- Tuesday, 9 February 2010
- INQ Mobile
- RSS
Politics is more dangerous than war, for in war you are only killed once - Winston Churchill
Supercomputer uses a GPU cluster
THE AUSSIE Commonwealth Scientific and Industrial Research Organisation (CSIRO) has just knocked together a supercomputer that's based around a cluster of graphics processing units (GPUs).
Using GPUs in this way gives the computer a processing processing capacity that competes with supercomputers over twice its size.
The as yet unnamed CSIRO system is one of the first supercomputers to combine traditional CPUs with powerful GPUs. The fifth fastest supercomputer on the latest Top 500 supercomputers list includes ATI Radeon HD4870-2 GPUs.
The CSIRO machine has 28 Dual Xeon E5462 Compute Nodes for a total of 1024 2.8GHz compute cores. It has 500GB of SATA hard disk storage and DDR InfiniBand interconnects.
The system also has 64 Tesla S1070 modules which means 256 GPUs with a total of 61,440 streaming processor cores.
It can manage 200+ TeraFLOPS (TF), more than the 140 TF capable supercomputer announced by the Australian National University last week, although that is a 64-bit machine while the CSIRO machine is a 32-bit machine.
The CSIRO machine's GPUs can deliver up to 4.14 TF of single-precision floating-point performance and 345 GigaFLOPS of double-precision floating-point performance for each Tesla on board.
They can be accessed with Nvidia's CUDA parallel computing architecture or by using compiler technology released by the Portland Group .
CSIRO science applications have already seen one to two orders of magnitude increases of speed by using Nvidia GPUs. µ
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
flooding @ home ;)
Or for Crysis so you can break the 60FPS barrier!
Would love to see this thing churn through RC5-72 blocks. :)
Glad to see Aussies are so keen on using antiquated proprietary solutions. Perhaps their next step will be using Apple IIs.
"500GB" ?
Yeah, but programming it is a bitch. The machine keeps fragging your code.
What is the news in this?
Was this for Balance to the HD4K GPU cluster being the first GPU based solution breaking the top ten (top 5) for the first time?
Because as such it's lame since the computer mentioned wouldn't break the top 15, and it's far from the first nVidia solution either as there are MANY peppered in the top 500, starting with # 56 TSUBAME;
http://www.top500.org/system/9883
which has 21600 GB (which seems more appropriate having 20TB) and is already IN the list, not a future contender.
There is no news in this other than it's a non-asian cluster, not that it's either GPU based nor an nVidia solution. Makes me question the relevance to the overall picture other than that.
Anyone else notice but you mention the supercomputer has 4870x2's and yet your article title talks about the 'green goblin', aka NVIDIA. You then mention CUDA, which as far as I know is NVIDIA technology not ATI. Jeez - a tech reporter should really proof read 1st, and a tech editor should have sniffed out those mistakes.
IS this story a load of 'ole bollox, or are is there a real story with correct facts behind it ? ...Mageek would be choking on his pint reading this twaddle.
@ Creesh
it does use nvidia GPU's. When he mentions the radeon cards hes talking about another supercomputer that uses GPUs aswell as CPU's.
"The system also has 64 Tesla S1070 modules"
TESLA are nvidia GPU's :)
That's got to mean one 500GB sata drive per node...
also, @Creesh: You're an idiot, the computer using ATI gpu's is a completely different machine than the one the article is about. Learn how to read.
Well, in a HPC, the compute nodes, and the storage nodes are always separate, however, since they are using infiniband interconnect and all, why create a bottleneck by using very slower SATA HDDs?
Perhaps, the amount of collective RAM in the nodes is massive, and the procession data overall is loaded into the RAM for faster access......yet, im just wondering that........