IBM Chip Processes Data Similar to the Way Your Brain Does

A chip that uses a million digital neurons and 256 million synapses may signal the beginning of a new era of more intelligent computers.

WHY IT MATTERS

Computers that can comprehend messy data such as images could revolutionize what technology can do for us.

New thinking: IBM has built a processor designed using principles at work in your brain.

A new kind of computer chip, unveiled by IBM today, takes design cues from the wrinkled outer layer of the human brain. Though it is no match for a conventional microprocessor at crunching numbers, the chip consumes significantly less power, and is vastly better suited to processing images, sound, and other sensory data.

IBM’s SyNapse chip, as it is called, processes information using a network of just over one million “neurons,” which communicate with one another using electrical spikes—as actual neurons do. The chip uses the same basic components as today’s commercial chips—silicon transistors. But its transistors are configured to mimic the behavior of both neurons and the connections—synapses—between them.

The SyNapse chip breaks with a design known as the Von Neuman architecture that has underpinned computer chips for decades. Although researchers have been experimenting with chips modeled on brains—known as neuromorphic chips—since the late 1980s, until now all have been many times less complex, and not powerful enough to be practical (see “Thinking in Silicon”). Details of the chip were published today in the journal Science.

The new chip is not yet a product, but it is powerful enough to work on real-world problems. In a demonstration at IBM’s Almaden research center, MIT Technology Review saw one recognize cars, people, and bicycles in video of a road intersection. A nearby laptop that had been programed to do the same task processed the footage 100 times slower than real time, and it consumed 100,000 times as much power as the IBM chip. IBM researchers are now experimenting with connecting multiple SyNapse chips together, and they hope to build a supercomputer using thousands.

When data is fed into a SyNapse chip it causes a stream of spikes, and its neurons react with a storm of further spikes. The just over one million neurons on the chip are organized into 4,096 identical blocks of 250, an arrangement inspired by the structure of mammalian brains, which appear to be built out of repeating circuits of 100 to 250 neurons, says Dharmendra Modha, chief scientist for brain-inspired computing at IBM. Programming the chip involves choosing which neurons are connected, and how strongly they influence one another. To recognize cars in video, for example, a programmer would work out the necessary settings on a simulated version of the chip, which would then be transferred over to the real thing.

In recent years, major breakthroughs in image analysis and speech recognition have come from using large, simulated neural networks to work on data (see “Deep Learning”). But those networks require giant clusters of conventional computers. As an example, Google’s famous neural network capable of recognizing cat and human faces required 1,000 computers with 16 processors apiece (see “Self-Taught Software”).

Although the new SyNapse chip has more transistors than most desktop processors, or any chip IBM has ever made, with over five billion, it consumes strikingly little power. When running the traffic video recognition demo, it consumed just 63 milliwatts of power. Server chips with similar numbers of transistors consume tens of watts of power—around 10,000 times more.

The efficiency of conventional computers is limited because they store data and program instructions in a block of memory that’s separate from the processor that carries out instructions. As the processor works through its instructions in a linear sequence, it has to constantly shuttle information back and forth from the memory store—a bottleneck that slows things down and wastes energy.

IBM’s new chip doesn’t have separate memory and processing blocks, because its neurons and synapses intertwine the two functions. And it doesn’t work on data in a linear sequence of operations; individual neurons simply fire when the spikes they receive from other neurons cause them to.

Horst Simon, the deputy director of Lawrence Berkeley National Lab and an expert in supercomputing, says that until now the industry has focused on tinkering with the Von Neuman approach rather than replacing it, for example by using multiple processors in parallel, or using graphics processors to speed up certain types of calculations. The new chip “may be a historic development,” he says. “The very low power consumption and scalability of this architecture are really unique.”

One downside is that IBM’s chip requires an entirely new approach to programming. Although the company announced a suite of tools geared toward writing code for its forthcoming chip last year (see “IBM Scientists Show Blueprints for Brainlike Computing”), even the best programmers find learning to work with the chip bruising, says Modha: “It’s almost always a frustrating experience.” His team is working to create a library of ready-made blocks of code to make the process easier.

Asking the industry to adopt an entirely new kind of chip and way of coding may seem audacious. But IBM may find a receptive audience because it is becoming clear that current computers won’t be able to deliver much more in the way of performance gains. “This chip is coming at the right time,” says Simon.

ORIGINAL: Tech Review

By Tom Simonite

August 7, 2014

IBM's Brain-Inspired Computer Chip Comes from the Future

Illustration: IBM

Brain-inspired computers have tickled the public imagination ever since Arnold Schwarzenegger's character in “Terminator 2: Judgment Day” uttered: “My CPU is a neural net processor; a learning computer.” Today, IBM researchers backed by U.S. military funding unveiled a new computer chip that they say could revolutionize everything from smartphones to smart cars—and perhaps pave the way for neural networks to someday approach the computing capabilities of the human brain.

The IBM neurosynaptic computer chip consists of one million programmable neurons and 256 million programmable synapses conveying signals between the digital neurons. Each of chip’s 4,096 neurosynaptic cores includes the entire computing package—memory, computation, and communication. They all operate in parallel based on “event-driven” computing, similar to the signal spikes and cascades of activity when human brain cells work in concert. Such architecture helps to bypass the bottleneck in traditional computing where program instructions and operation data cannot pass through the same route simultaneously.

“We have not built a brain,” says Dharmendra Modha, chief scientist and founder of IBM’s Cognitive Computing group at IBM Research-Almaden. “But we have come the closest to creating learning function and capturing it in silicon in a scalable way to provide new computing capability that was not possible before.”

Such capability could enable new mobile device applications that emulate the human brain’s capability to swiftly process information about new events or other changes in real-world environments, whether that involves recognizing familiar sounds or a certain face in a moving crowd. IBM envisions its new chips working together with traditional computing devices as hybrid machines—providing an added dose of brain-like intelligence for smart car sensors, cloud computing applications or mobile devices such as smartphones. The chip's architecture was detailed in a new paper published in the 7 August online issue of the journal Science.

Add caption

With a total of 5.4 billion transistors the computer chip, named TrueNorth, is one of the largest CMOS chips ever built. Yet the chip uses just 70 milliwatts while running and has a power density of 20 milliwatts per square centimeter— almost 1/10,000th the power of most modern microprocessors. That brings the new chip's efficiency much closer to the human brain’s astounding power consumption of just 20 watts, or less than the average incandescent light bulb.

“This is literally a supercomputer the size of a postage stamp, light like a feather, and low power like a hearing aid,” Modha says.

One reason IBM was able to minimize power usage is that its chip's computation only triggers when needed. Traditional computer chips have a clock that uses power to trigger and coordinate all the computational processes. But the IBM chip's digital neurons can work together asynchronously when triggered by the signal spikes. IBM also designed its chip to have low power consumption by creating an on-chip network to interconnect all the neurosynaptic cores and building the chip with a low-power process technology used for making mobile devices.

It’s also a supercomputer that can easily scale up in size. IBM designed its computer chip architecture so that it could simply add new neurosynatpic cores within the chip. The chips themselves can be arranged in a repeatable 2-D tile pattern to create bigger machines—IBM has already tested that idea with a 16-chip configuration. That’s the “blueprint of a scalable supercomputer,” Modha says.

Past brain-inspired neural networks have used a combination of both analog and digital to represent the individual neurons. IBM chose to represent the neurons in digital form, which provided several advantages. (At least one other project, SpiNNaker also depends on digital.)

First, the choice allowed IBM engineers to avoid the physical problems of dealing with differences in the manufacturing process or temperature fluctuations. Second, it provided a “one to one equivalence with software and hardware” that allowed the IBM software team to build applications on a simulator even before the physical chip had been designed and tested—applications that ran without problems on the finished chip. Third, the lack of analog circuitry allowed the IBM team to dramatically shrink the size of its circuits. (IBM fabricated its chip using Samsung’s 28-nm process technology—typical for manufacturing chips for mobile devices.)

IBM’s new chip represents the culmination of a decade of Modha’s personal research and almost six years of funding from the U.S. Defense Advanced Research Projects Agency (DARPA). Modha currently heads DARPA’s SyNAPSE project, a global effort that has committed US $53 million to making learning computers since 2008.

Now IBM has built an entire ecosystem around its new chip hardware and software, including a new programming language and a curriculum to teach coders everything they need to know. And the company is reaching out to potential customers, universities, government agencies, and IBM employees to fully explore the commercial applications of its chip technology.

“Our long-term end goal is to build a ‘brain in a box’ with 100 billion synapses consuming 1 kilowatt of power,” Modha says. “In the near future, we’ll be looking at multiple things for empowering smartphones, mobile devices and cloud services with this technology.”

Learn More DARPA SyNAPSE IBM cognitive computing brain-inspired computerslearning computers neural networks neuro-inspired computing

ORIGINAL: IEEE Spectrum

By Jeremy Hsu

Posted 7 Aug 2014 | 18:00 GMT

Can The Human Brain Project Succeed?

Image: Getty Images

An ambitious effort to build human brain simulation capability is meeting with some very human resistance. On Monday, a group of researchers sent an open letter to the European Commission protesting the management of the Human Brain Project, one of two Flagship initiatives selected last year to receive as much as €1 billion over the course of 10 years (the other award went to a far less controversy-courting project devoted to graphene).

The letter, which now has more than 450 signatories, questions the direction of the project and calls for a careful, unbiased review. Although he’s not mentioned by name in the letter, news reports cited resistance to the path chosen by project leader Henry Markram of the Swiss Federal Institute of Technology in Lausanne. One particularly polarizing change was the recent elimination of a subproject, called Cognitive Architectures, as the project made its bid for the next round of funding.

According to Markram, the fuss all comes down to differences in scientific culture. He has described the project, which aims to build six different computing platforms for use by researchers, as an attempt to build a kind of CERN for brain research, a means by which disparate disciplines and vast amounts of data can be brought together. This is a "methodological paradigm shift" for neuroscientists accustomed to individual research grants, Markram told Science, and that's what he says the letter signers are having trouble with.

But some question the main goals of the project, and whether we're actually capable of achieving them at this point. The program's Brain Simulation Platform aims to build the technology needed to reconstruct the mouse brain and eventually the human brain in a supercomputer. Part of the challenge there is technological. Markram has said that an exascale-level machine (one capable of executing 1000 or more petaflops) would be needed to "get a first draft of the human brain", and the energy requirements of such machines are daunting.

Crucially, some experts say that even if we had the computational might to simulate the brain, we're not ready to. "The main apparent goal of building the capacity to construct a larger-scale simulation of the human brain is radically premature," signatory Peter Dayan, who directs a computational neuroscience department at University College London, told the Guardian. He called the project a "waste of money" that "can't but fail from a scientific perspective". To Science, he said "the notion that we know enough about the brain to know what we should simulate is crazy, quite frankly.”

This last comment resonated with me, as it reminded me of a feature that Steve Furber of the University of Manchester wrote for IEEE Spectrum a few years ago. Furber, one of the co-founders of the mobile chip design powerhouse ARM, is now in the process of stringing a million or so of the low-power processors together to build a massively parallel computer capable of simulating 1 billion neurons, about 1% as many as are contained in the human brain.

Furber and his collaborators designed their computing architecture quite carefully in order to take into account the fact that there are still a host of open questions when it comes to basic brain operation. General-purpose computers are power-hungry and slow when it comes to brain simulation. Analog circuitry, which is also on the Human Brain Project's list, might better mimic the way neurons actually operate, but, he wrote,

“as speedy and efficient as analog circuits are, they’re not very flexible; their basic behavior is pretty much baked right into them. And that’s unfortunate, because neuroscientists still don’t know for sure which biological details are crucial to the brain’s ability to process information and which can safely be abstracted away”

The Human Brain Project's website admits that exascale computing will be hard to reach: "even in 2020, we expect that supercomputers will have no more than 200 petabytes." To make up for the shortfall, it says, "what we plan to do is build fast storage random-access storage systems next to the supercomputer, store the complete detailed model there, and then allow our multi-scale simulation software to call in a mix of detailed or simplified models (models of neurons, synapses, circuits, and brain regions) that matches the needs of the research and the available computing power. This is a pragmatic strategy that allows us to keep build ever more detailed models, while keeping our simulations to the level of detail we can support with our current supercomputers."

This does sound like a flexible approach. But, as is par for the course with any ambitious research project, particularly one that involves a great amount of synthesis of disparate fields, it's not yet clear whether it will pay off.

And any big changes in direction may take a while. Although the proposal for the second round of funding will be reviewed this year, according to Science, which reached out to the European Commission, the first review of the project itself won't begin until January 2015.

Rachel Courtland can be found on Twitter at @rcourt.

ORIGINAL: Spectrum
By Rachel Courtland
Posted 9 Jul 2014 | 17:00 GMT

Gordon Bell Prize Bubbles from Sequoia

Each year at SC, the ACM hands out one of the most coveted awards, the Gordon Bell Prize. The award, which became a regular feature of SC, began in 1987 and now carries a $10,000 prize sponsored by parallel computing luminary, Gordon Bell. Winners demonstrate high peak performance figures on real world applications or demonstrate other performance-geared achievements, including incredible advances in scaling, time to solution of scientific applications or other feats of HPC might.

This year the Gordon Bell award, chosen because of its demonstration of a high performance application went to “11 PFLOP/s Simulations of Cloud Cavitation Collapse,” by Diego Rossinelli, Babak Hejazialhosseini, Panagiotis Hadjidoukas and Petros Koumoutsakos, all of ETH Zurich, Costas Bekas and Alessandro Curioni of IBM Zurich Research Laboratory, and Steffen Schmidt and Nikolaus Adams of Technical University Munich.

The researchers, in collaboration with the Technical University of Munich and LLNL broke some serious computational fluid dynamics ground in their simulation, which maneuvered 6.4 million threads on the IBM Sequoia system. The simulation, according to IBM, stands as the “largest simulation ever in fluid dynamics by employing 13 trillion cells and reaching an unprecedented, for flow simulations, 14.4 petaflop sustained performance on Sequoia—73% of the supercomputer’s theoretical peak.”

The bubble bursting exercises are more than just interesting to watch in action. These simulations model complex events related to clouds of collapsing bubbles, which can yield new insight in manufacturing, medicine and beyond as scientists seek to understand how they might “shatter” tumors, kidney stones or even fuel injection fluid interactions.

The researchers described their award-winning effort by pointing to how the “destructive power of cavitation reduces the lifetime of energy critical systems such as internal combustion engines and hydraulic turbines, yet it has been harnessed for water purification and kidney lithotripsy.” They go on to note that they were able to “advance by one order of magnitude the current state-of-the-art in terms of time to solution, and by two orders the geometrical complexity of the flow. The software successfully addresses the challenges that hinder the effective solution of complex flows on contemporary supercomputers, such as limited memory bandwidth, I/O bandwidth and storage capacity.”

“We were able to accomplish this using an array of pioneering hardware and software features within the IBM BlueGene/Q platform that allowed the fast development of ultra-scalable code which achieves an order of magnitude better performance than previous state-of-the-art,” said Alessandro Curioni, head of mathematical and computational sciences department at IBM Research – Zurich. “While the Top500 list will continue to generate global interest, the applications of these machines and how they are used to tackle some of the world’s most pressing human and business issues more accurately quantifies the evolution of supercomputing.”

As IBM noted, These simulations are one to two orders of magnitude faster than any previously reported flow simulation. The last major achievement was earlier this year by a team at Stanford University which broke the one million core barrier, also on Sequoia.

This year the prize committee clarified their description of what it takes to render a winner, including the following criteria:

The prize winner is not selected simply on raw performance numbers. Rather, the Prize Committee seeks:

• evidence of important algorithmic and/or implementation innovations
• clear improvement over the previous state-of-the-art
• solutions that don’t depend on one-of-a-kind architectures (systems that can only be used to address a narrow range of problems, or that can’t be replicated by others)
• performance measurements that have been characterized in terms of scalability (strong as well as weak scaling), time to solution, efficiency (in using bottleneck resources, such as memory size or bandwidth, communications bandwidth, I/O), and/or peak performance
• achievements that are generalizable, in the sense that other people can learn and benefit from the innovations

Solving an important scientific or engineering problem is important to demonstrate/justify the work, but scientific outcomes alone are not sufficient for this prize.

ORIGINAL: HPC Wire
Nicole Hemsoth
November 22, 2013

Scientists to simulate human brain inside a supercomputer

ORIGINAL: CNN

By Barry Neild, CNN 

October 12, 2012 -- Updated 1313 GMT (2113 HKT) | 

Researchers say building a computer simulation could improve understanding and treatments of brain diseases like Alzheimer's. 

STORY HIGHLIGHTS 

• Human Brain Project will use supercomputers to mimic tangle of neurons and synapses that power our thoughts 
• Scientists say the simulator could offer new insight into the treatment of brain disease like Parkinson's and Alzheimer's 
• "Brain in a box" is unlikely to transform into sci-fi-style computer bent on world domination, scientists say 

(CNN) -- There's no escaping the fact that the Human Brain Project, with its billion-dollar plan to recreate the human mind inside a supercomputer, sounds like a science fiction nightmare. 

But those involved hope their ambitious goal of simulating the tangle of neurons and synapses that power our thought processes could offer solutions to tackling conditions such as depression, Parkinson's disease and Alzheimer's. 

The Human Brain venture is the next step in a long-running program that has already succeeded in using computers to create a virtual replica of part of a rat's neocortex -- a section of the brain believed to control higher functions such as conscious thought, movement and reasoning. 

Scientists at its forerunner, the Switzerland-based Blue Brain Project, have been working since 2005 to feed a computer with vast quantities of data and algorithms produced from studying tiny slivers of rodent gray matter. 

"This is a tool for research, not a giant simulated brain that is going to rule the world

Sean Hill, neuroscientist 

Last month they announced a significant advancement when they were able to use their simulator to accurately predict the location of synapses in the neocortex, effectively mapping out the complex electrical brain circuitry through which thoughts travel. 

Henry Markram, the South African-born neuroscientist who heads the project, said the breakthrough would have taken "decades, if not centuries" to chart using a real neocortex. He said it was proof their concept, dubbed "brain in a box" by Nature magazine, would work. 

Read: Mapping out a new era in brain research

Now the team are joining forces with other scientists to create the Human Brain Project. As its name suggests, they aim to scale up their model to recreate an entire human brain. 

It is a step that will need both a huge increase in funding and access to computers so advanced that they have yet to be built. 

If their current bid for €1 billion ($1.3 billion) of European Commission funding over the next 10 years is successful, Markram predicts that his computer neuroscientists are a decade away from producing a synthetic mind that could, in theory, talk and interact in the same way humans do. 

His bold claims have inevitably fueled comparisons to doom-laden popular fiction in which conscious machines turn on their creators and wreak havoc. 

The project's scientists have been referred to as "team Frankenstein" and their computer likened to "Skynet," the virtual intelligence that unleashes a robot war on humanity in the "Terminator" films. 

Sean Hill, a senior computational neuroscientist on the project, laughs at such comparisons. 

He says the computer will primarily become a repository for knowledge about the brain that will allow scientists to conduct experiments without the need to probe inside people's skulls. 

Read: $1 billion mission to reach the Earth's mantle

"This is a tool for research, not a giant simulated brain that is going to rule the world," he said. 

"Right now, we're in a crisis in neuroscience. There's a lot of wonderful data being gathered but we don't have a place where we can put those experimental results together and understand their implications. 

"We are just beginning to appreciate how complex our brains are, far beyond any other device in the known universe

Terry Sejnowski, Salk Institute for Biological Studies 

"The benefit of having this facility is you have a place to integrate the data into a model where you can test predictions and start to learn principles of how the brain operates." 

The computing power needed to build the model is phenomenal. Simply to replicate one of the 10,000 neuron brain cells involved in the rat experiment took the processing capacity usually found in a single laptop. To simulate a fully functioning human brain, it would take billions. 

Hill says that such computational power -- known as exascale -- will be available by the end of the decade. The Human Brain Project's scientists are hoping to work with supercomputer developers to ensure future machines match their requirements. 

But, even as the team touts its experiments as a possible solution to the brain diseases that affect about two billion people worldwide, they have attracted critics who say their work is far too broad in scope to achieve usable results. 

Professor Terry Sejnowski, head of the Computational Neurobiology Laboratory at the Salk Institute for Biological Studies in San Diego, has been quoted as saying the Blue Brain project is "bound to fail." 

Read: How the search for aliens can help sustain life on Earth

He told CNN via email that "progress is being made but there is still a long way to go before we will understand the computational capabilities of cortical circuits." 

He added: "We are just beginning to appreciate how complex our brains are, far beyond any other device in the known universe." 

Sean Hill said the team hoped it was answering skeptics with its achievements so far. 

"It's just a matter of keeping on doing it. Let's keep improving these tools and open them up so that many scientists are engaged and collaborating and using it as common point to bring the data together," he said. 

"The only way to address the critics is to keep working, showing the positive results and do the best we can -- and that is starting to happen."

Computing with Light

ORIGINAL: MIT Technology Review

David Zax

December 10, 2012

A breakthrough from IBM could signal a future for computing.

This image shows a falsely-colored integrated optical and electrical circuit. The blue wires carry optical signals and the yellow wires carry electrical ones.

IBM announced what it called a technological breakthrough today in San Francisco. The company verified in a manufacturing environment the feasibility of using light instead of electrical signals to transmit information. IBM had proven the concept of such technology, called “silicon nanophotonics,” back in 2010, but this announcement, following a decade of research, nudges the field towards commercial applications.

In its 2010 announcement, IBM described the invention succinctly: a chip that “integrates electrical and optical devices on the same piece of silicon, enabling computer chips to communicate using pulses of light (instead of electrical signals), resulting in smaller, faster and more power-efficient chips than is possible with conventional technologies.” The development forms a part of IBM’s “Exascale computing program,” which wants to build a supercomputer than can perform a million trillion calculations (a so-called “Exaflop”) in a second. IBM says the chip was made using a 90 nanometer manufacturing process, and that the optical data can travel through the chip at 25 Gigabits per second.

Basically, IBM hopes this will solve the problem that we are creating and transmitting data faster than our hardware has been able to keep up with. Silicon nanophotonics, says IBM, will help industry “keep pace with increasing demands in chip performance and computing power.” As one of the researchers, Dr. Solomon Assefa, explained: “For our computer servers to keep up with this growth, so that we can actually make sense of the data through analytics and so forth, we need to have a new technology.”

Why compute with light, rather than electrons? As TR explained in 2011 (see “Light Chips”): “The speed of supercomputers is constrained not by processing power but by limits on how fast data can travel down the electrical wires that link up different chips. Light signals move significantly faster than electrical ones, so using them could remove that bottleneck.” Reports today also pointed out that transmitting via light allowed further distances of data transfer while minimizing risk of lost data.

Again, today’s announcement has more to do with the process of fabricating the technology, which had already been demonstrated. IDG News Service sums it up well: IBM has shown that it’s possible to “bake optical circuitry into silicon processors using existing fabrication techniques, which could set the stage for radically faster and lower-cost computer communications.” The BBC explains that many data centers already use optical cables to shuttle around data, but they’ve had to have expensive equipment to convert photon-data into electron-data.

IBM presented the breakthrough at the 2012 IEEE International Electron Devices Meeting.

U.S. Lab's "Titan" Named World's Fastest Supercomputer

ORIGINAL: NatGeo

Marianne Lavelle For National Geographic News

October 29, 2012

The power of Titan, a supercomputer at the Oak Ridge National Laboratory in Tennessee, is akin to each of the world’s 7 billion people being able to carry out 3 million calculations per second. Photograph courtesy Charles Brooks, Oak Ridge National Laboratory

Update November 12: Titan named world's fastest supercomputer.

In a breakthrough that harnesses video-game technology for solving science's most complex mysteries, the U.S. government's new Titan machine was named the world's fastest supercomputer. Deployed just two weeks ago, Titan is the fastest, most powerful, and most energy-efficient of a new generation of supercomputers that breach the bounds of "central processing unit" computing.

The announcement in Salt Lake City marks a return to the top of the the closely watched, semiannual TOP500 list for the U.S. Department of Energy's (DOE) Oak Ridge National Laboratory in Tennessee. Its previous supercomputer, Jaguar, led the world for a year before being overtaken in 2010 by a Chinese system, which later was supplanted by a machine in Japan. But another U.S. government machine, Sequoia, has topped the world since June. Titan's performance on the TOP500's benchmark test was 17.59 petaflops, about 17,590 trillion calculations each second, edging out Sequoia, with a speed of 16.33 petaflops. Watch a video about Titan here:

It would take 60,000 years for 1,000 people working at a rate of one calculation per second to complete the number of calculations that Titan can process in a single second. Think of Titan's power as akin to each of the world's 7 billion people solving 3 million math problems per second.

But Titan's signature achievement is how little energy it burns while blazing through those computations.

Titan's predecessor supercomputer at Oak Ridge, the 2.3-petaflop Jaguar machine, drew 7 megawatts (MW) of electricity, enough to power a small town. Titan needs just about 30 percent more electricity, 9 MW, while delivering ninefold greater computing power.

"We're able to achieve an order of magnitude increase in our scientific computing capabilities, which is what we need for our challenges, but to do so at essentially the same energy budget," says Jack Wells, director of science at the Oak Ridge Leadership Computing Facility. "Titan puts us on a different curve with respect to the energy consumption for increased computing power."

Video-Gaming Efficiency

Titan's energy-saving secret is a "hybrid" architecture that boosts the power of central processing units (CPUs) by marrying them to high-performance, energy-efficient graphical processing units (GPUs)—the technology that propels and animates today's most popular video games. A few dozen supercomputers around the world have used GPU and CPU processing in tandem since the first hybrid machine, the one-petaflop Roadrunner, at Los Alamos National Laboratory in New Mexico in 2008. Titan is the largest, by far.

To update pixels rapidly enough to bring angry birds, soldiers, and athletes to life on game consoles and handheld devices, GPUs have to handle large amounts of data at the same time, in parallel fashion. "This is exactly what we need for the future in order to enable progress and manage the energy [in supercomputing]," Wells says. If Titan had relied only on CPUs, which are optimized to do just one task at a time rapidly and flexibly (serial processing), Oak Ridge estimates the electricity requirements would have been about 30 MW, or more than three times greater than the system now demands.

Titan's approach is not the only path to energy-efficient supercomputing. IBM's "Sequoia" BlueGene/Q supercomputer at the U.S. Department of Energy's Lawrence Livermore Laboratory in California, now No. 2 on the official Top500 list, is part of a family of supercomputers that have been leaders in low-power design. The Sequoia can boast energy efficiency similar to Titan's (it uses 8 MW, and its peak performance is 20 petaflops computing power) through a design using many small, low-power embedded chips, connected through specialized networks inside the system. Four of the current top ten fastest supercomputers are BlueGene/Q machines, but the design does not use widely available commodity processors.

But Oak Ridge and its machine designer, Seattle-based Cray, have built Titan with processors made by the same companies that make the processors in consumer personal computing and gaming products. The upgrade from the Jaguar system to the Titan Cray XK7 system, which cost about $100 million, relies on AMD Opteron CPUs (299,008 CPU cores in all) and NVIDIA Tesla GPUs. It's an approach that has allowed Oak Ridge to take advantage of advances in the broader information technology market—including the highly efficient processing needed for video games—to drive energy efficiency.

"There's an economic model here that really enables this to work," says Steve Scott, chief technology officer for NVIDIA, based in Santa Clara, California. "The high-performance computing industry has great demand but it's not a very large market. But we're able to leverage this very broad consumer technology and use that to enable power-efficiency breakthroughs and make this high-performance computational tool possible. (See "Supercomputing Power Could Pave the Way to Energy-Efficient Engines")

"So when you go out and download and play the latest video game," Scott says, "you actually are helping to advance science."

From Motors to Skin

Because Titan marks an achievement in energy efficiency, it is perhaps appropriate that one of its primary uses will be to advance science on the future of energy. Titan will be put to work on research into systems for more fuel-efficient automobiles, for safer nuclear power reactors with improved power output, and on advanced magnets that could drive future electric motors and generators. It also will be used in research to model more accurately the impact of climate change.

These projects were among 61 science and engineering projects awarded time on Titan and another U.S. supercomputer at Argonne National Laboratory outside of Chicago, the DOE announced today. Scientists in fields from molecular biology to materials science vie for time on the machine at Oak Ridge and other U.S. government facilities, in a competitive process in which projects are picked for "high potential for accelerating discovery and innovation." The deployment of Titan makes it the largest open science supercomputer in operation in the world today. (In contrast, Sequoia is dedicated to classified work on maintenance of the U.S. government's nuclear weapons stockpile.)

Although researchers use supercomputers to model staggeringly complex interaction of natural and man-made systems, some of the applications of these systems are commonplace, and even mundane. The giant consumer products company Procter & Gamble has its own supercomputer (often ranked in the Top 500, though not in the Top 10) to tackle such problems as how to make strong paper towels that tear easily at the perforation, how to make billions of diapers at blinding speed, and how to engineer containers that open easily but don't leak.

"Last year, we did over 50,000 calculations on plastic bottles," says Tom Lange, director of modeling and simulation corporate research and development for Procter & Gamble.

Now, P&G researchers, working in partnership with scientists from Temple University in Philadelphia, have been awarded time on Titan because they are tackling a project deemed of broad interest and stunning complexity. Their work will aim to develop the first molecular-based model for understanding how lotions or drugs are delivered through the skin.

Michael Klein, director of Temple's Institute for Computational Molecular Science, explains that only in recent years has it been understood that beneath the first layer, the human skin is made of a complex matrix of lipids, cholesterol, so-called free fatty acids and another type of long-chain fatty acid known as ceramides. "The structure of this matrix gives skin all these beautiful properties," of flexibility, resilience, water-resistance, and the like, he says. "Understanding how these processes work is of great interest to consumer products companies," he says.

Because it is conducted on the big government machine, the research will be published and shared with the scientific community at large, where the potential to advance medical science has a broad public benefit.

It's just one example of the surprisingly wide reach of supercomputing work. "The scope is as broad as science and engineering is broad," says Wells. "It is not so much about having the leading supercomputer in Top 500 list. That's significant, but it's not really what we're focused on. We're focused on the science and engineering applications. It's about clean energy, it's about clean air. It's about a sustainable future. We offer our resources to companies big and small to come work with us to take a look into the future."

This story is part of a special series that explores energy issues. For more, visit The Great Energy Challenge.

Supercomputer Recreates Universe From Big Bang to Today

ORIGINAL: Space

Clara Moskowitz, SPACE.com Assistant Managing Editor

11 September 2012

Large-scale structures in the universe form over time in these stills from a supercomputer simulation of the evolution of the universe.

CREDIT: Habib et al./Argonne National Lab 

Scientists would love to be able to rewind the universe and watch what happened from the start. Since that's not possible, researchers must create their own mini-universes inside computers and unleash the laws of physics on them, to study their evolution.

Now researchers are planning the most detailed, largest-scale simulation of this kind to date. One of the main mysteries they hope to solve with it is the origin of the dark energy that's causing the universe to accelerate in its expansion.

The new simulation is a project led by physicists Salman Habib and Katrin Heitmann of Illinois' Argonne National Laboratory, and will run on the lab's Mira supercomputer, the third-fastest computer in the world, starting in the next month or two. The program will use trillions of "particles" — elements in the simulation that stand in for small bits of matter. The computer will let time run, and watch as the particles move through space in response to the forces acting on them.

As the simulation progresses, these bits of matter will clump together under gravity to form larger and larger blobs representing galaxies, galaxy clusters and superclusters. To evolve the universe from the Big Bang 13.7 billion years forward to today, the simulation will take up to two weeks. [Video: Simulation of the Universe from Big Bang to Now]

Testing the theory

The ultimate goal is to compare the best telescope observations of structure in the universe to the structure displayed in the computer model, to test the reigning theory of cosmology.

"We are trying to look for subtle ways in which it's wrong," Habib told SPACE.com. "That’s why you need these very high-resolution, very large-scale simulations to see if the observations don't match the predictions."

Dark energy is the name given to whatever is causing the expansion of the universe to accelerate. When this acceleration was first discovered in the 1990s, it shocked the science community, because theories predicted the universe's expansion would be steady or slowing down, because of the inward pull of gravity.

The current reigning theory posits that dark energy is what's called the cosmological constant, a term Einstein first thought to put into his equations of general relativity to represent the vacuum energy of the universe. Although Einstein ultimately decided not to include the term, scientists later realized that it could explain the current observations of the expansion of the universe.

However, cosmologists aren't satisfied with this explanation, Habib said.

Another possibility

"It's just a single number entered as an extra term in the equations," he said. "The problem is that if you ask what its value should be, it's enormous — many orders of magnitude bigger than what is actually observed."

While simulations based on the cosmological constant so far appear to match what's seen in large-scale observations of the universe, scientists think that next-generation observations may reveal discrepant details.

If a cosmological constant is not to blame for the accelerated expansion of the universe, another possibility is that space contains some other type of mass or energy, such as a field, that is pulling everything apart.

"It's basically guesswork; it could be like this, or it could be like that," Habib said. "Either way it's very interesting."

Follow Clara Moskowitz on Twitter @ClaraMoskowitz or SPACE.com @Spacedotcom. We're also onFacebook & Google+.

Supercomputer will help researchers map climate change down to the local level

ORIGINAL: Washington Post

By Stephen P. Nash,

An advance guard of 18-wheelers is scheduled to roll into a business park in Cheyenne, Wyo., this week to unload components of a supercomputer called Yellowstone. This 1.5-quadrillion-calculations-per-second crystal ball will model future climate and forecast extreme weather.

“It’s a big deal,” said climate scientist Linda Mearns of the National Center for Atmospheric Research in Boulder, Colo. Yellowstone will help researchers calculate climate change on a regional, rather than continental, scale. With a better grasp of how warming may affect local water resources, endangered species and extreme winds, local and state governments will be able to plan more effectively.

Marika Holland, chief scientist for NCAR’s Community Earth System Modeling Project, said the new supercomputer is “close to a game-changer. We’ve had incremental improvements in our computational resources over time, but Yellowstone is a whole new scale, and there are things that we will be able to explore that just were not possible before.”

As climate models become more complex and detailed, limited computing power bottlenecks the research. Broad-brush models use less power, but they often cannot consider details that drive local climate, such as complex coastlines or the mountain ranges and valleys that affect rainfall. Researchers refer to this as a problem of “model resolution.”

“If you have an old digital camera that doesn’t have as many megapixels as a new one, you want to get that [new] camera because it takes sharper pictures, and you can store a lot more pictures on it,” said Richard Loft, a director in the computing lab at NCAR.

“It’s the same thing with Yellowstone. We’ve increased our ability to generate more-detailed pictures of the climate system, so we get a sharper, crisper view of things.”

Seven-square-mile ‘pixels’

Yellowstone will be able to generate climate projections for seven-square-mile “pixels,” instead of the 60-square-mile units typically in use now. “We’ve already had the model resolution to confirm the warming of the planet but not to talk about the winners and losers at the regional scale,” Loft said.

The new computer will cost about $30 million; it will be operated by NCAR, which is funded by the National Science Foundation. And just like your own computer, it is likely to be obsolete and ready for replacement by a faster model in about four years.

Yellowstone and its successors will allow modelers to provide a more reliable range of possible answers for local climate questions:

• How dry will it be by mid-century, say, in the watersheds that feed the Prettyboy and Loch Raven reservoirs, which supply much of Baltimore’s water? 
• How high will summer heat spike in the Shenandoah Valley? 
• Where can Florida panthers and loblolly pine, a major source of timber, survive as the climate shifts?

“Many climate scientists out there are itching to be able to do higher-resolution simulations and more experiments,” Holland said. “One of our working groups has strong links with water resource managers and water utilities. For them, models of sea-level rise, vegetation, snowpack, precipitation and stream-flow changes are all important. Those planners are living with the uncertainty of how climate changes are going to occur.”

Mathew Maltrud of the Los Alamos National Laboratory in New Mexico runs climate models that try to mimic the physics of rivers, vegetation and atmospheric moisture, sea ice and ocean currents.

“We’re moving into a realm where we have models that resemble the ocean, the atmosphere, the ice and the land to a high degree,” he said. Yellowstone will allow those components to interact more realistically.

“For example, what’s going to happen in the southwestern U.S.?” Maltrud asked. “Wetter? Drier? The models out there right now show a lot of agreement on global change, but when it comes to regional scales, there’s lots of variability. We hope to get far better representation of precipitation, which is definitely key for decision-makers. That could have a very large impact on understanding regional changes all over the world.”

The dinosaurs’ fate

Several modeling efforts intended to provide data for local and regional planners that are already underway may benefit from additional computing power.

Adam Terando, climate change research coordinator at North Carolina State University, is incorporating climatic factors into regional modeling as part of a federal effort to “make climate-change data more useful and relatable for decision-makers,” he said. His research also tries to prophesy a range of future climatic changes that could affect agriculture in the eastern United States: frost days, heat stress and the length of growing seasons.

Those decision-makers “need to know: How is climate change going to affect my wildlife refuge on the coast of North Carolina, or Virginia?” Terando said. The Southeast, for example, has high numbers of amphibian species. The amphibians are reliant on the little puddles that form in the winter and spring called vernal pools, where they hatch their young.

“If the water table changes, how will it affect them?” he asked. “We’re talking to these people already. They know they need to be doing planning now, not just for the next 10 years, but for the next 30 or 50 years, and beyond.”

Yellowstone will also allow researchers to produce climate snapshots with intervals of hours rather than days; such detail is difficult now because of limited data storage capacity. “That fine step is really what the folks that I work with — the ecologists and biologists and all the resource managers — really need,” Terando said, “because that’s much more useful for figuring out how climate change could affect different plants and animals.”

Loft has been most concerned about projections that raise the possibility of severe drought through the central and southwestern states and in Turkey and Spain — far worse conditions than the drought that has caused water rationing and the liquidation of cattle herds in Texas.

“Would you rather go to a doctor who tells you what you want to hear, or what’s really going to happen?” Loft asked. “Not everybody has the same answer to that, I guess. I’d like to know. And this machine’s going to help us know more about how bad climate is, as a threat to human civilization, and I think that’s a good thing.

“Dinosaurs didn’t have any technology to know something was coming, so they just woke up one day and found out that the planet was destroyed. But if we can know about it 30 years before it happens, and do some things to cope with it, we can still mitigate the impact on human beings,” he said.

Nash reports on climate science and teaches journalism at the University of Richmond.

Climate change around the world

A look at the biggest climate change stories of our generation, from the Gulf oil spill, Cancun climate talks, and flooding in Pakistan.

Smog and haze hover over Salt Lake City. The thick layer of smog lingering over Utah has fouled the state's mountain air so badly that health officials have warned people not to exercise outside and schools are keeping children inside for recess and sports. The smog is blamed on a weather phenomenon that pins pollution to the valley floors. Brian Nicholson / AP