that is the important thing.
11-02-2009, 05:59 PM
Gen

Quote:

Originally Posted by annihilat0r

If Fermi powers that one:

- Sir, we have a problem. the 5000 Fermi videocards we bought for our supercomputer...
- ...... yes? What's the problem?
- Wooden scre-
- DAMN YOU JENSENNNNNNNNNNNNNNnnnnnnnnnnnnnnnnnnn:::::::.:::: :...:..... .... . . (slowly disappearing scream)

ahaha :ROTF:

it will probably be used in a one time winner take all supercomputer crysis tournament, Fatal1ty vs. chinese gold farm.. or whoever is their best.

on topic though i'm sure if this can help with early warning weather systems then that's great.
11-02-2009, 06:03 PM
saaya

Quote:

the computer can do more than one quadrillion calculations per second at its peak speed

1.000.000.000.000.000 so in other words 1 peta flop... impressive... :D

so while nvidia talks gpgpu, ati makes it happen? :D
im curious though, how are they going to use the gpus?
open cl?
11-02-2009, 06:17 PM
Mungri

But can it run crysis?

Why dont they run 3D mark on that beast and hold the No.1 spot forever and ever?

I want benchmarks dammit!

And it had better be folding.

And can I go use it to play a game of overclock please? I'll try my hardest not to break it, promise.
11-02-2009, 08:16 PM
gumballguy

Quote:

Originally Posted by bhavv

But can it run crysis?

WTF is with the resurrection of that comment?

It's become stupidly popular again.

TIME. TO. DIE.

Not funny. Give it up. I'm looking at you *glares at all*
11-02-2009, 10:14 PM
lkiller123

Quote:

Originally Posted by Chumbucket843

why the hell would they use radeons? i think its fake. no one would deploy a massive scale of gpu's and not use the professional series. their is just too much risk.

Well, they are more than 50% cheaper and calculates the same (in terms of flops). Why not?
11-03-2009, 12:54 AM
Origin_Unknown

the real questions are;

1) can it run WCG? and why isnt it?
2) is movieman's personal server fam more powerful? i think so!
11-03-2009, 11:51 AM
Power5

Would be sweet to see a 3d06 or vantage score, but I would guess it is not possible to run on this thing without some serious rework to the program code.

Crysis saying may be dead to you, but I still think that exact same thing every time I hear about some new super tech.
11-03-2009, 02:59 PM
Chumbucket843

Quote:

Originally Posted by saaya

1.000.000.000.000.000 so in other words 1 peta flop... impressive... :D

so while nvidia talks gpgpu, ati makes it happen? :D
im curious though, how are they going to use the gpus?
open cl?

the gpu's are ~3 petaflops peak, probably getting ~450 teraflops running linpack b/c double precision. this would be implemented through ATi stream or whatever they are calling it now. nvidia tesla is in supercomputers too. they actually designed server racks with 4 gpu's. here is a video of tsubame.
http://www.computerworld.com.au/video/740796303
11-03-2009, 04:46 PM
revenant

soooo... I guess they're going to get right to the task of generating the largest prime number... weeee!
11-03-2009, 08:44 PM
lkiller123

Well, actually I think they should've used Nvidia GPUs instead of Radeons for the GPGPU purpose. I've heard that the CUDA technology had been way more mature than ATo Stream.

Was it probably because ATi sponsered this project?
11-10-2009, 08:02 PM
sniper_sung

Quote:

Originally Posted by WhiteFireDragon

i'm questioning if this guy is really qualified to be a professor. my netbook can do this as long as i have enough external storage drives, you don't need a super computer. he doesn't even know the difference between processing power and storage space :shakes:

I think you probably didn't calculate the storage space required for these photos. Suppose each photo occupies 1 MB, it takes 1.5 PB to store all these. Typically it takes a classroom to hold hard drives adding up to 1.5 PB today.

Secondary-storage space is another important performance index of a supercomputer, further to computation power. For example, to simulate ion diffusion in a single human heart, for 100 mm spacing we have 4.23x10^8 computational points. Storing 1 minute of simulation at 1 ms time steps requires 5.4 PB disk space.

I think it's pointless to mention your netbook here, as it's not relevant.
11-10-2009, 08:08 PM
cegras

They might be giving numbers so people can put things into perspective, although not the most relevant ones.
11-10-2009, 08:30 PM
LordEC911

Quote:

Originally Posted by lkiller123

Well, actually I think they should've used Nvidia GPUs instead of Radeons for the GPGPU purpose. I've heard that the CUDA technology had been way more mature than ATo Stream.

Was it probably because ATi sponsered this project?

Because if you actually code for AMD/ATi's architecture their cards are MUCH faster than Nvidia cards.
11-10-2009, 08:56 PM
MirageSys

This is suspect.

"Milky-way". You know Chinese milk is tainted with Melamine.....
/sarcasm

Last I checked, ATI architecture lays the beat down on current NVIDIA architecture when it comes to DP calculations. Why are people so surprised its being used?

All times are GMT -8. The time now is 01:50 AM.

XtremeSystems