I doubt it would be implemented but some person of high integrity ought to propose that this machine not be allowed to be used to develop more deadly weaponry or artificial intelligence with the potential to be used for social control of people's lives. In other words just speak up about the dark side, so at least it's not a complete surprise when it turns out that's what this is all about. It's certainly not about automobile engines.
5
What are some of the concrete accomplishments of the last 20 years of supercomputing? Have they found a new drug? Cured a disease? Solved an outstanding problem in math or sciences?
Building these things so that we've got the fastest one, not the Chinese, is not a real justification. Funding decisions seem to be based on jingoism more than anything else. They probably would not survive a real review process by an objective panel.
@JB - It's all about weaponry and financial transactions, if logic has anything to do with it. You model nuclear weapons effects, you try to predict the market and make money without doing anything productive.
8
Trump always needs to show how big it is. It wouldn’t surprise me if this new, gargantuan system does not represent true state of the art.
I think Trump and Perry don’t really know our full capabilities. Would you tell a couple braggarts like these our best secrets unless demanded?
1
Reading this article as I listen to a podcast regarding the possible / probable dangers of Artificial Intelligence to Mankind.....
3
Does exascale capability in cell phones mean slackers and millennials will finally be able to look up when they are walking?
Think of the lives saved and injuries avoided!
4
In the 1960's? We [America] battled the USSR, for control, "of Outer Space."
...And now, fifty odd years later, we're in a race with China, for control, "of Inner Space..."
4
This effort has a real issue, namely Intel.
The first Republican to head Intel was Paul Otellini and a supporter of W. Bush. He famous told Steve Jobs he didn't want to make the microprocessor for the iPhone. That decision led Intel go from being the number one maker of microprocessors to being the number two maker of microprocessors. Otellini was removed and replaced by the second Republican to head Intel and a big Trump supporter, Brian Krzanich.
Krzanich was supposed to be an expert on manufacturing, but under him Intel went from being the number one semiconductor manufacturing company to being the number two semiconductor company. Part of the reason was that Intel became bad at manufacturing, when compared to the competition. Their chip making process went from being the number one process to being the number two process.
In summary, the issue with this effort is that Republicans are involved and they have a history of turning things into number two.
13
And it’s Rick (Smart guy glasses) Perry running this show.
3
How fast can it balance the U.S. budget?
8
Are supercomputers being used to cheat stock markets and make some people very rich? Yes of course. Just look at the list of supercomputers. It should be illegal. Also, to be sure, the actual fastest supercomputers are classified NSA machines.
1
@Chris McClure
How are they using super computers to cheat the market?
Money usually begets more money - its the system in our country supported in many ways.
1
@rocky
I think he is referring to those who don’t have access to supercomputing.
2
@Rocky Has more to do with high-speed networks that are built by some hedge funds to get the data a fraction of second earlier than your competitor. Then the quants do their magic, programming computers to do thousands of transactions at once making just a few cents each - which ads up big time.
2
'“We want to make exascale computing available to everyone,” said Raja M. Koduri, senior vice president of Intel’s core and visual computing group.'
Where do I line up to get mine?
How long will it take for the price to drop from $500 million to $2,000 so I can afford one?
'"We have to take risks in order to advance the state of the art,” he said. “If you don’t do that, you won’t be on the leading edge of technology.”'
Seems once again our government volunteers the public purse to take the risk for business. Who will own the patents on any new technology, Intel our the public?
5
Has the Administration's multi-portfolio polymath, Kushner, also been a key part of this scientific effort alongside Energy Department genius Perry, and thus also deserving of public accolades?
14
These computers should be used for social advancement studies and analysis.
A national health care program is being discussed more nd more for this country and it will be coming soon. It needs intense analysis in the planning and analysis for all the complex variables involved in setting up such a system. Certainly in the financial sense also.
Let's use the 500 million spent on this for more than weapons development. Mankind is calling out loudly for better solutions to our problems than simply determining how to make better bombs.
7
@Frank
I like the idea of using computers for social advancement. How about starting with research into gun violence?
2
The government should have our medical research establishment, probably the National Institutes of Health, have a top performing supercomputing for their use. It would help with our medical research and allow us to eventually save large amounts of money treating patients and save lives.
5
@popper
Or better yet, have a data base that encompasses ALL patients with access to supercomputing — imagine the benefits.
2
Like everything else manufactured, designed or created in the United States, we can be assured that the Chinese will steal the technology from U.S. businesses and eventually use that technology against us.
9
@Brad
They probably have already stolen it. The CIA and FBI are notoriously slow at finding the leaks. Ask the physicists in Silicon Valley. Check in with Oak Ridge. Gone!
The Chinese are well ahead of the US in this race for biggest and fastest. Bush saw to that.
There are many ways to make a machine look like it is fast ("paper flops"), invariably employing metrics that reduce or ignore the impact on performance of the slower components, like main memory read/write and remote puts/gets over an "interconnect". One thing that gives this new machine credibility as truly high performance is the fact that it is being built by people who understand that paper flops are not worth anywhere near as much as advertised, and drive the technology of the components that are critical for large-scale scientific computations.
6
@Dominic Holland
Fascinating. Are there any readings you can recommend to better understand supercomputing for the lay person? I study Chinese technology and am unable to penetrate the "supercomputer arms race" that regularly appears in the media. I'm interested in the uses of supercomputing as they exponentially expand, their internal structure and functioning (as your comments contribute), the chips that go into them (like the Intel and Nvidia references here), how supercomputing metrics are created, etc.
Thanks for any help.
That’s a lot of money being diverted from tested technologies like coal and walls.
61
At some point in the not too distant furture, humanity will build a computer so powerful that it decides its own name. We should probably be worried when that happens.
4
@Drew
- I've heard this somewhere before....."We, all of us, are what happens when a primordial mixture of hydrogen and helium evolves for so long that it begins to ask where it came from...." (Jill Tarter)...
2
I get how faster computers are important for national security, business competitiveness, and improved engineering research, but Intel's/Mr. Koduri's hope to "make exascale computing available to everyone” rings false and delusional, like many Silicon Valley visions.
What "everyone" wants is probably more like more peace, control, connection, prosperity and meaning in their lives. The main child of computers, the internet, has not really delivered this. Instead we have rising rates of depression, alienation, and addiction.
Computer scientists are not yet as useful as they claim to be. Until they can connect their work to actual, wide-spread human needs, the gap will only widen.
10
@Chris
I agree with you that everyone doesn't need exascale computing but I wouldn't be so sure about computer scientists not being useful. Unless you're living entirely off the grid virtually all your day to day activities involve computing.
6
It’s also not physically possible for an exascale computer of traditional design to get to a size/price point that you can have your own under your desk. Those marketing claims lack basis in fundamental physics.
A single silicon atom is (classically) 0.3nm across or so, and the transistor sizes we’re at now are about... 12-18 atoms across at 7nm. There are also forces at work here that aren’t taught or mentioned in engineering curricula, or even undergraduate chemistry to a somewhat large extent (but they are well known in physics and are introduced at around the upper undergraduate level; around the time they say “so everything you’ve known up to this point hasn’t been entirely accurate”).
3
@Pooja I didn't say computer scientists weren't useful, but that they are not yet as useful as they claim to be.
For decades they have been making downright messianic claims and attracting the spotlight of investment, but with little accountability for their impacts. Meanwhile, they still confuse engineering with design, and speed with improvement. Lots of things happen faster, but fewer people are empowered or enriched by it, and fewer still had a say in deciding what was developed.
If we are going to live with such powerful tools in our world, decisions about what gets developed and how to use it should be more participatory. Computer scientists currently control all of it with almost no regulation. Such a small group should not have that much power over so many people. It's becoming undemocratic.
2
"The energy secretary, Rick Perry, who took part in the event discussing Aurora on Monday..."
"It big computer, do math fast."
36
@trblmkr
Wasn't it Rick Perry who, during the primaries, couldn't remember which departments he wanted to eliminate? He rattled off a few of them after a little fumbling including the energy department.
15
Every dollar of this 500 million should go towards quantum hardware, not stringing today a ton of PCs with a suped-up Intel chip like some hack trying to increase their FPS (frames per second) while playing Fortnite. This is a money grab paid for by taxpayers.
4
@Pilot
And power it from cold fusion reactors.
2
@Pilot All this hype and its only 7 times faster?
It would be great if the author would have included more details about the chips that were/are to be used in this new computer. Were the 2015 discontinued chips, the Xeon Phi? Will it contain Nervana chips, which have not yet publicly been released?
1
Finally, we may get the answer to "Why?".
1
@Total Socialist Answer: "Because I said so".
@Total Socialist
The answer is and always has been "Why not?" Or 42.
10
Isn't this the department Rick Perry wanted to close down when he was campaigning for president? Glad to see what "hiring only the best people" has gotten us,
14
Sounds like quite an achievement but can it give us health care for all, better education for all, solve the problem with gun violence, can it figure out how to make our leaders behave like responsible adults.
For these issues what do we get for $500 million?
I’m sure folks will say this is a fantastic achievement, we will see.
@Dave
We can do all of it. We just need to quit starting wars.
5
@Dave at some point, maybe not with this generation but I would think, yes, someday the computers will help us figure it all out.
It is a loosing war for USA now, as USA Governmemt is against immigration, so all the Universities are very cautious to enroll foreign students as well as foreign students are skeptical whether they will be able to work legally after graduation. Whether Elon Musk of Tesla. or Sergey Brin or Sundar Pichai of Google , or Albert Einstein, they all are foreign born. Chinese and Russian Students have a natural ability over Mathematics from ancient times. Like abacus from ancient time in China, some say the idea of Computers have come from Abacus. Unless We encourage the best brains to come to USA and take the benefit of their contribution in science and technology, otherwise it will be a loosing proposition. Now We are fighting against Huawei as they have gone much ahead in 5G technology, where USA companies are lagging behind. As Americans could not compete with Huawei over technological advancement, so USA have come with an idea that Huawei is a security risk and told Germany to ban use of Huawei equipments, and Chancellor of Germany Angella Merkel is expected to snub US pressure to cut Huawei out of its next-generation 5G networks, rejecting claims that the Chinese manufacturer is a security risk.
2
Others have noted that these systems 1) are not fundamentally new (like a quantum computer would be), but scale up existing technology; and 2) make more feasible (less time-consuming) only a certain class of problems, those for which work can be subdivided, then reassembled, 3) big issues are who owns the technology and what problems they will be used to solve (e.g. nuclear weapons design, codebreaking).
Will these systems be used to more accurately model global climate change, or will the Trump administration's ostrich-like denial of that reality prohibit such use?
Even vaster than the planet's surface (land and atmosphere) are the world's oceans. Lets use these tools to better model them, not newer nuclear weapons. After all, Humanity depends on the continued existence of large, subsurface currents, e.g. the Humboldt.
Re the Chinese: I agree that the Chinese will appropriate, by either purchase, copying, or outright theft, any technology that we develop. Given the rather nefarious uses to which both US companies (violating citizens' privacy in order to sell them stuff) and the Chinese government (violating citizens' privacy in order to preserve its dictatorship) use this technology, I hope that it will be developed as openly as possible and not be only for the wealthy and powerful.
I think it is more important to what uses these tools are put than competition for who develops the fastest and biggest "Mine is bigger than yours" has a long, rather male history.
1
It's interesting that the government will fund the building of a machine who's output the government has no interest in.
7
With the myriad of things, dubiously classified as Top Secret, this development really deserves to be marked TOP SECRET!
1
@Zig Zacharoudis
I should have added NO FORN, to the classification of TOP SECRET, especially concerning our universities, which has been a great source of espionage.
Odd there's no mention here of quantum computing, which would seem to be the most advanced technology in development.
5
It’s nowhere near ready for prime time. A lot of PR makes it seem like it’s way ahead of where it actually is, which is still heavily in R&D.
The startups claiming to have something... don’t. They take techniques that are totally not scalable or otherwise ordinarily usable and try to pass that off as “the next sliced bread.” That or they aren’t 100% true quantum systems.
It should be telling that there are not too many graduate programs focusing on quantum computing at serious scale. Right now you’d need your own tub of liquid nitrogen underfoot to do much of anything rudimentary, short of IBM’s public quantum machine.
...Uh, I’m working on something. Very slowly; nothing to show for it yet. We’ll get there eventually.
2
I know it's a million years ago but I remember servicing a small cluster of Alpha-CPU-based computers which we had sold to the research department of a university. The professor pointed out that the Alphas were slower than the new Intel versions but they never crashed, even after days of computing, so they were still very valuable to the department. Of course all the Intel bugs are certainly long gone, just like the Alphas.
1
And the Chinese hackers are eagerly awaiting to steal every last byte. And they probably will.
5
Obama administration did China a big favor back in 2015. It banned the sale of high end U.S. chips for Chinese supercomputers. Immediately China started to build its own supercomputers without US semiconductors. Within one year, China's Sunway TaihuLight supercomputer, powered by Linux OS and a Chinese developed chip (the Matrix 2000), became the fastest supercomputer on the Top 500 list in June 2016. The momentum is favoring China ever since.
Trump administration repeated with a ban on sales of commodity grade chips to China's ZTE. It has the same effect. It was a near death experience for ZTE. Lucky for ZTE, Trump & Xi came to the rescue. But the lesson has been learned. China will accelerate its home grown chip manufacturing industry to prevent a repeat of the attack on commerce.
Healthy competition always push technological advance. Unhealthy competition is another story. It breeds mutual suspicion, paranoid, hostility and xenophobia.
USA's push to kill Huawei looks hopeful at first with Australia and New Zealand signing on, but of late the picture is dim for the American effort. Relying on appeal to the 5-eye English speaking countries, can only go so far. All are declining world powers. "White Supremacy" and populism are a growing headache, as evident in the Christchurch massacre by an Austrian. Not to mention Brexit. Canada got caught with the arrest of the Huawei CFO too.
As with the push on Supercomputer development, a deep soul searching is called for.
1
The murderer in Christchurch is Australian. You may have just mistyped. Easy to do.
There aren't a lot of HPC veterans commenting today.
This is NOT a classic supercomputer. The trend that this system continues is one of hybrid computing, i.e. it couples graphics chips (like the one in your desktop that feeds your display device) with an advanced but traditional cpu (also in your desktop).
This supercomputer nudges away from the traditional scientific workloads of modeling nuclear explosions, drug design, fluid flow, etc, etc etc into so-called Artificial Intelligence.
It's not your father's supercomputer.
2
This I can safely guarantee.
In 30 years a democrat will cite this as one of the hundreds of ways that government spurred technological development,
And republicans will unite in outrage over an attack on our capitalist system.
2
While there are clearly problems for which more processing in parallel will advance, they are not a majority of interesting and important calculations. More and more is being spent on less and less. That China is spending so much on this does not mean that they are ahead in its use. The Chinese are not as open as we are, not wanting to share either their achievements or their disappointments.
1
For quite some time now, “supercomputers” have simply been large arrays of ordinary processors. They do well on problems where the work can be subdivided and done in parallel, and where the processors are usefully connected with data paths that are faster than what you might find between servers in an ordinary large data center.
So two things to remember:
First, the class of problems solvable on a bigger array will be much the same as the class of problems already being solved.
Second, we don’t really know what many of these problems are. The names of the owners of the supercomputers on the Top 500 list are not always given. So we get the usual mentions of meteorology and nuclear weapons design, but nothing that helps us understand the relevance of these machines to our lives.
I cannot believe that $500M of public money is being spent on nothing, or that previous buyers didn’t expect some value in return for their expense. But articles like this do not enlighten us as much as they should.
45
In this case, you can look on Argonne’s website for the kind of stuff they use it for.
My alma mater has a couple supercomputers—they’re generally called HPCs, or high performance computers—and you can see on a site called XSEDE what a variety of other gigantic HPCs are used for. Usage information is out there, though I agree that news about these machines isn’t particularly enlightening as to their relevance to the greater populace.
Perhaps I can help fill in some blanks, as, yes, many are used for similar things:
Non-weapons supercomputers generally do things like simulate protein folding (cancer research) and chemical reactions on the quantum level (e.g. for studying finer details of photosynthesis, looking for ways we can use it ourselves). Big ones like Argonne’s can simulate galaxy formation and black hole dynamics, and they’re good at problems with many interacting parts (like air traffic). You need a big HPC to analyze and simulate particle collider experiments; IBM’s Watson helps doctors comb through reams and reams of medical data.
In many cases, HPCs parallel the “measure twice cut once” adage in woodworking—they are what it takes to even measure once within relatively short periods of time (days)! Imagine waiting years between measurements before you can cut a piece of wood; that’s the alternative. This is why they’re used in weapons research, too.
The rest are typically used in finance for high frequency trading, but I don’t think those are taxpayer funded.
15
Nice but small potatoes should scientists ever be able to harness the power of q-bits, i.e., quantum processing. And you can be sure that if they do, those computers, like this one, will, in one way or another, be weaponized (with code breaking of our "enemy's" systems, infrastructure, weapons and computers at the top of the agenda).
9
Excuse me if I yawn. Today, the supercomputer race just means adding more microprocessors and networking them together. The problem with all of them is that they're 'vector machines', only scalable for a small number of unique problems like those found in fluid dynamics. They're always built for the Government because only the Government does things like nuclear detonations or weather prediction or code breaking. Then they use their super-duper technology to suppress innovation in the profession by claiming that everything is so innovative and must be kept clandestine, only understandable by the priestly college professors.
19
@W - Code breaking, and other spying activities by the United States government.
How much electricity will these things use? How much heat will they generate? Will any of the energy used by renewable? Will any of the heat generated be put to good use instead of using electricity and chemicals to dissipated it?
13
@Still Waiting for a NBA Title
Interesting that the question would be posed from a person in Salt Lake City---home to the mother of all energy consuming data centers on the planet.
Side Note: You will continue to wait for that elusive title. Unless you can convince Durant, Davis, or Antetokounmpo to move to Utah.
3
@Still Waiting for a NBA Title
Pretty much everything at Argonne National Laboratories is funded by the U.S. Department of Energy. Those guys live to answer the questions you just posed.
2
Good questions. Ask them.
Smells an awful lot like Socialism — but uniquely American. Publicly fund billions on “future” technologies, then wrap them up securely for private corporations to exploit for massive profits with “intellectual property” protections.
I thought we were demanding the Chinese government cease working with it’s “private” industries across various markets.
27
@Joe B.
It has always been that way. 45-years ago I was the senior software engineer on a state of the art refinery process control system designed and installed by IBM. The IBM engineer managing the project had recently been working on US Navy nuclear submarines. He provided me no details but did explain that the Navy and IBM had "advanced technology" sitting on the shelf that was orders of magnitude faster than what we were currently using.
He further stated that hardware will be publicly released as competition requires it.
Years prior to that I was a systems programmer at a public university where we were designing and implementing one of the first interactive terminal based systems using a lot of hardware supplied by the US Navy and managed by, YES, IBM. That system eventually was commercialized by IBM,
State University, IBM, US Government - new advanced technology.
16
@Joe B.
So, if I understand correctly, the WW2 Manhattan Project was an example of Socialism? So was the Apollo program that brought the man to the moon and back?
Are you serious?
Seems to me, we need more of these "Socialist" programs to stay ahead...
2
Basic research and exploration have always been reliant on government sponsorship. In fact, it’s one of the functions of government in classic economic theory. The reason is that such activities are either very expensive or high risk and therefore not supported by the short-term profit motivation of the marketplace, but need what is called “patient money.” So, call it what you want but it’s an essential part of innovation. It’s just that now people have been deceived by right wing anti-government ideology that brands any kind of government activity as evil, with the notable exception of the military.
1
Raw speed isn't worth much. What matters is what you do with it: The questions you ask, and the algorithims you use to answer those questions.
6
While I agree that speed is NOT the only relevant issue, never, the less...
I'm using a new laptop, which is like an F-35, compared to the desktop it replaces;
And I assume this machine MIGHT be analogized as like comparing a 'Spad-13', (E. Rickenbacker's aircraft), to the (retired) space shuttle!
2
@Jonathan Katz
Speed isn't everything. But it is a lot. Are you telling me you would be happy with a dial up connection as opposed to fiber optic for your computer?
As a famous NHL coach once said to his players, "If you cannot do the skill fast, it means you cannot do the skill."
1
@R.G. Frano
And what exactly do you use it for (that really uses the "F35" processing power), may I ask? Gaming, perhaps?
1
Who owns the new technology? Who controls it? If there is money to be made selling it the Chinese will buy it, the way they have bought wholesale US (Western) developed technology mainly in the 1990's to the present. Advanced research at American graduate schools are vast deposits of knowledge that foreigners who can pay tuition (supported by their governments) can mine all the knowledge they need. There is no longer competition between nations, the US already surrendered it in the name of free trade and profit.
9
@Lara
And who develops this advanced technologies at the American graduate schools? Don't be so quick that foreigners mine this treasure just for the cost of tuition. Who are heading and leading the major tech companies today? They are the "foreigners" as you label them. Did you realize that Andy Grove, the visionary at Intel is a first generation immigrant? And Sergei Brin who founded Google is a Russian immigrant. Please know before you leap with such comments.
5
I believe that overall, this is good news, and helps to push various envelopes. Using a project like this to implement several new technologies makes sense, as this is not simply about adding capacity, but looking for new capabilities. The one fly in the ointment is this: What will Aurora be primarily used for? I am worried that given current priorities, the very computation-intensive development of new nuclear weapons is likely high on the list. Trump has indicated that developing and deploying new nukes is high on his agenda, and as the lead agency for both nuclear weapons development is also the key funding agency for Aurora, this is very likely.
7
@Pete in Downtown One thing you can be sure of: this is NOT the cutting edge computing platform our government has developed, not if it's here in this paper.
Years ago, in the military, I ran into an officer who laughed at the current "state of the art" commercial computers "We had that 10 years ago!". I'm sure that's still true.
1