Print 7 comment(s) - last by KOOLTIME.. on Apr 4 at 10:33 PM

Roadrunner supercomputer is still fast but can't match today's best

One certainty in the technology world is that computer technology advances quickly. What was once of the fastest computers in the world only a few years ago is today well down on the list of the world's top performers. A perfect example is the Roadrunner supercomputer at the Los Alamos National Laboratory. When the supercomputer was installed in 2008, it was the world's fastest and the first supercomputer capable of breaking the petaflop (1 million billion calculations per second) barrier.

Roadrunner is still an incredibly fast supercomputer, but it was decommissioned Sunday. The supercomputer was operational for five years and was used in a wide variety of projects including providing key simulations for the Stockpile Stewardship Program and other unclassified scientific projects.

"Roadrunner exemplified stockpile stewardship: an excellent team integrating complex codes with advanced computing architectures to ensure a safe, secure and effective deterrent," said Chris Deeney, NNSA Assistant Deputy Administrator for Stockpile Stewardship. "Roadrunner and its successes have positioned us well to weather the technology changes on the HPC horizon as we implement stockpile modernization without recourse to underground testing."

Roadrunner uses a hybrid design with 6563 dual-core general-purpose AMD Opteron processors. Each of those processors was linked to a PowerXCell 8i processor, an enhanced version of the chip designed for the Sony PlayStation 3.

One reason that the supercomputer is being decommissioned has to do with the incredible amount of power it utilizes to achieve its impressive performance. While Roadrunner is now sitting around the 20th place on the list of the world's most powerful computers, modern computers use significantly less power for significantly more performance.

IBM had previously predicted supercomputers would reach the 20 petaflop range by 2012.

Source: LANL

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

Just a question...
By Iketh on 4/1/2013 10:47:56 AM , Rating: 2
Why did they use 6000+ dual cores instead of 3000+ quads?

RE: Just a question...
By hpglow on 4/1/2013 11:04:58 AM , Rating: 3
This thing was commissioned in 2008 which means its planning and build phase were 2007. Quad cores were available but were fairly new at the time so there may have been availability issues at the time. Also I think due to its design every Power processor had to be paired with an opteron so even if they had gone with quad cores they still would have had to purchase 6k of them. And for my final broad guess (because I have no idea how this system was designed) I think the Opteron processors were just there to feed data to the PowerXCell processors, they were likely not there to do any of the serious calculations.

RE: Just a question...
By RjBass on 4/1/2013 11:13:02 AM , Rating: 2
Ehh that wasn't how I originally understood it when DT first reported on it, but going back and reading the article, your hypothesis might be true.

RE: Just a question...
By alpha754293 on 4/1/2013 1:43:18 PM , Rating: 3
The Opterons were just (mostly) I/O processors while the Power processors did all the heavy lifting.

Petaflops and the brain
By Milliamp on 4/2/2013 3:12:51 AM , Rating: 2
The most recent top500 list is from November 2012 and can be found here:

This computer was about 1 petaflop and the current top supercomputer is 17.5 Rmax petaflops and 27 Rpeak petaflops.

IBM said in 2012 they believe the human brain can process about 36.8 petaflops of data so the top computer isn't quite there yet but if it continues to roughly double each year it might get there as soon as the end of this year.

There is a poster giving some cool stats and projected growth here:

RE: Petaflops and the brain
By Ktracho on 4/3/2013 6:48:19 PM , Rating: 2
Fortunately, the human brain doesn't require a power generating station to keep it operational. :-)

By KOOLTIME on 4/4/2013 10:33:44 PM , Rating: 2
Most folks dont know the FPU aka math caluclator in AMD's at the time was far superior then intel's these types of super computers use math as a main piece, so its obvious choice.

the calculator part aka FPU is the key component, and at that time frame amd's was far better then intel. Has nothing to do with over all system speed. that one piece is way better, thats the route to go when you need massive mathematical calculations of data.

“Then they pop up and say ‘Hello, surprise! Give us your money or we will shut you down!' Screw them. Seriously, screw them. You can quote me on that.” -- Newegg Chief Legal Officer Lee Cheng referencing patent trolls
Related Articles

Most Popular Articles5 Cases for iPhone 7 and 7 iPhone Plus
September 18, 2016, 10:08 AM
No More Turtlenecks - Try Snakables
September 19, 2016, 7:44 AM
ADHD Diagnosis and Treatment in Children: Problem or Paranoia?
September 19, 2016, 5:30 AM
Walmart may get "Robot Shopping Carts?"
September 17, 2016, 6:01 AM
Automaker Porsche may expand range of Panamera Coupe design.
September 18, 2016, 11:00 AM

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki