Slashdot Mirror

Perhaps I'm biased (I used to work at Thinking Machines) but pany more problems are parallel than you'd think at first, if you have a sufficiently fine-grained parallelism.

"even a compiler can benefit from multiple threads, though current compilers don't do it"

Actually, Parallel Make (i.e. gmake -j, http://developers.sun.com/solaris/articles/paralle l_make.html, or pmake, http://www.llnl.gov/icc/lc/DEG/pmake/pmake.html) can make project builds significantly faster.

Beyond that, any time that you're rendering graphics, or sorting data, or in fact using any large volume of data, or doing more than one thing at a time, multiple processors could help. This is why most graphics cards are highly parallel, and why all high-end databases run well on many processors. Heck, even booting a computer benefits from parallel execution. Of course, it may be hard work expressing the dependencies between processes properly (e.g. parallel makes can break makefiles that have implicit dependencies in instruction order that work serially, but which break when parallelized), but if the problem is worth thinking harder about, it can probably be parallelized.

Thereby making it trivial for anyone with Wal-Mart access to put together a "dirty bomb"?

Repeat after me: Dirty bombs don't work. They are a media scare and nothing else. Campaigns of FUD are designed to fool idiots into believing that everything they read in comic books is true.

Good. Now go here, read, and understand.

Hello? That's the NIF!

Why yes it is.

You've managed to go in full circle away from "the military wants this too" back to "the NIF is doing these experiments"!

To understand your confusion I was looking back at a post of yours in this thread in which you doubted that powerful lasers were to be used for nuclear stockpile stewardship:

Why would you go through all the trouble of testing the dueterium/lithium samples in a reactor when they could just as easily do a purity test?

and:

Maybe you have more info that would help, but I just can't fathom why anyone would attempt to test a fusion bomb in this manner.

Now, from the third paragraph of the article I linked:

By achieving ignition, NIF will allow weapons scientists to perform several kinds of experiments for the Department of Energy's Stockpile Stewardship Program to ensure that the U.S. nuclear arsenal remains safe and reliable.

If you read the article, you will see that NIF has indeed been sold primarily as a method to test the effects of aging on nuclear weapons.

But the idea is not to stick any part of an actual warhead in the middle of a room and blast it with lasers. It is to get better models for D-T fusion works.

We know the half-life of tritium pretty darn well, but we don't necessarily know how the yield of a warhead is affected by the gradually increasing ratio of deuterium to tritium.

As for the military.. some people may consider this a technical point, but the US military has not been in charge of the testing or development of nuclear weapons since 1946. This is a responsibility of the Department of Energy (a civilian agency), not the Department of Defense. The DOD helps out with tests and so on, but it is almost never in charge of any nuclear weapons work.

So it is trivially true that the military does not plan to test hydrogen bombs themselves. But they are very interested in it nevertheless. Similarly, the microlaser program which you linked to belongs to Sandia Labs, not the military. Sandia is a DOE facility, and therefore is a civilian operation.

HTH...

Actually timing, beamshape, and power issues are all the easy bits to solve. Repeatability is very high from shot to shot on other big fusion lasers. If you want to know more about the history of lasers used for high energy density experiments go to the LLNL library do and search on "shiva" "argus" or "nova" lasers to find papers published on the experiments of these first systems. Most of the timing/power issues were solved over 20 years ago.

If anyone *does* have a link to the military doing fusion testing with lasers, then by all means. Post a link!

"Designing for Ignition"

The link that you posted is for an entirely different program, all together.

It isn't hazardous if it's handled properly. See here for an article on target design. They need a low-Z material for the capsule.

Ever since the 20 beam IR Shiva laser at LLNL was built in ~1978 all of the giant multibeam fusion lasers (in no particular order:NOVA, ARGUS, OMEGA, JANUS, LMJ, GEKKO XII, CICLOPS, NIF, ISKRA-5, VULCAN, ..... etc.) right or wrong, are simply referred to as being "a" laser. Also, the fact that NIF will be capable of ignition scale fusion is VERY important in nuclear weapons testing. H-bombs by definition DO achieve ignition.

Huh. this story looks like its almost exactly what I have on my blog. Anyways, the direct link for the NIF is here.

Just think: it's better than half way to a fusion drive if it all works in 2010.

An A/C made the reasonable point that the H1-B program *could* lead to actual immigration, which *could* lead to a positive result. However, I disagree (and will do so politely, because I respect the poster's point of view, even though I disagree with it):

First of all, we already have programs for handling permanent immigration. People can apply for a green card and get admitted; it happens all the time and is how things are SUPPOSED to work. There are quotas, and a certain number of people are admitted, naturally -- you can't admit EVERYONE or half the rest of the world will be depopulated and this country will be an overcrowded hellhole.

I'm glad that you're being so polite now crazyphilman, though I do miss your entertaining "SO FUCK OFF foriegner" haikus.

I suggest that you actually go and talk to some people who've immigrated to America. Your opinions are basically out of touch with reality. Many people do go from an H-1B to permanent residency (a green card). It's substantially harder to get a green card when you're trying to come from overseas.

America is not as friendly to immigrants as you might think. The Bush administration, in particular, has made things very difficult for visitors, especially short-duration visitors. As a result several international scientific conferences, formerly based in America, have moved overseas (e.g. the protein structure prediction competition, CASP). This is a bad thing for America, which traditionally recruits heavily from abroad, especially in science. Yep, I know you don't like it, but it's true.

I predict 2 years ago.

http://wwwcg.in.tum.de/Research/data/Publications/ sig03.pdf
http://www.cs.utexas.edu/~nav/rtgraphics.html
http://www.llnl.gov/computing/tutorials/performanc e_tools/HighPerformanceToolsTechnologiesLC.pdf

The blog is a bit misleading: "Details are scarce about the STELLA supercomputer, built by IBM using some of its Blue Gene/L technology."

Details are plenty since what IBM gave to the project is a couple of racks of BlueGene so everything applies, scaled proportionately.

Here are some details: http://www.research.ibm.com/bluegene/ and http://www.llnl.gov/asci/platforms/bluegenel/

qz

thermonuclear ignition will be achieved in the laboratory in the year 2010 (+or- 2 years). so yes, there will be a breakthrough in fusion in the next 10 years but it will not be due to 60$ a barrel oil. it will be a result of slow and steady progress in the field over the last 40 years and a decision in the early 90s to build the NIF. is this the breakthrough you were talking about? no perhaps not. but the breakthrough you were probably referring to (economical fusion power) is possible and it is still a good bet that this type of breakthrough will also occur in the next 10 years. the Univ. of Rochester's lab for laser energetics will switch on the most powerful laser in the world in 2 years. this laser, at ~3 PETAwatts will be equivalent to about 2% of the total power recieved by the earth from the sun and will be used to attempt "fast ignition" experiments which may indeed make fusion power economical.

Check out http://www.llnl.gov/linux/chaos/ for the 'Clustered High Availability Operating System'...so much for new acronym, as some other people already pointed to other CHAOS projects as well

Actually CHAOS is already taken as well. It's the clustered operating system used by the Lawrence Livermore labs.

http://www.llnl.gov/linux/chaos/

nobody would be able to make a living writing software

I have managed contracts to fund developers working on open source software projects. My employer pays programmers to write software and to release it with an open source license. The Department of Energy (our funding source) has spent literally millions of dollars over the last few years on projects like this.

I contest the claim that writing open source software entails no monetary compensation to the software developer.

Here's an article describing some of the specs.

http://www.llnl.gov/asci/platforms/bluegene/talks/ gupta.pdf

It's from the days when BlueGene/L was still relatively small, but the basic design hasn't changed since then.

Turns out it's split into I/O and computing nodes. The 1024 I/O nodes run Linux. Each controls 64 dual-cpu nodes, which use simplistic microkernels written from scratch using Linux as an example.

The network architecture sounds funky: apparantly it's based on a torus!

Here's a link from Lawrence Livermore that shows some system pics and has some entertaining facts...i.e. what this beast is going to be used for (nuclear simulation)... http://www.llnl.gov/pao/news/news_releases/2005/NR -05-03-09.html

Pics of the Terascale Simulation Facility at LLNL that houses blue gene are available here. My office window is on the far left.

Lawrence Livermore National Laboratory (http://www.llnl.gov/) purchased a BlueGene/L all for themselves. They are not predicting the weather, they are using to simulate our nuclear weapon stockpile performance (http://www.llnl.gov/llnl/missions/nwss.jsp).

Lawrence Livermore National Laboratory (http://www.llnl.gov/) purchased a BlueGene/L all for themselves. They are not predicting the weather, they are using to simulate our nuclear weapon stockpile performance (http://www.llnl.gov/llnl/missions/nwss.jsp).

I havent done any calculations yet

but if you have one of these Inductrack And attach a few small cheap rocket boosters/electrical coils to get the thing up to an initial speed of Mach 1+, you could then get a scramjet fired right from the start and the ship could achieve a very high velocity. From their, once air density becomes too low to operate the scramjet, you could switch to onboard rockets that would provide the last amount of acceleration needed to get into orbit. Or even attached solid rockets.

See, you can learn/talk on Slashdot, it's not all trolls!

Sorry, but I fail to see what Blender and the GIMP have to do with real scientific visualization. Blender is for 3D modelling, and the GIMP is for image processing.

If you're looking for complete, open source scientific visualization and data analysis packages, try VisIt, which supports dozens of input formats and runs on Linux, Windows, and MacOSX. Pick it up at http://www.llnl.gov/visit, or get the latest binaries from FTP here.

I have less knowledge of ParaView, but it is also free: http://www.paraview.org.

Both of these are also developed in part by the national labs; they can run parallel to handle terabytes of data, so if you've got small dataset they should be smokin' fast, and if you've got your own cluster you should be able to visualize some huge data.

If you're looking for just a toolkit to build your own application, try OpenDX or VTK.

Sorry, but I fail to see what Blender and the GIMP have to do with real scientific visualization. Blender is for 3D modelling, and the GIMP is for image processing.

If you're looking for complete, open source scientific visualization and data analysis packages, try VisIt, which supports dozens of input formats and runs on Linux, Windows, and MacOSX. Pick it up at http://www.llnl.gov/visit, or get the latest binaries from FTP here.

I have less knowledge of ParaView, but it is also free: http://www.paraview.org.

Both of these are also developed in part by the national labs; they can run parallel to handle terabytes of data, so if you've got small dataset they should be smokin' fast, and if you've got your own cluster you should be able to visualize some huge data.

If you're looking for just a toolkit to build your own application, try OpenDX or VTK.

It's not botulinum toxin, but it IS some pretty nasty stuff to have in you.

...one ten-thousandth of a gram (0.1 milligram) inhaled can cause cancer. This is correct: we have already estimated that 0.08 milligrams inhaled will have 100% probability of causing a fatal cancer. To inhale 0.1 milligram of plutonium, however, a person would have to inhale more than seven hundred thousand particles. (A single 0.1-milligram particle would have a diameter of over 260 micrometers, about 90 times too big to be readily inhaled.)

you're doing some serious 3d work you should have some professional SGI style equipment.

Last decade's technology. No one doing serious 3D work is using SGIs any more, at least not in the DoE. More precisely, SGI is in bed with nVidia and ATI at this stage of the game, so a good number of people are "rolling their own," as it were. Simple fact, a cluster of Linux nodes with nVidia 6800s can toast an SGI any day. And it's a lot cheaper. Check out this article for more information.

does this mean we can pull the plug on this monstrosity.

Seriously. Look at the pictures, does anyone else get the idea that this thing is WAY more powerful than any nuclear power plant..... (sans radiation of course..)

sonoluminescence has not been proven to actually be fusion. It's just a lot of light and some heat in water that's been compressed by sound. much more interesting than that though.

And they claim that this process that isn't fully understood yet will get break even fusion in 5 years? Doubt it seriously.

I know that that what impresses me in addition to course work is someone interested in computer science. People who learn on their own and do things on their own. People who have contributed to open source projects or published some of their own software.

Also, internship experience. This allows you to talk about accomplishments in something like a real world setting.

And on the topic of internships... Let me give a plug for the internship programs at the National Labs and in particular at my employer, the Lawrence Livermore National Labs (LLNL).

LLNL has excellent internships in a variety of areas, including physics, materials science and computer science. A friend's daughter did an internship here last summer and was given access to the microfabrication (silicon) fabrication facility. Another student did some work to add the Reiser FS to a distributed computing system. Some of the departments have seminar programs for summer interns which are interesting as well.

If you are interested in applying see the LLNL web pages (don't send me email, I can't help). You should apply NOW, since some of the programs stop accepting applications by the end of December (some allow applications through January and Feburary). In many cases you will need letters from professors, so get those letters now before winter break.

Wow very few remote exploits
Well I feel much better now. Unix might have less high profile automated attacks against it, but don't kid yourself into thinking its any safer on the Internet then anything else.

Its not like I really had to look hard either, it took longer to write the little HTML in this post. Results 1 - 10 of about 150,000 for remote linux root exploits. (0.30 seconds)

POSIX threads; not only cross-platform and essentially built into C (remembering that while C itself eschews the One True Library approach, there are certain libraries so pervasive they might as well be part of the language, and POSIX libraries would be on the top of that list), but also cross-language as many other languages built on C have inherited that model.

Java is great... I guess, personally I think it's the worst thing to happen to systems programming since C++, but since popularity and power seem to be inversely related, Java is great, I guess, or people will shoot me... but to date, it has not done one original thing, or done it first. It shouldn't, that's not its thing. If you ever think Java did something first, odds are, it's your ignorance showing.

I think most folks in the /. world consider IT to be the 'tech' industry. Not surprising due to the backgrounds of the people who read/post here. As for 'tech' jobs, there are quite a few in my region of the technology world:

LLNL has 20 open S&E positions.

INEEL in the middle of transitioning contractors, but will undoubtedly need S&Es to complete missions for DOE and the Navy.

LBL has 95 open S&E positions.

BNL has 7 open S&E positions.

SNL has 20 open S&E positions.

LANL has 107 open S&E positions.

ORNL has 28 open S&E positions.

PNNL has 36 open S&E positions.

ANL has 32 open S&E positions.

There complete list of laboratories is here. All of them have job postings in the S&E categories. These just happen to be the largest insitutions.

I haven't even started searching Monster.com

The differences between the Human Proteome Folding (HPF) project and Folding@home have already been mentioned. The differences between HPF and the recently completed Distributed Folding (DF) project should also be mentioned. HPF and DF attempt to predict the 3-dimensional, or folded, structure of protein sequence data. Both projects are well suited to parallelization. DF used an in-house algorithm to predict the structures of small proteins (which may or may not be in the human genome) with known structures and of proteins with previously-unknown structures in the CASP5 and CASP6 structure prediction contests. HPF uses the Rosetta software package, developed by The Baker Laboratory at the University of Washington, to predict protein structures for proteins which occur in the human genome.

DF is currently redesigning its folding algorithm using the results from its first project, and may begin another project in the future. See my summary of DF for a quick history of the project.

The differences between the Human Proteome Folding (HPF) project and Folding@home have already been mentioned. The differences between HPF and the recently completed Distributed Folding (DF) project should also be mentioned. HPF and DF attempt to predict the 3-dimensional, or folded, structure of protein sequence data. Both projects are well suited to parallelization. DF used an in-house algorithm to predict the structures of small proteins (which may or may not be in the human genome) with known structures and of proteins with previously-unknown structures in the CASP5 and CASP6 structure prediction contests. HPF uses the Rosetta software package, developed by The Baker Laboratory at the University of Washington, to predict protein structures for proteins which occur in the human genome.

DF is currently redesigning its folding algorithm using the results from its first project, and may begin another project in the future. See my summary of DF for a quick history of the project.

No, going from sequence to structure is a big problem; see e.g. the CASP competition. The fundamental difficulty is that protein folding involves many complex interactions between amino acid side chains and solvent molecules, getting you into a world of nightmarish quantum chemistry where energy landscapes are rugged and rules are made to be broken.

In general there are two ways to approach structure prediction. The most reliable is homology modeling where you basically find a similar protein sequence (i.e. a close evolutionary relative) whose structure is known. Current protein database searches generally rely on probabilistic models borrowed from natural language processing and speech recognition, primarily hidden Markov models. Essentially, these models address the evolutionary process (which describes how different proteins are related), rather than the folding process (which describes how individual proteins fold).

If there aren't any similar proteins with known structure, you're into the domain of novel fold prediction, the second (harder) way to predict structure. The current best novel fold prediction methods begin by breaking the protein sequence into lots of tiny fragments (think words), then doing homology modeling on these fragments... e.g. the ROSETTA program from David Baker's group.

Simulating the full folding kinetics, as folding@home does, is even harder, and involves wading knee-deep into all that nightmarish quantum chemistry (or approximating it). Here you are interested in not only the final folded structure of the protein, but also its intermediate structures (hence the applicability of this approach to study of misfolding diseases, such as those involving prions).

Thank you DeepStream for pointing out the difference between folding@home and this ROSETTA-related project... teach me to respond without rtfa...

Considering that the origional source was Lawrence Livermore National Laboratory, I think that it's probably trustworthy.

1 milligram of plutonium spread on a field would kill the grass, no matter how you dilluted it and grass wouldn't grow again for a long time.

I'm sure I didn't explain this as well as he could have, but I hope you get the gist of it.

Your concept is correct, but your facts are horribly incorrect and it distracts from your point.

WIkipedia describes the myth of Pu toxicity you refer to.

A Perspective on the Dangers of Plutonium also deals in reality on the effects and dangers of Plutonium. Plutonium's danger lies in it's radioactivity and a Mg spread out over a field of grass is all but inconsequential. Junkscience.com has a short blurb about the effects of low-level radioactivity that would suprise many who have been led to beleive that radioactivity is a large and deady threat.

Toxic is a relative term, not an absolute, and there are multiple avenues of toxicity. Most laymen use the term to mean a substance's chemical toxicity.

Plutonium's chemotoxicity is less than that of caffiene, acetiminophen, and so on. It's radiotoxicity is 1/200th that of Radium, a naturally occuring substance in soil.

So basically, that horse urine is a greater threat to that field of grass than that Mg of plutonium.

Some people are working on an earthquake alert system that detects an earthquake at the epicenter and sends an wireless signal out to others. Because radio wave travel faster than ground waves, the alert reaches people seconds before the quake hits. Its not much of a warning, but it may be enough time to shut down some processes, park the heads on the disk drive, turn on the backup generator, etc.

That is something that you just cannot rely on.

For example, US Senate ratified the chemical weapons treaty. That's good. But: "When the U.S. Senate ratified the CWC, it implemented a mandate, Condition 18, which states that no sample taken on U.S. soil shall leave the U.S. for analysis during an OPCW inspection." (http://www.llnl.gov/str/May03/Alcaraz.html) So basically US can now ignore a neutral inspection and play with the samples as it likes, making it always look like there is no chemical weapons deployed or stored on US soil. That is pretty much counter to the spirit of the treaty. But that is what you will get, so it is best to not to trust the good intentions.

This "can't trust" is also why Russia and Europe are making their own GPS clone.

A quick Google search has netted the following: OS - Linux, HPK (High Performance Kernel) Complilers - Fortran95, C99, C++ Math Library - a subset of ESSL If you would like to read the article, it can be found at http://www.llnl.gov/asci/platforms/bluegene/talks/ gupta.pdf

I think if your kid could find a gram of plutonium the fact that he could even find A WHOLE GRAM is more serious than him eating it.

For other hazards, this link is helpful

I wouldn't worry about that too much. Manufacturers would tend to be smart enough to choose materials that are not water soluble. In addition, they'd probably melt the materials inside a block of non-reactive metal to make sure the materials stay in a solid form.

As long as the materials are treated with respect by the manufacturer, consumers shouldn't have too much to worry about. Even if the manufacturer DOES screw up, it's doubtful that so little material could cause much of a problem. You might be interested in this link. :-)

I've been harping on the idea of using nuclear batteries in cell phones and laptops for the past year or so. To date I've been called a variety of names for it, the least of which is "crazy". Yet here we are. Researchers are SERIOUSLY talking about using radioisotopes as power sources!

In case anyone is wondering how these work, the idea is that the radiation from a small amount of radioactive material (NOT fissable material!) is captured and converted into electricity or other forms of energy. There is very little radiation emitted by these devices, because the radiation IS the power! Letting it escape would be poor economy.

NASA has used these sorts of devices in spacecraft for 40+ years, starting with the Apollo missions. NASA's earlier designs produced about 75 watts utilizing a few pounds of Plutonium-238. Pu-238 was an excellent choice because it is useless for bombs, and has a short half-life (~80 years). With the public finally calming down about nuclear technology, NASA is now developing a more efficient device called an SRG. These devices get about 55 Watts per 600 grams of PU-238. This is way more efficient than current RTGs, like the ones used on Apollo.

The primary downsides to Nuclear Batteries is that they are expensive and they don't scale. They are expensive because the nuclear materials are very rare and expensive to process. If we started using these materials in massive quantities, it's a certainty that the prices would drop. They are not scalable, because the amount of materials required means that a few hundred watts is the largest device one could construct with a reasonable size, weight, and expense.

As for anyone who's worried about dirty bombs, I suggest you read this and this. The threat has been greatly overstated, and is actually less effective than a regular bomb. The real problem is the issue of keeping the materials out of landfills. Even today, there's a big problem with Lead, Cadium, and other dangerous materials ending up in landfills. Radioisotopes wouldn't be much worse, but there is an upper limit on how much you want to add to the sub-soil.

I don't know whether that particular configuration has been observed experimentally though.

It has

I wonder how much "noise" there is with all the natural seismic activity in Japan (3 large earthquakes in last week).
If anybody is interested there is a way to determine if the seismic activity is from an explosion or from an earthquake, or nuclear blast.

IANAGP (IANA Geophysicist), but the images on this page seem to indicate that the wave measured went through the earth's mantle (a teleseismic wave), though I haven't been able to find any details on the velocity of such a wave. Any geophysicists in the house?

So, can you make an RTG out of U-235, or does it not emit the right kind of radiation?

You could use U-235, but you wouldn't get much power out of it. You see, U-235 has a half-life of 7.038E8 years. That means that in 7.038E8 years, 1/2 of the U-235 will decay into Thorium. Of that Thorium, it will decay into other products such as Radon, Radium, and Lead. These processes also take a certain amount of time. The remaining mass of the U-235 will become Alpha and Gamma radiation.

Now each radioactive particle has a certain force and penetration power. Gamma rays penetrate very deep, but have no mass. (It's basically a photon.) Alpha particles are very heavy, but are easily stopped because of their +2 charge. (i.e. Helium ions) Collision events from the radiation causes kinetic energy to become heat.

Here's the problem. If you've got a half-life in the range of millions of years, you've probably only got a few radiological events every minute. A single alpha particle doesn't carry much kinetic energy. :-(

This is where Pu-238 comes in. Pu-238 has a half-life of 87 years. That means, that in 87 years, 1/2 of the plutonium you have will become Uranium-234. 87 years may sound like a long time, but that's enough radiological events to produce massive amounts of heat! If you were to hold a chunk in your hand (not a big deal!), the plutonium would feel hot in your hand. Placing that same chunk in water would rapidly cause the water to boil.

Obviously, anything with a shorter half-life is going to be hotter than Pu-238. Pu-237, for example, has a half-life of 45.7 days. That stuff would be so hot, that you'd probably be cooked to a crisp by a small portion of it. Even worse, the fast decay rate would make the Gamma effects more pronounced, thus increasing your internally absorbed radiation dosage.

Then there's the thermal output, which seems like it should be considerable. How many BTUs would something like this put out?

You'd have to do the calculations to convert it to BTUs (something I'm too lazy to do right now). However, you can't tap the waste heat, because that's what the RTG is already doing. If you fail to cool the RTG, you'll no longer have a heat differential. Without a heat differential, the pelter will not operate.

Look up a Stirling engine sometime. I think you'll find that a bit easier to understand. Basically, the engine makes use of the fact that heat wants to equalize. Thus the heat will move toward the cold side of the engine. By putting a piston in the way of a heated medium, you can tap the differential for power. BTW, radiological Stirling engines do exist as something called an "SRG" or "Stirling Radioisotope Generator". They're actually quite a bit more efficient than RTGs. The trick is that they're rather new tech, despite the fact that Stirling engines are about as old tech as you can get.

There's some note that the material could be used for a "dirty bomb," which is a concern in this age of terrorism

Dirty bombs are stupid and won't work. Sure, they might cause some panic, but there are better ways of doing that. Amazingly enough, the terrorists seem to be smart enough to realize this, and have pretty much given up on the "Dirty Bomb" concept. It now exists only as a phrase the media likes to use to stir things up.

So in summary, Pu-238 is used because it produces a lot of thermal power. U-235 produces negligible thermal power from radioactive decay, and is much more useful as a fissable material. We also learned that half-life has everything to do with how radiologically "hot" a substance is.

While we're nitpicking about the comparison to a VW, I thought I'd point out that while the article says the largest version weigh 500 tons and is 15 meters in length, Science and Technology Review just says that the reactor will be no larger than 500 tons and 15 meters in length. I'm not sure there's any reason to expect a smaller reactor. I think that's a lot closer to the size of a locomotive than a VW.

With nice pretty pictures. LLNL/a