SoylentNews

The World's Most Powerful Supercomputer is Tailor-Made for the AI Era

posted by martyb on Saturday June 09 2018, @12:29AM   
from the but-can-it-play-Crysis? dept.

AnonTechie writes:

Since 2013, Chinese machines have occupied the number one slot in rankings of the world's most powerful supercomputers. Now America is back on top again. Engineers at the US Department of Energy's Oak Ridge National Lab in Tennessee have just unveiled Summit, a supercomputer with enough processing power to surpass the current record holder, China's Sunway TaihuLight.

The new machine is capable, at peak performance, of 200 petaflops—200 million billion calculations a second. To put that in context, everyone on earth would have to do a calculation every second of every day for 305 days to crunch what the new machine can do in the blink of an eye. Summit is 60 percent faster than the TaihuLight and almost eight times as fast as a machine called Titan, which is also housed at Oak Ridge and held the US supercomputing speed record until Summit's arrival.

[...] Jack Wells of Oak Ridge says the experience of building Summit, which fills an area the size of two tennis courts and carries 4,000 gallons of water a minute through its cooling system to carry away about 13 megawatts of heat, will help inform work on exascale machines, which will require even more impressive infrastructure. Things like Summit's advanced memory management and the novel, high-bandwidth linkages that connect its chips will be essential for handling the vast amounts of data exascale machines will generate. Scientists at the national lab say they've already leveraged Summit's AI smarts to conduct what is effectively an exascale comparative genomics calculation. Thanks to these and other advances, Summit will help us reach even more impressive peaks of computing power.

The world's most powerful supercomputer

Also Covered by

Move Over, China: U.S. Is Again Home to World's Speediest Supercomputer (The New York Times)

Oak Ridge National Laboratory Launches America's New Top Supercomputer for Science

The World's Most Powerful Supercomputer Is an Absolute Beast (GIZMODO)

---

Original Submission

• Re:Why do they need it?(Score: 2, Interesting) by sea on Saturday June 09 2018, @04:45PM

  by sea (86) on Saturday June 09 2018, @04:45PM (#690830) Homepage Journal

  You ask: Why do they need the ability to do more calculations faster?

  Here are some fun puzzles I can think of. They're outrageously difficult, so after a short time working on them you'll see that at the moment and until we can find better algorithms the best thing we can do is throw more computing power at these problems in the hope of approximating better solutions.

  • Puzzle 1: Can you find a collection of molecules which store an enormous amount of energy, is easy to 'burn' to release that energy, and has minimized negative side effects? (Explosive, Toxic, produces insane amounts of greenhouse gases on burning, etc.). This is the problem of finding better fuels for our society, and since there are such a large number of possible chemical combinations and methods of releasing their energy, you can see why you'd need massive computational power to search the space.
  • Puzzle 2: Given tonnes of genetic data (and I really do mean tonnes. Think *many* terabytes worth), can you find and classify sequences that appear to cause particular illnesses and disorders while minimizing overlap with important sequences? Can you do it precisely enough that we can build CRISPR-powered tools to target these subsequences and cure the disorders, without wreaking untold havoc on the body? Bear in mind that if you make a mistake, or you didn't do your homework to the fullest extent, people might die.
  • Puzzle 3: My friends want to build a fusion reactor, but gosh they're having a hard time finding ways to minimize the input energy needed to stabilize the plasma without destroying the containment vessel. Can you think of a better design? Something that allows you to use less-powerful magnetic fields, but perhaps organized in a novel pattern that coincidentally keeps the plasma from obliterating your fusion chamber. Maybe the design for the chamber itself needs optimization. Who knows? You have free reign to try every combination you can think of. There are a lot of combinations of shapes, materials, and energy inputs. Maybe you even want to pulse your input energy or alter the field strength in some areas by a time-varying function. The search space is enormous, so I think you'll need a pretty big computer to take a crack at any part of it.
  • Puzzle 4: Find me a better alloy and a construction method for getting a hold of it. I want a lighter version of steel, but ten times stronger, and it would be nice if you could give me a list of all of its thermal and mechanical properties along the way. It'll come in handy for all of those bridges and skyscrapers we want to build. Bonus points if you can find a way to make it out of cheaper and more plentiful metals or components. (Gee, what if you could make it out of carbon? Science fiction books seem to say it's possible..)
  • Puzzle 5: I'll give you some historic weather and climate data. There's a lot of it. I want you to piece it together and tell me what's correlated to what, and if possible, predict how things will change if I vary the parameters a bit, say, by increasing CO2 concentrations in a particular way. The better your predictions are, the smarter I can be about planning for the future of climate change, and the more money I can save in the long run. By I, of course, I mean the species.
  • Puzzle 6: I want to build a quantum computer, but the parts are so small that I can't realistically probe them all and design it the old fashioned way. We know the behavior of particles quite well, so could you find me a design that holds them in the right state for long enough to get my computational results out? Even the slightest design tweak seems to have massive cascading effects..
  • Puzzle 7: There are a *tonne* of mathematical puzzles out there. A lot of them could benefit from a great big whack on the search space to find counterexamples. We don't know what any of these problems are useful for today, but the thing is, all the evidence suggests that mathematical insights always come back with a payoff in 70 to 100 years...
  • Puzzle 8: This is the hardest puzzle of all. Can you think of a puzzle that we haven't solved yet, but which, if we did, would improve the quality of life on earth (or someday, elsewhere?) As long as there's the chance that someone will come up with another need for a supercomputer, I say we have them primed and available. You never know when someone will be hit by insight and come up with the golden algorithm that can compute what we need to find that cure for cancer..We'd better have the tools they need ready for that day.

  You get the idea.

  -
  Daniel McChlery

  Parent

  Starting Score:	1		point
  Moderation		+1
  Interesting=1, Total=1
  Extra 'Interesting' Modifier		0
  Total Score:		2