Sequoia Supercomputer Sets Record With 'Time Warp'

Nerval's Lobster writes "The 'Sequoia' Blue Gene/Q supercomputer at the Lawrence Livermore National Laboratory (LLNL) has topped a new HPC record, helped along by a new 'Time Warp' protocol and benchmark that detects parallelism and automatically improves performance as the system scales out to more cores. Scientists at the Rensselaer Polytechnic Institute and LLNL said Sequoia topped 504 billion events per second, breaking the previous record of 12.2 billion events per second set in 2009. The scientists believe that such performance enables them to reach so-called "planetary"-scale calculations, enough to factor in all 7 billion people in the world, or the billions of hosts found on the Internet. 'We are reaching an interesting transition point where our simulation capability is limited more by our ability to develop, maintain, and validate models of complex systems than by our ability to execute them in a timely manner,' Chris Carothers, director of the Computational Center for Nanotechnology Innovations at RPI, wrote in a statement."

Comments

Planetary scale computing?

[RandomFactor]

"enough to factor in all 7 billion people in the world, or the billions of hosts found on the Internet."

It's all fun and games until someone puts in a hyperspace bypass...

Re:Planetary scale computing?

[Anonymous Coward]

Yes. You've read a book. We get it. Thanks for letting us know.

This sounds familiar...

[NoImNotNineVolt]

I could've swore I heard about this less than 23 hours ago somewhere...
http://tech.slashdot.org/story/13/05/02/2119214/llnlrpi-supercomputer-smashes-simulation-speed-record
Yup.

Re:This sounds familiar...

[deains]

The system's so efficient it managed to travel back in time by 23 hours.

Re:This sounds familiar...

[Anonymous Coward]

... and no "Time Warp" pun? You're doing it wrong...

(It's so obvious I feel like they intentionally duped this just for laughs.)

Re:This sounds familiar...

[wolfgang.groiss]

It's just a JMPLFT and then a STPRGT

Re:This sounds familiar...

[ElderKorean]

I could've swore I heard about this less than 23 hours ago somewhere... http://tech.slashdot.org/story/13/05/02/2119214/llnlrpi-supercomputer-smashes-simulation-speed-record

Maybe they could use 0.000000001% of it's processing power to look for duplicate stories on here.

Spoiler

[DarthVain]

You live in a computer simulation.

Re:Spoiler

[slew]

You live in a computer simulation.

And so does this supercomputer...

That raises up the question of how can a supercomputer simulate what our species would do if it had access to a supercomputer that could simulate what our species would do if it had access to a supercomputer that could simulate...
Out of memory [core dump]
% rm -rf *

Which of course begs the question...

Psychohistory

[cmeans]

Finally, we can start computing where we're going to end up, and who/what we have to manipulate to end up somewhere else.

Damn you IBM!

[Anonymous Coward]

But what about OS/2 Warp?

Really?

[Dripdry]

The time warp? Again?

Re:Really?

[DougOtto]

I see what you did there.

Re:Really?

[deains]

10 LET JUMP = "to the left"

Re:Really?

[arfonrg]

20 LET STEP = " to the right"

Re:Really?

[ArcadeMan]

Maybe laserdiscs are making a comeback for some reason?

so.. how much power does it draw?

[wierd_w]

Other factors: based on the parallelism model used, and the current state of electronics, what is the maximum number of cores that can be included before performance degrades from additional nodes?

(Eg, it takes x time to transmit data over a bus (any bus). How may cores, before the time penalty for transmitting the data over the bus to the allocated processor becomes greater than the penalty for just waiting for a processor to become free?)

There *must* be an upper bound on parallel computing potential before we need pure unobtanium semiconductors.

I am curious what that limit is, and how close we are to it.

Re:so.. how much power does it draw?

[Anonymous Coward]

It's most likely dependent on the problem size / nature. Apparently, they were able to achieve super-linear scaling on their particular problem because more and more of the data structures they used fit into CPU cache. For problems that use less memory or more inter-processor communication, I imagine the sweet spot would be very different.

Re:so.. how much power does it draw?

[ceview]

Isn't the solution to this to use qubits? Just use a quantum computer techniques to 'transmit' the data? Though the computer probably would need to be running at some crazy cold temperature.

Re:so.. how much power does it draw?

[wierd_w]

That would be close to running on pure unobtanium. (Energy costs would become untenable to maintain the quantum states of many thousands of entangled particles, and keep them cold.) Also, quantum computing can only efficiently serve a subset of parallelized tasks, and are a poor fit for general parallelism as is. (Improvements may fix this in time however.)

The brick walls that I see looming are:

Energy requirements VS what can feasibly be delivered
Thermal waste generation rate VS rate of removal
Computational gain from parallelism VS loss from computational overhead

The first one has a snag: even a super conductor cannot conduct unlimited energy flow rates. Clenching happens, and when it does, things go very bad. Further, there are limits to what can rationally be sustained as an energy source. There is a looming restriction where we simply won't be able to provide the energy needed to continue performance increases.

The second one is related: as energy needs increase, the amount of thermal waste generated also increases, and the need to eliminate it efficiently increases. Materials lose valuable electrical and structural properties as they heat up, and rates of thermal exchange are not infinite. There is a fundemental limit where our computer produces heat faster than we can ever safely/reliably remove it.

The third I already explained.

Quantum computing eliminates obstacle 3 (for some applications) but rushes headfirst into the second two. The more qubits you use, the more energy needed to sustain entanglement, and the more aggressively it has to be cooled. This implies that there is a finite possible number of qubits one can string together before it becomes impossible to sustain more, (unless you discover a way to pull a Maxwell's Demon, and escape the second law.)

Re:so.. how much power does it draw?

[Anonymous Coward]

They're using Reverse Polish Notation.

Repost?

[Anonymous Coward]

It seems that this news has already been posted... like... today?

http://tech.slashdot.org/story/13/05/02/2119214/llnlrpi-supercomputer-smashes-simulation-speed-record

I for one...

[arfonrg]

I for one welcome our Matrix overlord... (or MCP... or Skynet, which ever achieves sentience first...)

I was hoping to....

[mark-t]

... be the first to make the admittedly very obvous reference to RHPS, but then I noticed the submitting department.

Re:I was hoping to....

[Anonymous Coward]

And you weren't the first either.

Simulate people?

[ArcadeMan]

Hard to do, with free will and all...

New Time Warp Protocol?

[Anonymous Coward]

So the protocol is no longer

It's just a jump to the left
And then a step to the right
With your hands on your hips
You bring your knees in tight.
But it's the pelvic thrust.
They really drive you insane
Let's do the Time Warp again
Let's do the Time Warp again

Now We Wait...

[FurtiveGlancer]

For the Russians to complete Guardian!

Colossus:_The_Forbin_Project

It's astounding

[houghi]

Time is fleeting;
Madness takes its toll.
But listen closely...

Not for very much longer.

I've got to keep control.

I remember doing the time-warp
Drinking those moments when
The Blackness would hit me

And the void would be calling...
Let's do the time-warp again.
Let's do the time-warp again.

It's just a jump to the left.
And then a step to the right.
Put your hands on your hips.

You bring your knees in tight.
But it's the pelvic thrust

That really drives you insane.
Let's do the time-warp again.
Let's do the time-warp again.

So shut up and make an AI already

[gestalt_n_pepper]

You've got the tools. You've got the know how (sort of). First one to intelligence wins the world, more or less.

So MOVE!

The Answer is...

[ceview]

Well you are not going to like it. Though I'm pretty sure they need to build this at the end of the universe too, just to be sure.

Clearly the big bang was a giant core dump

[rsborg]

You live in a computer simulation.

And so does this supercomputer...

That raises up the question of how can a supercomputer simulate what our species would do if it had access to a supercomputer that could simulate what our species would do if it had access to a supercomputer that could simulate...
Out of memory [core dump]
% rm -rf *

Which of course begs the question...

Questions I've posed to astrophysicist friends that have never gotten good answers:
1) Why should we not think of galaxies as simply accretion disks - ie, we're all circling the giant 10M solar mass black hole drain that's the center of the galaxy.
2) Why is it irrational for me to think of the big bang as basically the opposite of a black hole - and how do we know it's not continuing to spew matter?

Re:Clearly the big bang was a giant core dump

[Anonymous Coward]

1) Why should we not think of galaxies as simply accretion disks - ie, we're all circling the giant 10M solar mass black hole drain that's the center of the galaxy.

Being part of an accretion disk sort of implies that our best physics would predict that ultimate destination of the world-line of the stars surrounding a galaxy intersects the event horizons of the black holes at the center. If that is not the case, it's easier to think of the stars as simply in an orbit around the center (there'd be no "accret'n" by the black holes). I'd think of it as simply the difference in terminology: a stable orbit vs a stable point.

2) Why is it irrational for me to think of the big bang as basically the opposite of a black hole - and how do we know it's not continuing to spew matter?

Well, according to most common theories, matter didn't really exist during the big bang (energy only, matter came later). The current model is that the big bang occured *everywhere* in the universe and what most folks think of as the big bang was actually "inflation" (the planck-length sized universe itself got bigger, not any sort of explosion in some specific place in an otherwise "empty" universe). There is also a model of mass-energy conservation, so the "spewage" of matter would necesarily come from some external energy source (outside of the universe?)

Faith or belief is not irrational (you might think of those as axioms of a logical system), however a thoughts within a context (such as a logical system) can often be shown to be rational or irrational thoughts (think of them like theorems which can have proofs), but some might be undecideable. Thus it is not irrational to think about the big bang as the opposite of a black hole, it is merely a viewpoint that is inconsistant with commonly accepted theories. This might make you a crackpot (someone who uses different axioms) or simply misinformed or something in-between.

A truly irrational thinker will not change their thinking when confronted with the flaws in their reasoning**. So, basically, with irrationality, put out your axioms and reasoning for thinking a certain way, let people critique your reasoning, and then being irrational is your choice ;^)

** Sadly, simply proving your axioms to be inconsistent is not a sign of irrational thought. Human beings have a tremendous capacity for cognitive dissonance which isn't considered irrational until it results in anti-social behavior.

howly shyt.

[Anonymous Coward]

past a certain probability, anything is possible.

welcome to 1993

[Anonymous Coward]

it must be running. plan9

Mind

[joh]

Google should invest in such things. Data is worthless if can't do something useful with it. Then Google in the best case could become something like a Mind or at least an embryo of one. In the worst case... Facebook... no! We're doomed!

Asimov's Foundation?

[wayne1932]

Sounds like Psycho-History might be in the making. Now we just have to characterize 7 billion people.

fuck you all.

[Anonymous Coward]

Sure wish someone here had explained what an "event" in this context is.

I guess everyone here knows.

Except me.. and I'm significantly smarter than (very probably) anyone else on Slashdot.

A toast!

[TheGoodNamesWereGone]

A toast!