back to article Uncle Sam's nuke-stockpile-simulating souped-super El Capitan set to hit TWO exa-FLOPS, take crown as world's fastest machine in 2023

The US Department of Energy has revealed a few more details about the supercomputer it has commissioned to simulate and study America's stockpile of nuclear weapons. This is the machine that is set to be the world's fastest publicly known super, and is expected to be 10 times more powerful than today's biggest beasts. Aptly …

  1. IGotOut Silver badge


    "will unlock solutions to society's most complex issues and answer questions we never thought were possible."

    Knowing if they can destroy the world* 10x over or just 5x over is right up there on my list of things I need to know.

    Still,nice to see it being rented out to one of the richest companies in the world to help them make drugs many won't be able to afford.

    *Yes I know all the nukes on the planet will not actually destroy the Earth.

    1. bob, mon!

      Re: Uggghh

      Better to simulate them, instead of setting them off for real.

    2. Anonymous Coward
      Anonymous Coward

      Re: Uggghh

      How about if the military donates the processor time to GSK, with the catch that anyone in need gets their critical drugs for free. Not just the ones developed on El Capitan (otherwise the best drugs will coincidentally have been developed on other systems) but all of GSK' lifesaving drugs. Go ahead and sell your ED meds at high margins, but cancer drugs, etc.

      The argument is that drugs are expensive to create, so they need to be priced high to recoup the cost. Well, some free time on the world's biggest supercomputer (x600) should knock off a bunch of development costs.

      1. GrumpyKiwi

        Re: Uggghh

        ***The argument is that drugs are expensive to create, so they need to be priced high to recoup the cost. Well, some free time on the world's biggest supercomputer (x600) should knock off a bunch of development costs.***

        The biggest chunk of costs are getting FDA approval - something that can take up to 10 years. No amount of super/ultra/mega-computing time will cut that type of paperwork.

      2. robidy

        Re: Uggghh

        It's $600 not 600 time faster...regardless, that's a pretty cool idea...would widen it to other drug companies and allow startup in test to keep say half profits.

      3. Chemist

        Re: Uggghh

        "Well, some free time on the world's biggest supercomputer (x600) should knock off a bunch of development costs."

        Well I take your point but actually the R cost (high though it is ) is utterly dwarfed by the D cost. And, no, the computer will not have any great impact on the D cost

      4. Tigra 07

        Re: Uggghh

        That's certainly one argument. The other is that these companies spend billions in research to create the drugs, but then only get around 6 years to charge what they want and recoup that money. After that timeframe any company can create the drugs cheaper as the patent expires.

        If you charge a bargain price while your patent is active, you won't recoup costs and will ultimately collapse. I'm not excusing greed from some of these companies, just pointing out the obvious business model.

    3. PhilipN Silver badge

      New drugs

      Taking the GSK connection at face value, and with optimism, for the sake of humanity.

      But it still does not negate the fact that the prime mover for this and much of “progress” generally is and has always been militaristic. Greek fire. Gunpowder. Dynamite .....


    4. Alan Brown Silver badge

      Re: Uggghh

      Just to set your mind at rest a little:

      Oak Ridge is ALSO doing a shedload of research on restarting molten salt fuelled nuclear reactors (The Oak Ridge Experiment) and because MSRs are effectively banned in the USA (you can play with molten salts to your heart's content, but adding anything nuclear to them was explicitly made impossible when Nixon killed off Weinberg's safer mousetrap in 1972(*)), this is about the only way to model the best/safest/most efficient ways of setting up core geometry.

      Bear in mind that whilst Thorium LFTR tech is reasonably well understood and not pressurised, the Oak Ridge Experiment used a graphite core - which when push comes to shove is a fire hazard if things go really _really_ wrong. That's highly unlikely in a MSR because you can and do flood the reactor with inert gas if you drain it, but after Chernobyl still a kneejerk response item - and means that less flammable alternatives need to be tested and modelled to provide the necessary criticality.

      (*) Standard water moderated nukes are a bloody great steam bomb waiting to go off - Alvin Weinberg invented the water-moderated nuclear reactor for the Nautilus and didn't like his small design being scaled up to Heath-Robinson scales for civil power as he felt it was dangerous, so built a "better mousetrap" in the form of a hotter, much safer, unpressurised reactor.nuclear loop that couldn't burn or contaminate the biosphere. His reward was to be kicked out of the nuclear industry - and the project that Nixon promoted in favour of the MSR (the fast breeder reactor) was a dud.(**)

      (**) Bad things to cool or moderate nuclear reactors with:

      - water - steam bombs, not hot enough, corrosive and biosphere contamination

      - sodium - burns furiously when exposed to air (Monju)

      - lead - rots brains (some soviet sub reactors and a few others)

      - Helium/CO2/other gases - containment issues

    5. Anonymous Coward

      Re: Uggghh

      Knowing if they can destroy the world* 10x over or just 5x over is right up there on my list of things I need to know.

      Knowing that without having to run an experimental test is indeed something you probably should want to know, since the experimental test involves everyone dying (or at least a significant possibility of that).

  2. Anonymous Coward
    Anonymous Coward

    But will it play Crysis ?

    1. Ima Ballsy

      ..OR ....

      Fortnight ....

      1. Anonymous Coward

        Re: ..OR ....

        Yes, but it's so fast it will play Fortnight in under 2 days.

  3. Snake Silver badge

    That article title

    would be far easier to read with a comma after "stockpile-simulating", FWIW

    1. diodesign (Written by Reg staff) Silver badge

      Re: That article title

      Commas are for wimps, and multimillionaires.


      1. Tom 7

        Re: That article title


      2. Snake Silver badge

        Re: That article title

        Lol. The new hyphens work perfectly, though.

  4. Marcelo Rodrigues

    This computer, whose name I'm not worthy of saying

    Will be the biggest and greatest ever, and will calculate the ultimate answer to life, universe and everything.

    1. John Savard

      Re: This computer, whose name I'm not worthy of saying

      Well, our technology isn't anywhere near that advanced. No, for all its power by our standards, I'm sure that Deep Thought would dismiss it as a mere abacus - and, indeed, many orders of magnitude less powerful even than the computer than it had actually so dismissed.

      1. Ima Ballsy

        Re: This computer, whose name I'm not worthy of saying

        But the answer is ..42

        1. Alan Brown Silver badge

          Re: This computer, whose name I'm not worthy of saying

          "But the answer is "

          ...dependent on the number of fingers you have.

  5. John Savard


    Supposedly, rather than the next generation of EPYC, it will be using the generation after that. So instead of EUV 7nm+, it will be using 5nm chips. I wonder what the HPC and AI optimizations of the custom EPYC chips are both on this and on the earlier Frontier supercomputer.

    1. Unicornpiss


      AMD keeps that piece of the chip they recovered from the wrecked terminator under lock and key.

  6. Sgt_Oddball

    Anyone heard

    How happy Intel is over AMD taking their lunch money?

    Also nice to see their speed projections going up for the same money.

    With regards to the nuclear option, they also test if the bombs still work and how safe they are currently without actually... You know.. Testing them (I.e. Unless we decommission them who knows how everything in there will age, the explosive, the fissile material, even the cases. I'd also be interested in how often these things are maintained for stuff like battery supplies, leaky capacitors and dry joints?)

    1. Tom 7

      Re: Anyone heard

      If you have a big pile of bombs people will take it seriously and ensure that no-one fucks with it. If you take bombs apart and make them 'safe' (dilute) and put them somewhere out of the way then people will forget them and in a few hundred thousand years someone might just dig them up again and go 'wow we can make our hands and faces glow'!

      1. The Bobster

        Re: Anyone heard

        This place is a message… and part of a system of messages… pay attention to it! Sending this message was important to us. We considered ourselves to be a powerful culture.

        This place is not a place of honor…no highly esteemed deed is commemorated here… nothing valued is here.

        What is here is dangerous and repulsive to us. This message is a warning about danger.

        The danger is in a particular location… it increases toward a center… the center of danger is here… of a particular size and shape, and below us.

        The danger is still present, in your time, as it was in ours.

        The danger is to the body, and it can kill.

        The form of the danger is an emanation of energy.

        The danger is unleashed only if you substantially disturb this place physically. This place is best shunned and left uninhabited.

      2. Alan Brown Silver badge

        Re: Anyone heard

        "If you take bombs apart and make them 'safe' (dilute) and put them somewhere out of the way"

        Most "out of the way" places so far happen to have been "the core of civil nuclear reactors"

        The stuff is equally valuable as fuel as weapons (although Thorium makes more financial sense at $150k/kg vs $40k/kg for 3% U235 civil reactor grade(*) or millions/kg for 50% military reactor uranium)

        Weapons plutonium is "dirt cheap" by comparison, as it's made from "depleted uranium" (U238) exposed in the military reactor (U238 also makes cases for "H-bombs" and gives them their added kick - which is why it's kept under lock and key). Thankfully it makes reasonable reactor fuel, so getting rid of it this way is a viable proposition (a Thorium MSR reactor could go 2 better by digesting U238 AND most "high level nuclear waste" from conventional reactors, reducing waste output by at least 99%)

        (*) Uranium and Thorium out of the ground are about the same price, but you 'throw away;(**) at least 89.5% of your natural uranium as "depleted uranium" to make civil grade fuel and it takes stupendous amounts of energy to run the centrifuges to get that far. $40k/kg is believed to be a highly subsidised price.

        (**) Except you don't - depleted uranium is part of the weapons cycle used to make plutonium and MIRVs. This is the shell game of weaponisation in the uranium process. Forget reprocessing. Everything's far too crosscontaminated to weaponise at that point. All the evil shit happens here - and THAT is why the military hates thorium cycle MSRs - it would expose the entire cost of uranium systems as military, not civil. (Civilian plant operators hate Thorium MSR because a commercial deployment would render water-moderated steam bombs obsolete, too expensive to build and too dangerous too allow to keep running at a stroke)

  7. AIBailey

    Is it just me?

    It's great to see the scale of performance numbers being floated around, and the idea of supercomputers has always held a mystique, but I can't help but feel disappointed that this continues the trend set many years ago, where supercomputers are just rack mount machines in rows and rows, where the only distinguishing features are the design on the doors and side panels.

    Bring back the spirit of the 70's and 80's, supercomputers looked other-worldly, instead of looking like any other data centre.

    1. Mo'Fo B'dass

      Re: Is it just me?

      Id like to see - for appearances sake - a pretend magnetic tape machine from days of old that would spin if machine is doing something and slow down when idling along and stop if system crashed. Perhaps rewind on a reboot... Just for the oppos (a quick look up from the screen to see if all is well). A row of PDP 11/70 LED's would do nicely (where's the harm?). Otherwise, where's the charm?

      1. robin48gx

        Re: Is it just me?

        On the PDP-11 in RSX-11 you could tell when the OS was idling because of the LED pattern.

        1. Nige

          Re: Is it just me?

          ICL 2900 series mainframes had a speaker on the central CPU (in ICL parlance, the Oder Code Processer, or OCP) that would chitter and chirp something similar to the old phone modems and fax machines. An experienced operator could tell what stage a batch job was in - extract, sort, number-crunching, spooling, etc - just by the sound coming from the speaker.

  8. Anonymous Coward
    Anonymous Coward

    I wonder...

    Will GSK use this computational power to create or solve problems? Seems to me a military/government operation may have more use of something like Covid-19 than its cure.

    1. LB45

      Re: I wonder...

      Having once worked for GSK (before my jerb was sent to Mumbai) I can assure you nothing would come of it. It would sit idle waiting for the endless meetings to discuss what should be done with it and how to actually, sorta, maybe come to a conclusion.

      Then it would have to wait until after the re-org of the re-org to be completed so the group now responsible could determine if they should move forward or wait until after the upcoming re-re-org.

      Meanwhile someone would be trying to get Crysis up and running on it so they could appear busy during the next series of meetings to decide what should be done with it.

  9. j.bourne

    And the most unintelligible article title award goes to....

    This complete mess.

    El Reg does realise that the purpose of a headline is to attract people to read the article aren't they? I've no idea what the article title means - hell, I only got half way through it before deciding it would be more fun to do my own dentistry.

    1. Champ

      Re: And the most unintelligible article title award goes to....


      It was pretty damn obvious to me. Which is why I clicked on it from the daily headlines email

    2. Jimmy2Cows Silver badge

      Re: And the most unintelligible article title award goes to....

      Uhhh... you know this is a tech publication right? Pretty sure most people here can follow it no problem at all. Hardly word salad.

  10. Anonymous Coward

    So that explains one thing...

    > Uncle Sam has turned to HPE – via newly acquired HPC specialist Cray – to make it all happen.

    In the same way that the US Gov want a wall with Mexico which the Mexicans must pay for, so the US Gov wants a new supercomputer that Mike Lynch must pay for.

    1. Alan Brown Silver badge

      Re: So that explains one thing...

      "the US Gov want a wall with Mexico which the Mexicans must pay for"

      Didn't you hear?

      Mexico said they were happy to pay for a wall along the 1848 border.

  11. Anonymous Coward
    Anonymous Coward

    Pity the poor guy peddling the dynamo to power it due to having to meet CO2 targets.

    Lance are you still doping and available ?

  12. batfink

    What exactly are we simulating?

    Sticking to the nuke question: what exactly are we now simulating that we didn't simulate before, with those piddly little earlier machines?

    1. Yet Another Anonymous coward Silver badge

      Re: What exactly are we simulating?

      Same thing ultimately, but more configurations and with fewer simplifying assumptions

    2. Anonymous Coward

      Re: What exactly are we simulating?

      They will be running simulations with, for instance, smaller element sizes in the finite-element model. That means that the simulation is more accurate and can also deal with things like turbulence &c better. There is essentially no limit to the amount of computer power you can throw at problems like this and continue to get better results.

  13. TimMaher Silver badge

    Mac Pro.

    My early 2008 Mac Pro runs El Capitan.

    And it runs Mojave guests in VMWare Fusion (very slowly), even though it cannot run Mojave or High Sierra itself.

    So what have they achieved?

    I’ll get my coat. It’s the one made from graphene. (See different thread)

  14. Androgynous Cow Herd

    Bad news..

    While initially named "El Capitan" it was recently upgraded to "Catalina" and a bunch of stuff that used to work now doesn't.

  15. Bonzo_red

    World's Fastest Machine?

    What are the chances that in 2023 the Chinese will have surpassed the planned figures for El Capitan?

  16. Bitsminer Silver badge


    Look, it's HPE (ex-Cray), and we know they used to be able to make products, but, really? Three years from now?

    After Cray gets homogenized into being yet another HPE subsidiary, what will happen? Will the staff that make things happen still be there?

    Look at the former SGI product line (SGI was also assimilated, remember) and see if you can find any of it on their website. Or any news releases bragging.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like