Slashdot Mirror
Earth Simulator Sees Green Light
burbs writes "Big Blue's dominator is getting closer to being turned on. The Earth Simulator in Japan is, supposedly, the world's fastest parallel-processing supercomputer. Designed for the Earth's weather, the computer should be able to predict climate for the entire planet for thousands of years in a short amount of time."
"Its not important" a scientist was quoted as saying. "I mean, its not perfect, but its close enough."
Besides, to be able to precisely model the earth's climate, they would need to have measures for about every (mathematical) point of the earth at a given time, which is not possible... Unless they go for an approximation, and then chaos theory kicks in and their 'thousands of years in advance' prediction is worth nothing. (the butterfly - hurricane thing anyone?)
Am I missing something there?
One shall speak only if what one has to say is more beautiful than silence
1000 Years? I dont think so.
The weather is a chaotic process. that means little errors in the input of any weather formula produce big errors in the output. Over 1000 years, even if you know the exact position of every molecyle exept one, this one will produce such a large error, that the results are unuseable.
Or did i miss something?
hmmm. well, i think i don't get it right here... please tell me where my logic is wrong.
1. the earths climate is dependant on the waether
2. weather can't be exactly forecast because of the chaos theory which says that in wheather there is an infinite number of variables.
so, either this earth computer can deal with infinite numbers of variables ot it just takes a wild guess... (predicting waht the climate will be thousand years is easy, i say it will be cold - proove me wrong in 1000 years)
massive parallel computing is a cool thing, but there are certain things that can't be solved with computers and i think weather forecasting is one of those things.
".Sig Stealer" was here
Predicting the weather for more than a month in advance isn't a matter of computing power. You could build a planet-sized computer and still not be able to predict whether it's going to rain in 30 days time.
The problem is not how well you process all the data: it's a question of finding all those bloody butterflies and stopping them from flapping their wings.
There has been a major scientific break-in
I'm old enough to remember when discussions on Slashdot were well informed.
you utter twat
Its easy to predict the weather from this: since it has so much processing power, once they turn it on, it will become self aware, declare war on all mankind, and launch the all missiles, thereby plunging the world into a nuclear winter that will last a 1000 years.
-- We don't understand software, and sometimes we don't understand hardware, but we can *see* the blinking lights
I'd be happy for next week. Here's a prediction: It'll be getting colder in the months to come. Japanese computer will maybe add another day to the current (and low quality) three day forecasts. Or more likely some unforseen factor will make it not work at all. Such blatent claims of success against a foe as worthy as chaos itself, indicates flawed method of thought that probably extends to all facets of their operation.
I remember one day when it was blazing hot sunshine in the morning, afternoon it was pouring with rain, then it snowed in the evening.
Only in Britain i tell you.
Thats why the British always comment on the weather, because you never have any idea what its going to be like, also the 'be prepared for any condition' attitude.
heh.
it's supposed to predict the climate
Oh my god, just imagine they would use this machine for something really useful, like, say, SETI? No?
This sig is stolen from someone who had a much better idea than I had.
Seems that if each one of these wonders requires a "city-scale" power generation plant that they would have to model themselves into the equation too.
Result: Global Warming is indeed occuring, but apparently it is mainly IBM's fault.
* Winners compare their achievements to their goals, losers compare theirs to that of others.
I suppose if it "sees green", that's the outcome we were hoping for.
But it must be one heck of a computer, to see the result of the simulation before they ever power it up.
Sheesh, evil *and* a jerk. -- Jade
I wonder if it takes into consideration how man affects climate... Will it be able to predict new technologies in 100 years that could alter the earth's climate forever?
WWJD.... for a Klondike bar?
It's called the planet earth. My calculation will be done in 1000 years.
So.. it will try to predict weather around the globe for years... uhuh.. sure...
Does it take into account unforseen disasters that will change the nescesary variables.. for instance vulcanic eruptions or global warming which is not predictable at all.
I would like to know what they do about these things so they can really predict things...
The article also states that the supercomputer can and probably will be used for all kinds of different modelling/simulations.
It also says that most software is probably flawed for now but it won't be for long i guess..
Will it be possible to open this baby for all kind of researchers all over the world instead of only a few japanese research groups?
Cached copy at Google: http://www.google.com/search?q=cache:www.feer.com/ 2001/0109_20/p034innov.html
A number of posters are confusing climate modelling with weather prediction. Weather prediction -- working out if it will rain tomorrow -- is very difficult because weather systems are chaotic. Climate prediction, however -- working out how large an effect increased CO2 emmisions will have on global warming -- is easy by comparison... at least in theory.
The problem with climate modelling is that the models right now incorporate large numbers of "fudge factors", and by setting those appropriately you can get whatever outcome you want from your modelling. Of course, without those fudge factors the Earth would be somewhere around -40C most of the time, so you can't just throw them out.
In short, good models exist for weather, but weather is chaotic so you can't predict much anyway. Bad models exist for climate, but at least it isn't chaotic (as far as we know).
Tarsnap: Online backups for the truly paranoid
predicting the weather is nice, but how fast does it do a kernel compile?
My guess is that "city-scale" is an exageration of the reality.
They have 640 nodes, let's say 2000 watts per node, that will take 2.5 MW. That's the power for a really small city (I'd say less than 15,000 inhabitants)
-Yep imagine a beowulf of this and them the global warming caused by it, "Earth Simulator will require the undivided attention of a nearby city-scale power station".
Does it calculate 2 futures, one where it is kept running and one where it is turned off to conserve power?
-I could be working but it's not in my job description.
even less, in fact... :)
damn MS-calculator
Forecasting the weather for 1000 years? This can only be slashdot-typic journalism.
Ever heared what metrologists call the butterfly effect?
The puff a butterfly makes during flight will alter local weather a little, and this change will continue to influence in weather mechanics, until some months later this butterfly can originate an tornade on another continent. This is a very common example used in metrology.
I suppose they mean they will calculate -climate-, and this only for the next -decades-, instead of 1000 years. I remember that these climate modules also had the difficulty that very small changes in input can dramatically change the output, like the emulation cell size. It's not that the output of the simulation grows more accurate as the cell size (of sky) grows smaller, the way it is is that the output changes totally on the choosen size in a chaotic way.
--
Karma 50, and all I got was this lousy T-Shirt.
Tom.
Oh arse
a beow... oh wait it is a cluster of powerful computers
it was a bit slower, and it was alway a day late - but it was dead accurate.
We do not live in the 21st century. We live in the 20 second century.
It would be interesting if these computers were used for something like the Metaverse in Neal Stephenson's book Snow Crash. i.e. Virtual Reality you can walk around in. The weather would be realistic at least!
Rohan
Permutation City is a story about the computation of a Virtual Reality environment down to atoms and molecules, complete with Artificial Intelligence, without the computer actually needing to be running! A very interesting read.
Rohan
This is an incredibly powerful machine. In fact, when they tested it a while ago, there way a programming error, and before they succeeded in turning the thing off, it had caused a heatwave in much of Europe!
I have it on good authority that worms, maggots, and other parasites are feasting on Douglas Adams's eyeballs and penis as we speak. He is a rotting pile of putrifying flesh. Oh! and the stink!
"Big Blue's dominator is getting closer to being turned on."
Ooh, can I do it? I like to turn on dominators!
Sticks and stones may break my bones
But whips and chains excite me.
--TheOrangeSquid Is it any wonder things seem so awry? We swim in a sea of confusion and don't have to think to survive
- don't worry. What they are talking about is climate, not weather. It's a bit like predicting that this winter is going to be colder than this summer - it doesn't require as much precision as predicting the weather: when it will rain or snow, the windspeed, cloud cover etc etc.
They must be talking about predicting the weather for L.A. for the next 1000 years.
That's easy. Sunny.
What the "nodes" consist of - the article wasn't very specific - two multi-processor PC's per rack cabinet - sounds a lot like the RS-6000 based "nodes" that ASCI White uses. Those are two per rack cabinet, each with 8 CPU's and a shared memory structure - sounds similar.
:)
As for 1000 years of weather prediction, fine, but can it accurately predict the weather TOMORROW is what I want to know
-- You are in a maze of little, twisty passages, all different... --
They are not trying to see if there will be 3" or 4" of rain in 2509. They are determining things such as major climate shifts. We know that the northern part of Aferica used to once be covered in grasslands...now it is harsh desert. We know that iceland used to be much more mild than it is now (there are still many green plants frozen stiff deep under ice). We know that the climate has a tencancy to shift over time affecting things such as rainfall patterns and temperature flucations. These are not caused by a butterfly's wings, these are caused by things such as the earth's tilt/wooble over the course of time. Wouldn't it be interesting to understand the "big picture" more than we do now?
Sig Return: 204 No Content
reminds me of a joke regarding maps. The people wanted the most accurate map possible, so they made it on a 1-1 scale. Only problem was they had to unfold it over the country to look at it...
J-aims
--
Yo, whatever happened to peas? Join T( H)GS
imagine a beowulf cluster of those things!! w00t!!
does it run linux too?
Reading the article, it seems the 1000 years bit is just looking at trends ( thats ignoring the typical journalistic simplistic spin ).
If you put a kettle filled with water on a stove, you know that in x minutes it will boil. You don't need to know the exact position of every water molecule ( except of course, the position is somewhere inside the kettle ). That is akin to the 1000 year forecast.
If you want to calculate all the eddies (sp?) and convection currents within the kettle ( ie like an accurate daily forcast ) you do need to know positions and state of each part of the water mass, down to molecules if you want to be realy accurate.
Of cause, rounding errors will probable give greater problems.
WINDOWLESS environment. Does it mean what I think it means, does it ?
Some like it with bugs..... I don't!
The weather thing seems to me to just be an advert to the world that there is a new giant abacus. The comments about other stakeholder interests like weapons development, nuclear physics and biochemistry research taking a lower priority has a cautionary tone for cloud-watchers. More noteworthy than whether Kasparov's great grandchildren will need sunblock, is whether the mutants who survive the nuclear blast of the Bush-o-tron 3000 will have 2 or 3 finger mittens.
~~~
"Well-washed and well-combed domestic pets grow dull; they miss the stimulus of fleas."
They think that politicians, who can't predict what will happen in the next few months, and scientists, who can't predict what will happen in the next few days, can forecast the weather for the next few thousand years just because one finances the other's uber-expensive simulation software.
Its amazing that the public never suspects this scam.
"Look at me, I invented the stove!" -- Ben Franklin
Isn't this what quantum computers are designed to do? I tell ya, screw conventional computing and focus this computer effort to creating a better quantum computation effort.
Eventually, we will have computers that will be able to figure out every course of action from any other given course of action... and the results will be orders of magnitude faster.
--donabal
Safety First Day?
I have read that some of the earliest data points are from before world war 2, when we left a lot of bouys (sp?) to track the directions and temperature of the currents for trade. They were picked up decades later. Most of todays data, of course, probably comes from sattelite.
Wouldn't it be more usefull to be able to predict the weather for a location 2 days in advance with a greater than 50% accuracy?
As x approaches total apathy I couldn't care less.
Software willing, the Earth Simulator will accurately model the earth's oceans and atmosphere by calculating weather data collected by various, land, sea and space-based sensors at 10 kilometre-spaced points around the entire earth.
Perhaps. But my pentium 100 can do the exact same thing in near real time.
Don't know about your country, but using one of the fastest computers here in denmark, the danish weather institute can't predict the weather 100% correctly even a week ahead. This new computer can't be THAT fast!
Thomas S. Iversen
Designed for the Earth's weather, the computer should be able to predict climate for the entire planet for thousands of years in a short amount of time."
It is good to see that the climate for the coming millennia can be predicted in a short amount of time, since the calculation will have to be repeated every 4 to 5 days. Chaos is such a party pooper.
Scary though: Imagine getting new weathercasts each day, predicting for the coming millennium. That weather girl had better be pretty gorgeous to keep it interesting.
I intend to live forever, so far so good.
March through November: rain
December through February: cold rain
I'm a leaf on the wind. Watch how I soar.
There is a principal in control system theory that states that for a model to be accurate the model must be at least as complicated as the thing it is modelling. e.g. to truely model the universe you need something as complex as the universe. The same thing applies to the weather system.
Being chaotic does not (in principal) deny the possibility of modelling. In fact we already have an operating implementation of the weather (look up).
The weather is modelable, a distinction should made that modelable does not mean guarenteed accurate. Even if this system says that there is a 75% probablitity that it will rain in spain (mainly on the plain) tomorrow then that is still worth having.
It seems unfair to dismiss work like this so quickly just because it is inherently impossible to predict with 100% accuracy.
Carpe Daemon
Two words:
Power Ball.
'nuff said
People automatically assume a warmer climate is bad, something that can't exactly be claimed given we have no good scientific model for weather and climate modelling. Not enough understanding exists, and by some theories which are quite compelling (for example, read the book 'Chaos'), it may never be possible to predict the weather.
Now, a warmer climate might be bad for a specific area or populace. Islands will probably dissapear as oceans rise, and coastal areas will be changed. There's more evidence that some areas might become wetter and better for agricultural production; that some marginal farmland might produce much higher yields, and that previously inhospitable areas in northern climates might become much more temperate. Of course, by the same blade, storms will possibly be more frequent and of higher magnitude.
We just don't know. Global temperature change is an inevitable result of modern civilization. It's entirely possible that we're headed for huge disasters as a result of the dominance of man, and there's nothing "wrong" about that. We just need to develop technologies that prolong our stay here as long as possible until we can do something else. There -are- 6 billion people here, and most models project it stabilizing at around 12-20 billion. That's also a lot of minds working on the problem.
..don't panic
So it's a Network of computers keeping track of the atmospheric conditions, like say the Sky.
Hmmm.
GCM d+ s+:+ a- c++ U? P! L E-- W++ NM+ V PS- PE+ Y+ PGP- t 5+ X?+ R+++$ tv+ b+ DI++++ D---- G e
The weather in Afghanistan tomorrow is expected to be sunny in the morning with increasing mushroom clouds by afternoon. The temperature looks to be a moderate 2000 degrees with cool winds upwards of around 700 miles per hour. It will definitely be a day for the sunblock, and it wouldn't hurt to shake the dust off that old lead suit in the closet. If you're planning on venturing outside in beautiful Afghanistan tomorrow, don't forget to drink plenty of fluids, such as barium which shows up nicely when blasts of radiation flow through your body. Most of all, have fun out there in dusty Afghanistan, and enjoy the country while... well, while it's still there!
As my Geography teacher never tired of telling us, climate != weather.
Climate is big and long-term. Weather is here and now. Not even the people who built that machine think it can predict world weather for one thousand years. There's just been a bit of a misunderstanding.
but meanwhile BSD is still dying !
OK, from what I understand of chaos theory, the 'butterfly' affect as it is called is about something small and overlooked that has a capability of affecting the system to a great degree over time. Fair enough. There is also a part of it (It has ben a damn long time since I have studied chaos math) that states the larger the system and longer the time the more of these events are likely to happen and counter eachother to some degree leaving the general trend the same. The butterfly affects abnormally spike periods of the system not the entire 1000 years.
As x approaches total apathy I couldn't care less.
Hey. Wasn't The Matrix an earth simulator? Should we be worried?
LOL. Thanks for the cheerful earful.
Designed for the Earth's weather, the computer should be able to predict climate for the entire planet for thousands of years in a short amount of time.
If they predict it ahead of time, we won't be surprised when it happens! Please write your representatives and ask the government to stop manipulating the weather!
"Climate is what we expect. Weather is what we get."
PHEM - party like it's 1997-2003!
I dunno about this one. Wouldn't you be solving some pretty harsh nonlinear d.e's in order to model weather correctly? AFIAK, the only real way to solve these things is to make assumptions about the boundary conditions and the parameters in the equations that simplify the equation in such a way as to make it solvable by means we do know. Not only that, you have mixed fluid equations - one equation for each type of particle in the atmosphere. You'll have boundary conditions on each layer of the atmosphere, and where things behave like plasmas you'll have an entirely new set of equations to consider. It sounds like there are way, way, way to many considerations that go into an exact solution of an atmospheric model. So, solving a 1000 years in advance seems ridiculous. I'd be happy if we could solve a day in advance!
Anyone ever seen the 13th Floor?
This sounds quite familiar to that....an earth simulator...hmmmm.
Yes, my girlfriend is a BitchX
correct me if I am wrong but wouldn't the climate be altered in case of an atmospheric nuclear explosion, somewhere ?
In such case, declaring that IBM's machine may predict the world weather thousands years from now would mean it could also predict nuclear conflicts ?
OK, now let's be serious, I believe that by mentioning thousands years from now, the author's intended to explain us that the dominator (isn't that a motorcycle name ?) could reajust its parameters to process new long-term forecasts in real time. Am I right this time ?
Trolling using another account since 2005.
I believe how it works is that every 1000 years, the Earth Simulator comes out of his hole. If he sees his shadow, then it is 1000 more years of global warming.
Washizu
OddManIn: A Game of guns and game theory.
Now, that's geek stuff...
Uhh, theres a flaw in that soup.
"...I'll need guns" --Chow Yun-Fat in 'Replacement Killers'
BTW, it's not Big Blue's dominator. This machine comes from NEC and i believe is based on the SX-5 supercomputer.
The NEC SX-5 has the fastest CPU's in the supercomputing world. The main innovator for NEC Tadashi Watanabe is known as the Seymour Cray of Japan. Currently Cray has an agreement with NEC to resell the NEC supers, which are one of the only parallel vector machines still being produced. Cray stopped producing it's own version of these classic large PVP the Cray T90, and is now concentrating on the parallel, multithreaded and smaller PVP machines. Which left a gap at the high end, so Cray made the resellers agreement with NEC about 8 months ago
This machine should be able to acheive a much higher percentage of peak performance in production codes because of the huge memory to CPU bandwidth and because of using a smaller number of very very fast CPU's, there should be less parallel sync overhead. This should reduce the non-parallel times (see Amdahls Law).
-- Straights are for fast cars, corners are for fast drivers.
My local weatherman can't give an accurate precition of tomorrow's forecast, now I'll be able to get an inaccurate forecast for Australia too.
Now that's technology!
'Life is like a spoonful of Drain-O, it feels good on the way down but leaves you feeling hollow inside'
Gee, you mean like this?
I think what you mean is that chaotic systems have to be non-linear (note that not all non-linear systems are chaotic). Also, you do not need differential equations.
Basically, you can define chaos as sensitive dependence on initial conditions. Consider the following very simple example:
Xn+1= A*Xn*(1-Xn)
This is known as the Logistic Map. You can think of it as a population model where X0 is the initial population, Xn is the population at generation n, and A is the rate of growth. The (1-Xn) term models death in every generation.
Now Play with this:
start with x0=0.1 and A = 4 and iterate
the X values look something like:
0.1
0.36
0.9216
0.28901376
0.821939226
for 5 generations.
Change X0 to 0.11 and rerun the simulation
0.11
0.3916
0.95299776
0.179172118
0.58827788
There's your "butterfly effect." No matter how little you change X0, the solution will diverge.
Just imagine this with several thousand more variables!
(Note: play with A and you'll see that some cases of this system are not chaotic)
Now IANAMeteorologist (I only run computers for them), but predicting climate change is not a sure thing, either. There are lots of basic parameters that we don't have hard numbers for yet.
One example: the amount of energy released by rainfall in the tropics. First of all, we don't know exactly how much rain does fall in the tropics. We have some gross estimates, but most of these are derived from land-based measurements, yet most of the tropical region is ocean. Then use satellites, you say. NASA is already running an experiment to attempt to measure tropical rainfall (TRMM, if you're interested). One of the major problems in any experiment like this is determining ground truth: comparing what you see from miles above the earth to what is actually happening on the ground. Ground radar is used to make comparisons, but mapping radar return to rainfall is yet another ground truth exercise: there are hundreds, if not thousands, of measured Z-R relationships (mapping reflectivity to rainfall). And the tropical energy budget is just one parameter. I'm not a climate modeller, but I'm sure there are hundreds of parameters more.
Of the many uncertainties we think we know how to quantify, there are many others we can now only begin to take a stab at. How exactly does one measure the amount of carbon dioxide exuded by biological matter decaying in a square kilometer of forest? If you're making measurements, how much do you attribute to living flora and fauna? If you think about these problems in enough detail, you see the difficulty.
Tiny errors in fudge factors are magnified by years of projection and extrapolation so as to be nearly useless for long term prediction. What will be useful for the short term is discovering where the model sensitivities lie. It sure would be helpful to see what climate looks like a thousand years from now if you tweak the CO2 parameters a fraction of a mole...
I left my sig in my other pants.
a) forecasting weather will be thousands of times more accurate
b) they can run climate simulations for a thousand years in three days.
running climate models for 1000 years is not equivalent to predicting the weather for 1000 years.
where i live (colorado), we have climate (every day the same, hot and sunny in summer, cold and sunny in winter, with occasional precipitation). where i used to live, in u.k., we had weather (hot and sunny in the morning, arctic gales by 3 p.m.).
p.s. can we get rid of that flashing gif for planetharddrive.com please?
Just my two cents, I know very little about meteorology, but here goes:
Surely, the weather patterns that occur can only do so naturally or as a result of pollution, any outside influence, e.g volcano, effect the weather in a knock on effect (from what I understand) in the area of affect.
So how can this machine and software give us any more of an understanding of the future weather conditions?
Me opinion in the pub, based on the weather yesterday and last week on the same day has the same value as a 400m computer?
If it is working on intangibles (which it has to) then what is its true value?
Come on people, this thing hasn't been made to guage the climate for the next 1000 years.
Its power *could* be better utilised working out the relay times on super kryton switches or radical chaotic dispersal.
Weather prediction, and the general forecast we see is based on past performance, obvious trends and what people want to make of it.
Its nothing more than fortune telling IMHO.
Just my two cents, I know very little about meteorology, but here goes:
Surely, the weather patterns that occur can only do so naturally or as a result of pollution, any outside influence, e.g volcano, effect the weather in a knock on effect (from what I understand) in the area of affect.
So how can this machine and software give us any more of an understanding of the future weather conditions?
Me opinion in the pub, based on the weather yesterday and last week on the same day has the same value as a 400m computer?
If it is working on intangibles (which it has to) then what is its true value?
Come on people, this thing hasn't been made to guage the climate for the next 1000 years.
Its power *could* be better utilised working out the relay times on super kryton switches or radical chaotic dispersal.
Weather prediction, and the general forecast we see is based on past performance, obvious trends and what people want to make of it.
Its nothing more than fortune telling IMHO.
Imagine a beo of these babies... if they only got maybe three days right, it would be enough for now..
If it can track all the butterflies in tokyo and predict the tornadoes in Texas at the same time, it's got my tax dollars. http://www.cmp.caltech.edu/~mcc/chaos_new/Lorenz.h tml
BTW, if it's able to complete this task in 'a short time' what will the computer be used for after that? 42.
Kind thoughts do not change the world
Climate modeling SOFTWARE AND ASSUMPTIONS to date have been a pathetic group of politicaly driven desired-result-generators intended to support the latest alarmist's demand for control. When the known parameters of prior years is input to these programs they do not come even close to "predicting" recent climatic history. So why do people still think a HARDWARE upgrade is going to make these bad assumptions and fudge-factors work out correctly for 10, 100, or a thousand years into the future?! A cheap computer can repeat your questionable algorithm and self serving assumptions a million times a second. A rediculously expensive computer can repeat the same ten-trillion times a second....so?
Any technology distinguishable from magic is insufficiently advanced. - Geek's corollary to Clarke's law
It's been estimated that if you took the pressure, temp, humidity, and wind velocity every cubic foot for a hundred miles up, you could only reasonably predict the weather 30 days in advance.
Climate is the average temp, rain, etc. over a period of decades, if not longer.
I am for the complete Trantorization of Earth.
*ducks*
Not to mention chaos theory. The CPU fans in Japan can cause hurricanes right around florida...
Good quote, too many chars. Seriously, the slashdot 120 char limit sucks!
Chaotic? Oh come on. It's only fluid dynamics! ;-)
By networking hundreds of old farmers together...
Man, I can tell today is going to be a long one...
crazy dynamite monkey
Imagine a Beowulf of these :)
But most fortune tellers would love to make the money these guys are making. At least the not so good ones. To live off the public trough in the guise of science. What a wonderful way to live.
At least these guys answer to massively parallel computing is better than the one I got from the computing center manager for Westinghouse Bettis 8 years ago. He showed us his 30,000 processor computer. We oohed and awed. He then turned and said "this is our new one, it has 120,000 processors."
When asked "Can you use the 30,000 processor one yet?," he replied "No".
Brad
God: "I don't leave footprints!"
A useful and interesting application for a massively parallel supercomputer such as this would be to run a simulation of the WTC attack and possibly pinpoint where the black box recorders are, and if they survived.
can you imagine a beowulf cluster of these?!
Does this factor in the global warming caused by the computer itself running?
Winter. Nuclear winter. Yeah I need a computer to tell me that.
After I have received the wisdom of good teaching, I will untiringly teach all people. - The Teachings of Buddha
Every time there's a hurricane, we can blame it on one of those overclockers that put 2 dozen fans in their computers. Or punish them for causing it. The punishment: if they have a faster computer, they give it to me. :)
it clearly should be used to run quake!
If these great tools are available, why don't these scientists actually do something useful with them, like figure out exactly where and how much ozone to inject into the atmosphere to fix the hole? That would be much more useful than finding out if its going to rain sometime in 3545 AD.
Every single one of you who mention how "the chaos theory" prevents this from working, need to shut off your computers NOW. Stop pretending you know what you're talking about--you don't. Pick up a good book on chaos (I reccomend "Does God Play Dice?" by Ian Stewart but plenty of others will do). Read it. Now you may turn your computer back on and come back.
If you're incapable of doing so, allow me to explain. There is no such thing as "the chaos theory". Stop saying it. Chaos is a mathematical concept, a way of categorizing things, not a theory. Certain things are chaotic, others are not. Weather, for example, is chaotic. This in no way means its unpredictable, it just means we can predict only probabilities in the long term. Another example would be the movement of gas molecules. We can never tell exactly where all of them will go. However, there are billions upon billions of them, so we can use statistics to get an overall picture of them that is extremely accurate. Perhaps you're starting to see the analogy here: weather and climate are the same way. We can't accurately predict each day's weather, but we can predict the statistical sum of all of them (climate). Not quite as accurately as we can with gases, but its still possible to get a fairly high degree of precision.
Let's see - 320 cabinets, 2 nodes per cabinet, 8 processors per node gives 5120 processors. And the stated capacity is 40 teraflops - 40,000 gigaflops. So that's almost 8 gigaflops per processor.
Now I'm trying to figure out what processor can actually give 8 gigaflops. It's not 2GHz P4 or 1.6GHz Athlon, they're about half the power needed even at those clockspeeds. Itanium has multiple way issue, but the clock speed isn't there.
For that matter, with any of those, the power consumption would still be under 1KW per processor - and 5MW is NOT
a 'City sized power plant'.
Maybe the Alpha isn't dead - doesn't API have a new version being manufactured by a Japanese company that might meet this speed?
Or perhaps IBM's 4 processor on a chip Power series might be counted as a single processor by the people writing the article? (That would bring it down to 2 gigaflops per processor - which is in the right ballpark.)
Liquor
Sanity is a highly overrated commodity.
I bet I can predict the weather for thousands of years in a short amount of time also, without using a supercomputer. I can also predict that the world will end on March 17th, 2309 at approximately 4:58pm PST. It will be a partly sunny day with chance of showers and a high around 73 degrees farenheit in San Diego, CA. It will be warm and dry in Arizona.
It doesn't matter how sophisticated the equipment is; you can use a bag of specially sanctified stones, a deck of magical cards, or you can use the world's most powerful supercomputer. The fact remains that meteorologists are the soothsayers of our time.
You had me at "dicks fuck assholes".
how many parallel processors are needed to find the meaning of the universe then?
I'm sure it'll all run perfectly well until some idiot starts hitting the "tornado" and "fire" buttons over and over again. They'll probably even end up destroying that Mayor's Statue I had to work so hard to get.
That green slime had it coming.
does this thing give off? And does the climate model take this heat into account? And if this heat depends on the amount of computation done by the computer, you'd have to model that too. Ahh, my brain hurts.
JET Program: see Japan, meet intere
How can a computer predict the behavior of a complex system such as the Earth? I can see a simulator such as this being somewhat useful in predicting large-scale phenomenon such as global warming, but I don't see it predicting, for example, next year's Atlantic hurricanes, with any reasonable accuracy.
We have about as much chance of seeing intelligent androids such as those in A.I. as we have a machine that can predict the weather a thousand years out.
Republicans are idiots.
The guy that posts a bit of satire or the guy that anonymously posts that the satire poster is a twat? Back 'atcha! Twat!
In fact, in the atmospheric community we use the nonlinearity to our advantage by doing something called ensemble modeling. Running ensembles is just a fancy term for multiple simulations with slightly perturbed initial conditions and/or model parameterizations.
As long as our model assumptions are not too far off, then the ensemble results give us statistical information about the forecast (i.e., most probable value, confidence intervals, ...).
With respect to climate simulations, we run ensembles corresponding to various greenhouse gas scenarios. The results yield the most likely temperature change plus/minus our uncertainty in this temperature change.
The Earth Simulator in Japan will help out immensely for two reasons:
(1) We can run more ensembles, which helps us to better define the model uncertainty.
(2) We can go to higher spatial resolutions to better capture the nonlinear processes in the model (such as cloud related processes).
Modeling the Earth. Sounds familiar...
When they run this thing they will find out that the answer is 42.
As long as our model assumptions are not too far off, then the ensemble results give us statistical information about the forecast (i.e., most probable value, confidence intervals, ...).
With respect to climate simulations, we run ensembles corresponding to various greenhouse gas scenarios. The results yield the most likely temperature change plus/minus our uncertainty in this temperature change.
The Earth Simulator in Japan will help out immensely for two reasons:
(1) We can run more ensembles, which helps us to better define the model uncertainty.
(2) We can go to higher spatial resolutions to better capture the nonlinear processes in the model (such as cloud related processes).
In my not so humble opinion, all we'll see at first when the Earth Sim is fired up is that the best climate evolution models are awfully wrong. The nice thing about this machine is that it's powerful enough to run a large number of model variations and see how they produce the expected result. But the researchers will follow this time-honored pattern of numerical simulation:
The problem is that every of these steps is hard. Step 1 requires a very accurate, very fine-grained depiction of initial parameters. Even today, we just don't have that! In meteorological (weather forecast) computing, the initial condition matrix is created by complex non-linear interpolations from a limited set of observations that also are not all collected at the same point in time. Creation of this initial condition data set takes actually longer than the weather forecast run! This is a major hurdle. Now, can you guess how such a data set would be prepared for, say, year 1800, where weather stations were rether scarce? Yep, that's right, by guessing. Which means the very accurate computations will be based on very rough guesses. Discutable results at best.
Then, the "apply model" run. Hah. Our climate models are, to say the least, inaccurate. We don't understand the physics of climate yet. Also, we assume some things like "the solar constant", which says that the amount of energy coming from the sun is constant. It's not. You have variations on an 11-year cycle (solar spots) and a bigger variation on a 208-year period. Oh, not much, it varies by less than a percent. But it's enough to strongly affect the climate. See paper "Solar Forcing of Drought Frequency in the Maya Lowlands" in Science of 18 May 2001 to see how this solar oscillation wreaked havoc in the Maya civilization. And last time I checked, this phenomenon was not taken in account by standard climate models.
Step 3 is easy... if you're comparing your results with modern-era climate. But if you are running a simulation on the evolution of an ice-age 120,000 years ago, it's pretty hard to check the result's accuracy.
Step 4 requires more money. And that's where the current hype about climate change and warming was useful for scientists (and not just for companies who market costly, unsafe replacements for the cheap, inert, non-toxic, banned freon). Climatologists have finally been able to get big funding in some places, instead of being an arcane, underpowered science.
The cautious ./er will remember, if he's old enough, that during the Seventies, the "experts" warned us about the upcoming ice-age. I remember reading sci-fi books about state-cities trapped in underground caves below the 2-mil thick ice and fighting each other for proteins and uranium fuel... The same experts now warn us about the upcoming warming. When should we believe them?
--
Mad science! Robots! Underwear! Cute girls! Full comic online! http://www.girlgeniusonline.com/
This isn't designed to predict the weather in Omaha next Tuesday... or tell you the weather in Texas 1000 years from now....
It's designed to analyze global climate change... which you CAN do, with a reasonable amount of accuracy.
The article says that NEC built this thing. Not a single mention of IBM. Why does the blurb say IBM? I don't know.
[|]
This sounds like a Computational Fluid Dynamics (CFD) program. If so, this is a subset of finite element analysis (FEA) widely used in industry. This is an incredibly powerful technique that is used in all types of engineering.
"Chaotic" effects aside, what the program seeks to do, is solve the governing equations of temperature and fluid motion with user applied boundary conditions. To do this, the physical model is broken down into a number of "finite elements" of simple geometry, for which an analytical solution is known. These are assembled into a large matrix, and then the computationally intensive task is to invert the matrix. The matrix may be hundreds of thousands of elements on a side.
The main types of errors in this type of analysis are:
1. Modeling error - wrong inputs or boundary conditions, or bad software
2. Discretization error - the error associated with breaking down a smooth functions into a finite number of separate pieces.
3. Numerical error - the round-off or truncation error in the numerical processing.
I hope they get good results!
Quoting from the article:
IN A WINDOWLESS building opposite a crumbling government-housing block
Don't tell me I'm the only one who is reminded of Hitch-hiker's by the claim of a computer that can simulate the EARTH!!! :-)
The bitter lessons of a veteran coder: http://bitterprogrammer.blogspot.com
Hey, this is a site for computer geeks. What are the cpu's they are using in this computer? what is the operating system?
A press release from NEC, dated May 30, 2000, is at <http://www.nec.co.jp/english/today/newsrel/000 5/3001.html>. There are also links to more current information, in both Japanese and English.
According to the press release, the system has 640 nodes with 8 vector processors each, for a total of 5120 CPUs. Each node has 16GB of shared memory and has a peak performance of 8 GFLOPS.
The system runs NEC's SUPER-UX Unix-based operating system, and supports Fortran90, C, C++, HPF (High Performance Fortran), and MPI2 (Message Passing Interface).
I wonder if they have thrown in the continental drifts into the equasion.... if they are going to predict thousands of years into the future, then that has to have an affect, yet I saw no mention of that....
I also wonder if we will see any earthquake preditions comeing out of this.
The Code Ninja is swift with his tool, precise in his delivery, and deadly accurate in his execution.
is indeed impossible.
As a well known phenomenon in climate simulation and prediction community, general circulation models, when running after a sufficiently long period, will have the syndrome of a so called "climate drift" effect. This effect is often caused by biased air sea interface dynamics and thermodynamics.
The heat flux/moisture exchange at the air sea interface has long been an active subject in the atmosphere ocean climate modeling community because of its complicated nature and lack of physical description and parameterization. Many GCMs choose to use either atmosphere or ocean component and use a fixed sea surface temperature/flux boundary condition to avoid severe bias in that matter. Coupled GCMs often suffer from inadequate boundary layer parameterization and often ad hoc fix were used to make the model produce realistic result for a certain amount of time.
For thousands of years? I can only hope the Earth will survive that long to embrace the fate these GCMs have predicted for them, a hell or a glacier.
Together, we are strong; Apart, we are stronger.
Designed for the Earth's weather, the computer should be able to predict climate for the entire planet for thousands of years in a short amount of time." if they knew the algorithms that determine the weather (and if no unknown variables are introduced)
...the computer should be able to predict climate for the entire planet for thousands of years in a short amount of time."
Okay, then what do you do with the computer? You spent billions of dollars on it, and it chugs and churns for 3 days straight, gives you the climate for the next 4000 years -- then what? Shut it down? Play Quake?
(And yes, I know, the climate will need to be constantly recalculated, so the machine will never be not used)
...What a dull name
Davo -- Free speech, free software, AND free beer.
Oh, I see, "feer.com" is the Far Eastern Economic Review. That makes more sense than it being some 31337 haxor rag, which was my first reaction.
Hrmmm who else thinks that Taco is probably drooling about running Slashcode on that machine right now... And I'm sure Tech Squares former AI Lab gurus wish they had it 'back in the day'... of course all I wanna do is see how many fps I can get on it...
Remove *your pants* to send me email.
"Designed for the Earth's weather, the computer should be able to predict climate for the entire planet for thousands of years in a short amount of time."
That's like the biggest load of bullshit ever! I'll give them the benefit of the doubt, and assume it's just another ignorant slashdot misquote.
Read it again, they said _climate_, not weather!
Climate is all about long running averages and suchlike. Weather is about exact low/high pressure zones, exact temperature/moisture profiles in the atmosphere at given times, and so on.
Climate models take into account long-running observations and averages and attempt to predict where the trends in the averages will go. Weather models attempt to predict where the aforementioned weather-related phenomenon go. Totally different ballgames. Weather cannot be _reliably_ predicted out beyond about 5 days at present (ie. with a 50% chance of being right), and will take a _long_ time to go out any further.
I have a Ph.D. in this area. There are certain attractors that weather patterns do return to in the current climate. However, in order to predict the weather more than 7 days in advance, we'd need initial conditions, globally, at a resolution of like, 1m. So, yeah, put a 1m grid of weather-station buoys out there across the oceans, across antarctica, throughout the Himalayas and the Atacama desert, over the polar ice caps and throughout the far north reaches of the North Atlantic (watch out for icebergs!). Then you'll be able to predict the weather like a week and a half in advance. Climate is the accumulation of weather, so good luck predicting it a millenium in advance. You'd need weather buoys and stations on a tighter than 1cm grid. Bonne Chance guys!