Britain's warhead-watcher to simulate Trident nukes with Atos supercomputer The Atomic Weapons Establishment, which provides warheads for the UK's nuclear weapons, is to use a Bull supercomputer to simulate Trident nuclear warhead explosions. Warhead components change over time through ageing, obsolescence and redesign, which affects their operational efficiency. The AWE can't run test explosions any …
"...the Comprehensive Nuclear Test Ban Treaty (to which the UK is signed up), which bans emission of nuclear yield (radioactive debris)..."
Umm... the Comprehensive Nuclear Test Ban Treaty bans all nuclear explosions, for all purposes, in all environments (the 'comprehensive' bit).
If the treaty was just about the release of emissions then we'd still be seeing underground testing.
So you aren't even allowed to use them for real?
Well - yes. But, if you ensure that the first one makes sure to target the organisation what enforces the ban, it wouldn't really matter.
Even lawyers might have problems taking you to court once they are radioactive particles drifting in the wind..
Makes you wonder what the output is like. Is it teraflops of pretty fiery explosion gfx? or a little ticker tape pops out of the thing saying:
Results for 50kt test 2:
Blast Raduis: 2 miles
EMP: 3000 kg m^2 / A s^3
Suntan factor 50: 20 miles
Commie scaring factor: 6/10
Mutant Duration: 1000 years
Given that bomb codes tend to coded in FORTRAN I'd guess in principle porting them to ARM would not be impossible. IIRC those codes have been under continual development (and re-hosting) for decades.
Getting them to realize the theoretical speed of the processors is a whole different story.
FORTRAN?
Yeah, T'was nice to code linear algebra operations on single-CPU vector machines, like, you know Cray Y-MP and that stuff.
I STRONGLY doubt FORTRAN is still a good fit for today's massively parallel systems.
If they are still using FORTRAN, it's just another case of "we have always done it this way around here", followed immediately by a horse terminal removal wagon.
Fortran is a very good fit for big supers. The language has semantics very well designed for good float performance & has evolved to be a lot less horrible than it was, there are extremely good compilers (Intel's is very good, and vendors usually provide their own which may be better), and MPI / OpenMP support is very good indeed (again: vendor libraries help here). And there are really substantial libraries of course.
Source: my day job involves running big numerical simulations, written in Fortran, on large HPC systems (not atomic weapon simulations).
"Given that bomb codes tend to coded in FORTRAN I'd guess in principle porting them to ARM would not be impossible."
Far from it; the gcc suite is available on ARM. I think porting between different versions/standards of FORTRAN would be a bigger issue.
In any case, you wouldn't be running the heavy lifting on ARM; all the heavy stuff would be run on the arrays of GPU-type accelerators that the ARMs manage.
if it goes off, then it will be bad. why do we really need to know how bad? their answer is just going to be somewhere between "really bad for humans and unbelievably bad for humans " somewhere near the explosion.
if one of those things goes off, there are going to be a lot more of them flying around shortly after. is it going to model that as well?
https://www.nrc.gov/reading-rm/doc-collections/fact-sheets/plutonium.html
I'm not a nuclear physicist and don't know which type of plutonium they pulled out of Windscale back in the 1950's, but it's certainly one or the other.
As nobody set off any old nukes, the AWE need to understand if they will still reach the critical mass to cause a chain reaction... or if they'll open a vortex into another dimension where beasties and people much scarier than the fat Nork bloke (with the bad hair) live.
I suppose one of the applications will be to simulate how warheads degraded through ageing will perform. The deployment scenarios - and I do hope they are filed in binders like General Turgidson's "World Targets in Megadeaths" - are bound to be based on a more or less reliable yield per warhead. At one point you'll have to update either the warhead or the contingency plan.
It's all about warhead reliability with degraded cores. As the cores age the amount of decay products inside rise, some of them potential fission poisons. It's the reason the US is so heavily investing in things like the Z-machine and NIF. It's not about fussion power generation per sé. It's about a very specific kind of very short lived high energy fussion process.
Because no-one of he western powers has set one off (so to speak) in the last 3 decades or so there is serious doubt if any of them will even work. Thus all the testing. Not just for reliability when it comes down to it but also upgraded maintenance procedures to keep them in a safe condition.
It would be somewhat amusing if nuclear war broke out and all the worlds weapons failed to go off through lack of use.
Yes, but there would still be a lot of highly radioactive material scattered about (think "dirty bombs" instead of "big-boom bombs").
"if it goes off, then it will be bad. why do we really need to know how bad?"
That "if" being exactly why it needs simulating - nuclear weapons are fairly complicated things, and as the article notes there have been plenty of changes made over the years, as well as other factors like deterioration of components. You can argue all you like about whether we should have nukes, but given that we do have them it's probably a good idea to check that they will actually work as expected and won't, for example, explode on the launchpad or leak radioactive material everywhere while in storage.
"if one of those things goes off, there are going to be a lot more of them flying around shortly after. is it going to model that as well?"
No. Nuclear physics simulations are surprisingly unhelpful at things like socio-economic analysis and strategic planning.
Nuclear physics simulations are surprisingly unhelpful at things like socio-economic analysis and strategic planning.
Well everything is ultimately physics. If we had a really big computer and knowledge of the current state of the world we could simulate all the sociology and economics you could possibly want.
"If we had a really big computer and knowledge of the current state of the world we could simulate all the sociology and economics you could possibly want."
Well, for all we know, somebody might be doing just that right now, and we're part of that simulation.
(You know, I could probably even live with that it it was true; as long as I can file bug reports.)
(You know, I could probably even live with that it it was true; as long as I can file bug reports.)
Well, we're obviously in a simulation. Just look at Trump's hair: it's just obviously not rendered properly. And they've attached the hands of a much smaller person to him, and the whole colour balance on his skin is just hopeless. We're not just living in a simulation, we're living in one which is being done on the cheap.
Let 's see now.. the half-life of Tritium used in boosted weapons is about 12years give or take a month or so. Production in the UK at Chapelcross shut down in 2004 and the Americans only have enough low enriched uranium stockpiled to keep their needs sated another decade. I suppose the bombs will still go bang but they won't be crowd pleaser's, assuming that is they actually detonate.
Has anybody checked the expiry dates on the PAL's ? (Permissive Action Links), they're Tritium powered.
That's why we run lots of different models (and lots of incarnations of a given model with different inputs) and compare their output with each other and with what actually happened (when the models are run in the past), yes. That's the whole fucking point of all this simulation: you don't think we write some golden model, run a copy of it and write a report, right?
> Are you listening, Climate Change?
No, they're not listening. Whatever the truth (or otherwise) of climate change (other than that it's been happening since before humans evolved), climate change models seem remarkably impervious to any alteration in the input parameters.
One wonders whether they are all just minor variants of "sleep 10000; print $MY_ANSWER"?
Atos? they're there as the government gaurantee that the nuclear weapons are fit to work.
Even if when you view them with the Mk 1 eyeball they have clearly rusted through, primers are missing, a puddle of red fuming nitric acid has burnt through the floor, the oxidiser tanks are showing 2% of fuel remaining and the bit where the warhead is supposed to be has been hacked off with a crowbar...
But Atos will simply declare that it is fit to work and that's the nuclear deterrent sorted. move along.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
