Austin (TX) - The Texas Advanced Computing Center (TACC) at the University of Texas at Austin is now running the world’s most powerful supercomputing system for open science research. Ranger is expected provide more than 500 million processor hours of computing time to the science community, performing more than 200,000 years of computational work over its four-year lifetime.  

The new supercomputer went live already on February 4 (about one month later than originally planned), but is made official by the TACC and the National Science Foundation (NSF) on Friday. Ranger achieves a peak performance of 504 TFlops, which makes the world‘s second most capable supercomputer, after system is after IBM’s 596 TFlops BlueGene/L, which is installed at the Lawrence Livermore National Laboratory.

Ranger will be joining NSF’s Teragrid and is expected to ring in the age of Petaflop computing. A performance of 1 PFlops translates into one thousand trillion floating point operations per second.

The system uses Sun’s Constellation platform and integrates 3936 compute nodes. There are 15,744 AMD Opteron processors (62,976 cores) and 1.7 PB of storage in Sun Fire x4500 servers.   

"Ranger is the first of the new “Path to Petascale” systems that NSF provides to open science. It is out in front on the pathway to sustained petascale performance," said Daniel Atkins, director of the NSF’s Office of Cyberinfrastructure, in a prepared statement.

“Ranger provides incredible new capabilities for computational researchers across the nation and world. Its computational power, memory and storage capacity greatly exceed anything the open science community has had access to,” said TACC director Jay Boisseau. “It takes tremendous expertise to deploy and support research on such a system as well as to use it effectively, but it is an awesome honor and responsibility for us at TACC.”

The NSF provided $59 million for Ranger, which covers the system and four years of operating costs.

operating costs?

Feb 21, 2008 20:43     

I remember the college that I graduated from turning down a "free" super computer, because they couldn't afford the operating costs. Its good to see that the funding to build this beast also included the operating cost.
stancilmor report abuse vote down vote up Votes: +0

Barcelona?

Feb 22, 2008 00:59     

Based on the numbers in the article, these are quad core processors. That leads one to believe that these are AMD Barcelona Opterons with the much-reported hardware erratum. Anyone one know the story with this "Ranger" system?
samw report abuse vote down vote up Votes: +0

TLB erratum Linux kernel memory patch

Feb 22, 2008 03:55     

"Ranger was built from 3,936 Constellation blade servers, each containing four quad-core "Barcelona" Opterons running at 2.0 GHz. The 15,744 Barcelona processors in Ranger come from the batch that suffers from the highly publicized translation lookaside buffer (TLB) problem. TACC has used the recommended Linux kernel memory patch to work around that particular problem. Reportedly, the patch has little, if any, impact on performance." http://www.hpcwire.com/hpc/2111550.html further reading: https://www.x86-64.org/pipermail/discuss/2007-December/010260.html

AMD fan report abuse vote down vote up Votes: +0

If you made a simple search....

Feb 22, 2008 04:26     

http://www.tacc.utexas.edu/ser...es/ranger/ Go to General System information... Wasn't that hard, was it?
EJ report abuse vote down vote up Votes: +0

Must be Barcelona

Feb 22, 2008 07:46     

AMD's supercomputer customers are not worried about that bug. They use custome software and also they will not encounter that bug. Also, they would probably use the 1.8 or 1.9GHz proc's. The main reason is low power and initial cost. AMD server cluster are much cheaper to build than Intel's non Hypertransport. This is still AMD's territory - maybe not for long with Nahalem. AMD's advantage is that their system is very mature and tested. There should be more more servers coming out soon using Barcelonas. There are other super computers already running Barcelonas.

Johan Steyn report abuse vote down vote up Votes: +0

First

Feb 22, 2008 07:58     

Sorry, This is the first Barcelona supercomputer: http://news.softpedia.com/news/Ranger-the-Barcelona-Powered-59-Million-Supercomputer-74928.shtml
Johan Steyn report abuse vote down vote up Votes: +0

Re: Barcelona

Feb 22, 2008 08:46     

They could have patched them, but I doubt they bothered. I've never heard of anyone actually encountering that erratum in the wild. AMD probably offered TACC a real bargain for those CPUs to prove the Opteron is still professional class.
Shaun report abuse vote down vote up Votes: +0

One Node's Power Consumption in Watt

Feb 22, 2008 10:34     

Do you know what the each node's power consumption is ?
CoolnQuiet report abuse vote down vote up Votes: +0

RE: If you made a simple search....

Feb 22, 2008 18:38     

I don't need the "just-******-google-it" talk as I spend plenty of time doing basic research on the web. I was busy and figured someone would comment soon and could save me some time. so there.
samw report abuse vote down vote up Votes: +0

samw

Feb 25, 2008 03:57     

You're just a lazy bastard who is unable to make a search... Cheer up, the learning curve is not so hard comes to internet.
EJ report abuse vote down vote up Votes: +0

RE: If you made a simple search...

Feb 25, 2008 10:08     

I agree with samw, the attitude problems are getting bad here. Just help when you can, and keep the gloating to yourself.
Narg report abuse vote down vote up Votes: +0

re: First

Feb 25, 2008 21:04     

certainly not the last Barcelona supercomputer. Kyoto University ordered one from Fujitsu, with 1,664 Barcelona procs, enough to put it in the top 20s (61TFLOPS) http://www.networkworld.com/news/2008/021508-kyoto-university-orders-amd-based.html
Kin report abuse vote down vote up Votes: +0