DailyTech - Phase 1 of the ORNL "Jaguar" Supercomputer Complete

Submit News

Science Phase 1 of the ORNL "Jaguar" Supercomputer Complete
Tim Thorpe (Blog) - August 29, 2006 5:14 PM

32 comment(s) - last by exdeath.. on Aug 31 at 2:08 PM

ORNL "Jaguar" has been upgraded to 54 teraflops and will run at 100 teraflops by years end

Oak Ridge National Labs (ORNL), under contract with Cray Inc., has announced that engineers have completed the first phase of upgrades to the labs Cray XT3, named Jaguar. Jaguar now has the capacity to perform at 54 teraflops up from the supercomputers previous peak of 25 teraflops.

The upgrade was part of a multiphase $200M USD contact between the Department of Energy -- which runs ORNL -- and Cray Inc. The first phase included replacing Jaguar’s 5,212 processors with new dual core Opteron parts as well as doubling memory capacity. Engineers also upgraded the high speed interconnects between the machines to increase overall throughput.

In November Cray Inc. and ORNL hope to complete phase two of the project which will increase performance to 100 teraflops. Phase three will further increase Jaguar's performance to a final 250 teraflops by the end of 2007. In 2008 ORNL will install a new supercomputer currently code-named Baker that will be able to perform up One petaflop 4 times faster than Jaguar after all of its upgrades are complete.

Comments

Threshold

Username
Password
remember me

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

Spinning $200M wheels

By The Boston Dangler on 8/29/2006 8:11:31 PM , Rating: 1

The machines are used to estimate the reliability of the US nuclear arsenal, since live-fire tests are a thing of the past. Russian weapons were designed to be servicable and upgradable over time (including disassembly and recycling), just in case Doomsday was delayed. The US lacked such forsight. In fact, the experts in Los Alamos and Oak Ridge have no idea what kind of condition the guts of our bombs are in. All they can do is simulate what 30-40 years of cold and hot standby does to a warhead, which may or may not work right out of the box.

I'd like to see the money spent elsewhere, but those things will never happen. If the money is nukes-only, I'd rather see it go towards making fresh bombs, but that might be prohibited by treaty.

RE: Spinning $200M wheels

By SysAdmin on 8/29/2006 8:19:02 PM , Rating: 5

Actually, this machine is used for open science research, and in general is not related to weapons research. More information on our allocated projects for this year can be found at http://www.nccs.gov/leadership/projects/fy06.html. Weapons research requires a classified facility, which we are not.

Parent

RE: Spinning $200M wheels

By The Boston Dangler on 8/29/2006 8:32:22 PM , Rating: 2

Thank you for clarifying that for me. I guess I wouldn't read about highly classified weapons-testing platforms on DailyTech.

Parent

RE: Spinning $200M wheels

By imaheadcase on 8/29/2006 10:34:15 PM , Rating: 2

About half of the USA nuke weapons are now upgraded. The project to do this started in 1991 it is to be completed in 2010.

Parent

RE: Spinning $200M wheels

By The Boston Dangler on 8/29/2006 10:24:59 PM , Rating: 2

Some of these projects are budgeted millions of processor-hours. Is the entire system devoted to one project at a time, or can some run concurrantly?

Parent

RE: Spinning $200M wheels

By PlasmaBomb on 8/30/2006 10:34:13 AM , Rating: 2

I would love to know how they work processor hours out? Is it a case of one hour on one processor = 1 processor hour? In which case 3 million processor hours on a 5000 processor system would only be 25 days flat out... or it could be divided into smaller lots- say 10 500 processor batches.

Parent

RE: Spinning $200M wheels

By SysAdmin on 8/30/2006 5:24:50 PM , Rating: 2

That's almost how it works, except one processor hour is really one "core" hour with the dual core upgrade. Some jobs span the entire machine, and some are smaller, with several smaller jobs running concurrently.

Parent

RE: Spinning $200M wheels

By SysAdmin on 8/30/2006 5:22:27 PM , Rating: 2

The projects run concurrently. Basically there is a batch system where they submit their jobs, and based on various factors, their jobs get scheduled for some point in the future.

Parent

RE: Spinning $200M wheels

By Desslok on 8/29/2006 8:42:54 PM , Rating: 2

Do you have any links on how the old Soviet bombs are "upgradable"?

ANY warhead can be taken apart, so I am not sure what you are talking about there.

As for "recycling" the guts, I am pretty sure that the Uranium of any bomb can be taken out and used for power generation purposes.

Parent

RE: Spinning $200M wheels

By The Boston Dangler on 8/29/2006 9:20:02 PM , Rating: 2

Sorry, my knowledge of Soviet atomic weaponry is based on many documentaries on Discovery, Disc. Science, et cetrera, and not hands-on in Grozny. The "Trinity and Beyond" series, narrated by Mister William Shatner, is also quite cool.

Many of our bombs use a highly radioactive gas (I forget what), and people far more knowledgable than I have decided it's best not to open the sealed containers unless absolutely neccessary. That's good enough for me.

As for the rest of the device, time has taken it's toll on the many electrical parts, as well as the chemical explosives used to trigger fission. Stability is a major concern, thus the need for constant testing. I'm willing to accept the odds of an accidental thermonuclear detonation are zero, but the chance of a defective bomb fizzling while in the shop is very real.

The plutonium and weapons-grade enriched uranium used in bombs is suitable for use in breeder reactors, of which there are few.

By recycling, I was referring to the parts being reused as-is or with minimal reworking, in a new bomb.

Parent

RE: Spinning $200M wheels

By Desslok on 8/29/2006 9:54:28 PM , Rating: 2

Thanks for the heads up on the documentry, I will check it out.

Parent

RE: Spinning $200M wheels

By PrinceGaz on 8/30/2006 7:26:57 AM , Rating: 2

quote:
The plutonium and weapons-grade enriched uranium used in bombs is suitable for use in breeder reactors, of which there are few.

Well why not sell it to a country that would pay good money for it? I'm sure Iran would be more than happy to buy it for their civilan nuclear programme.

Parent

RE: Spinning $200M wheels

By PlasmaBomb on 8/30/2006 10:25:04 AM , Rating: 2

The gas they use would be tritium or 3H (as in a proton and two neutrons giving atomic mass 3 atomic number 1) which has a half life of about 12.3 years. So if they have been sitting around for nearly 50 years without service there would only be ~25% of the tritium left.

The tritium is used to produce hydrogen bombs, as when the nuclear weapon detonates the intense heat it produces causes the tritium to fuse, releasing yet more energy and improving the destructive potential of the weapon.

Parent

RE: Spinning $200M wheels

By Chernobyl68 on 8/30/2006 11:52:25 AM , Rating: 2

a half life of 12.3 years for 50 years would be closer to 1/16th, not 1/4.

Parent

RE: Spinning $200M wheels

By exdeath on 8/30/2006 2:08:31 PM , Rating: 2

Lol newbs.

The half-life of tritium was taken into account when fusion bombs were first developed.

The primary fuel is lithium deuteride, or LiH where H is the H2 isotope deuterium and while corrosive and very reactive, it is not radioactive or dangerous to humans. Deuterium is more stable and safer.

Neutrons from initial fission, and then created and sustained during fusion chain reaction, spontaneously create the needed tritium on site as needed out of the lithium compound, thereby solving the need to constantly refuel the tritium.

Parent

RE: Spinning $200M wheels

By Chernobyl68 on 8/30/2006 11:49:12 AM , Rating: 2

I wouldn't say "any" bomb, because the same type of fuel most reactors use probably isn't the same type that a bomb uses. weaponized uranium is designed to do one thing well, and fuel for a reactor is designed to do something else well.

Parent

RE: Spinning $200M wheels

By exdeath on 8/30/2006 2:26:16 PM , Rating: 2

It is exactly the same, only the concentration of the unstable U-235 isotope is different.

Natural uranium contains only .7% U-235. Commercial reactors require 3-5% U-235 by mass to be sufficient for power production, while 90%+ is considered weapons grade.

Enriching is the process of using mass dependent means such as centrifuges or gaseous diffusion to separate the uranium by mass in gas form, since U-235 and U-238 have the same chemical signature.

So you can see it takes ALOT of raw uranium and lots of time and equipment to produce weapons grade fuel.

Parent

RE: Spinning $200M wheels

By The Boston Dangler on 8/30/2006 10:42:43 PM , Rating: 2

Would the byproduct be the seemingly large quantity of depleted uranium used in armor, armor piercing rounds, and anything else lead doesn't pass muster for?

Parent

RE: Spinning $200M wheels

By exdeath on 8/31/2006 2:08:03 PM , Rating: 2

Yeah U-238 is the remaining byproduct after *most* of the U-235 has been removed (since when is anything in science exact after all?) And since only .7% of natural uranium ore is U-235... well there’s a lot of left over U-238. Might as well use it for something :P

Useful for kinetic energy weapons because it is harder than lead and because of its higher density (70% denser than lead) allowing use of heavier projectiles of a given size or smaller diameter projectiles of a given mass (focused AP rounds).

U-238 is a relatively long lived and weakly radioactive isotope (4.5 billion year half-life) so the primary danger is that it’s poisonous as a heavy metal. The real source of dangerous radiation exposure is dependant on the trace levels of U-235 remaining after enrichment.

The media attention to DU weapons causing birth defects and medical problems is no more a threat than vaporized lead dust from normal munitions, lead water pipes, or lead paint.

Parent

1 petaflop, dear god!

By Fenixgoon on 8/29/06, Rating: 0

RE: 1 petaflop, dear god!

By brownba on 8/29/2006 8:19:55 PM , Rating: 2

quote:
i can't even imagine what 1 petaflop is in terms of real computing power.

...one quadrillion floating point operations per second

Parent

RE: 1 petaflop, dear god!

By lemonadesoda on 8/29/2006 9:00:15 PM , Rating: 2

oh great! that really helps..

Parent

RE: 1 petaflop, dear god!

By augiem on 8/29/2006 9:27:47 PM , Rating: 2

Doesn't really matter how fast they get. We CG artists will always be able to take a computer to its knees. :P

TOO SLOW!

Parent

RE: 1 petaflop, dear god!

By brownba on 8/29/2006 9:38:27 PM , Rating: 4

interesting you mention that, because I read that DreamWorks wants to have time on this machine to help develop new ray-tracing algorithms

Parent

RE: 1 petaflop, dear god!

By obstreperous on 8/29/2006 9:43:14 PM , Rating: 2

Here here, Give it stacks of ray-traced shadows, reflections, refractions, a bunch of volumetrics, not to mention collision detections, cloth simulations at HD resolution with 30 anti-aliasing passes and this sucker will grind to its knees. Never enough grunt for 3d.

Parent

RE: 1 petaflop, dear god!

By peldor on 8/30/2006 9:34:09 AM , Rating: 2

Or roughly a few % of the computing power being devoted to World of Warcraft at any given time. 5200 dual core opterons are no match for a fully-armed and operational MMORPG!

Parent