Climate Change Research Gets Petascale Supercomputer 121
dcblogs writes "The National Center for Atmospheric Research (NCAR) has begun has begun using a 1.5 petaflop IBM system, called Yellowstone. For NCAR researchers it is an enormous leap in compute capability — a roughly 30x improvement over its existing 77 teraflop supercomputer. Yellowstone is capable of 1.5 quadrillion calculations per second using 72,288 Intel Xeon cores. The supercomputer gives researchers new capabilities. They can run more experiments with increased complexity and at a higher resolution. This new system may be able to reduce resolution to as much as 10 km (6.2 miles), giving scientists the ability to examine climate impacts in greater detail. Increase complexity allows researchers to add more conditions to their models, such as methane gas released from thawing tundra on polar sea ice. NCAR believes it is the world's most powerful computer dedicated to geosciences."
What's the carbon footprint of this machine? (Score:4, Funny)
Hey look, when we model the city where the machine is, there's a hot spot. What could be causing it?
Re:What's the carbon footprint of this machine? (Score:5, Funny)
Re: (Score:3)
Here is a direct response from the MET Office [wordpress.com] on that subject. There's a nice graph at the bottom that ranks years from hottest to coldest colour coded by decade. To quote them:
Over the last 140 years global surface temperatures have risen by about 0.8C. However, within this record there have been several periods lasting a decade or more during which temperatures have risen very slowly or cooled. The current period of reduced warming is not unprecedented and 15 year long periods are not unusual.
15 years of data is simply too short a time to make definitive statements about warming in the face of natural variability.
Re: (Score:2)
No climate models "successfully predicted 15 years of stagnation" because that's not what they're designed to do. In fact you probably couldn't write a model that would be successful at predicting a 15 year stagnation because of natural variability. The climate models I'm familiar with generally use 30 year averages for their projections and have since the 1980's. A paper from last year [agu.org] statistically analyzed the issue and found it requires at least 17 years of temperature records to separate the signal
Re: (Score:2)
The hot spot will only be actual heat generated which will probably be on the order of a small town. The electricity generated to run it may or may not be from non-CO2 producing sources (hydro, nuclear, etc) so that could possibly up the CO2 output at the generating station or on the grid, or not.
Congrats to NCAR!
Re: (Score:3)
Re: (Score:2)
*golf clap*
Climate research vs. weather prediction (Score:4, Interesting)
All this computer power is going to climate study/prediction, while weather prediction is limping along with .07 petaflops. See much more discussion on the topic here: http://cliffmass.blogspot.com/2012/05/us-climate-versus-weather-computers.html [blogspot.com]
next step is weather control and you need to resea (Score:2)
next step is weather control and need to research in a lab setting be for taking it full scale
Re: (Score:2)
With the amount of energy that thing will release? I'd say this step is weather control...
Re: (Score:2)
We'll need a white fuzzy cat supply.
Re:Climate research vs. weather prediction (Score:4, Informative)
Re: (Score:2)
All the good that does them. They're still wrong more than half the time, that's worse than Environment Canada.
Re: (Score:2)
Re: (Score:2, Informative)
Why is this modded down? (Score:1)
Why is this modded down while the original post at +5 is wrong? The computer is not going to be used only for climate modeling, just the media isn't going to get all thrilled to talk about more mundane uses like weather prediction and research.
Re: (Score:2)
Random numbers that fit your model, yes.
Re: (Score:2, Insightful)
Weather prediction = 7 days (mostly wrong)
Climate prediction = 1000 years (no idea on accuracy)
Gee I wonder why you would need more processing power ...
I tidied that up a bit.
Re: (Score:2)
Re: (Score:2)
Weather prediction is mostly limited by the amount and precision of data and the fact that it's impossible to accurately predict a chaotic system after a certain point, not computing power.
Re: (Score:2)
With all the money invested into this they better find out that Global Warming is real and Man made otherwise they'll have the budgets cut pretty hard.
Yeah, but... (Score:1)
... how many apps does it have?
Re: (Score:2)
and how much thinner is this new version?
Re: (Score:2)
The keyboard on this one glows!
global warming (Score:1)
I wonder how much heat that machine emits.
Re: (Score:3)
Improving the speed of inevitability (Score:3, Funny)
NCAR - We confirm your still f#cked, only faster!
Re: (Score:2)
Re: (Score:2)
No it's coal-fired
Re: (Score:2)
Re: (Score:2)
You might want to look up mesh density and what it does for simulations.
The Ultimate Question (Score:1)
Re:The Ultimate Question (Score:4, Funny)
Sure, but Crysis only gets 3fps.
Re: (Score:2)
Does it run NetBSD?
In other news... (Score:5, Funny)
Climate deniers have rejected the results of the new, higher speed climate models in 3 femtoseconds, proving even faster than the new supercomputer.
Re: (Score:1)
Re: (Score:2)
But it won't. They're the same models, only faster. And its really easy to test them... put old data in, see if you can accurately predict what happened next. None of them can.
Well, sure you can! You have data from yesterday (literally) to plug in now. That'll change those results, it will!
Re: (Score:2)
In my opinion, they still won't be able to build an accurate model until they can calculate the interactions of every single particle in the solar system.
Quantum physics says even that wouldn't work. It's like predicting the weather, it'd be reasonably accurate for a short period of time and quickly deteriorate in accuracy at time went on.
Re:In other news... (Score:4, Insightful)
Show me ONE climate model that has accurately predicted anything, ever.
With impossibly high standards like yours, it's a wonder any other physical model still holds up.
QED, for instance, will never accurately predict where a photon is going to land, but it will give you the probability of a photon hitting a specific area. Probability is a huge part of science and no scientist will tell you anything is 100% certain. Very high certainty for a range of conditions is what a model is intended to provide.
Re: (Score:1)
With impossibly high standards like yours, it's a wonder any other physical model still holds up.
There's nothing particularly high about the standard of "accurate" prediction, although with climate models we're in the peculiar position of being asked to take the results at face value long before anything like a decent empirical check can be made.
The comparision with QED is entirely wrong-headed, because QED is an exact theory that can be used to compute results in its domain with astonishing accuracy and checked by controlled experiments, whereas climate models are a pile of approximations that cannot
Re: (Score:1)
I think his point is that most (all?) models that have been presented and taken into consideration as input for things like regulations, taxes, and other major social and economic changes, have predicted in no uncertain terms things that should have happened by now but haven't.
I'm fine with the fact that it's not an exact science but if you're going to turn the world on its head because of some doomsday predictions, you better not be several orders of magnitudes wrong with your figures. I don't think the p
Re: (Score:2)
It needs to take into account all the heat it generates and the CO2 produce to calculate the heat it generates and the CO2 produced to calculate the heat it...
That data is called "anomalous" and discarded.
And just how much heat does it generate?! (Score:3)
If climate scientists run a supercomputer in a room full of warming skeptics, does it give off any heat?
Re: (Score:2)
If climate scientists run a supercomputer in a room full of warming skeptics, does it give off any heat?
Which one generates more heat? I need a grant here; come on!
And, still... (Score:1)
And, still, it won't provide enough computational power to discriminate between natural phenomena and anthropogenic global warming.
Yellowstone - with that name, does it run Caldera? (Score:2)
Doesn't Matter (Score:3)
Re:Doesn't Matter (Score:4, Funny)
Oh, they're running things off of Oracle then?
Re: (Score:2)
Re: (Score:3)
They've got 72,000 cores, but their software license only allows them to use 2 at a time.
Bah, doomp, tsii!
Re: (Score:2)
I think dcblogs had a typo and meant to write "or".
Is that peta-monster green enough? (Score:2)
Quite disappointing (Score:2)
Bad math (Score:1)
1.5 petaflops is not "roughly 30x" 77 teraflops; it's just under 20 times.
Garbage in, garbage out (Score:1)
It doesn't matter how powerful your computer is, when you've got only the barest idea of the inputs and the parameters of your model, the output is still going to be crap. Assuming your model is any good in the first place, which is unlikely.
Re: (Score:1, Informative)
In other news.. no global warming for 16 years now...
http://notrickszone.com/2012/10/16/luningvahrenholt-comment-on-hadcruts-16-years-of-no-warming-tough-times-ahead-for-climate-science/
Counterpoint [guardian.co.uk].
Re:GW? (Score:4, Insightful)
Re: (Score:1)
What the hell is angaging?
Re: (Score:3)
Surprising nobody has identified the purchase price in a fraud lawsuit - yet.
Re: (Score:2)
Re: (Score:1)
I'm a little confused, since I am not an expert on climatology, but this [noaa.gov] [source [noaa.gov]] seems to suggest that the global oceanic thermal energy (aka "heat content") has risen. The point of my post is that "the temperature is constant" is only one part of a complicated issue. Your ice tea is "warming up" while it is sitting out, but its temperature stays constant as long as the ice cubes haven't melted. "I'll give you 1/4 success on that one."
Re: (Score:1)
Re:Maybe (Score:4, Informative)
About 4 times as much sea ice has been lost from the Arctic as has been gained in the Antarctic. The Antarctic ice sheets have been losing more ice than than the Antarctic sea ice has gained so the net there is still negative.
The gain in Antarctic sea ice is interesting. It has to do partly with the ozone hole over Antarctica and partly to do with global warming. The ozone hole causes stratospheric cooling which strengthens the circum-polar winds, blowing the existing sea ice around which opens up leads which subsequently refreezes. Global warming causes more precipitation which when falling on the ocean surface freshens the water making it less dense which reduces the mixing between the warmer saltier waters below and the colder surface waters reducing the ice melt at the surface and making the water easier to freeze.
Re: (Score:2)
The MET Office [wordpress.com] has refuted this story.