It’s a schoolboy’s dream: a laser-powered computer. But now thanks to the development of a novel super-stretched germanium-based semiconductor it could form the basis of future ultra-high speed optical computers.
It’s long been known that microelectronics faces seemingly impossible challenges from the immutable Laws of Physics the smaller components get. That’s led to some researchers believing that the future of computing lies in optical systems and lasers.
The problem is that traditional silicon semiconductors are not much use when it comes to the emission of laser light, says Richard Geiger, a doctoral student at the Laboratory for Micro- and Nanotechnology at the Paul Scherrer Institute, in Switzerland.
Gieger and colleagues at technology college ETH Zurich instead turned their attention to germanium components.
“Germanium is perfectly compatible with silicon and already used in the computer industry for the production of silicon chips,” said Geiger.
They were able to develop a novel manufacturing process which introduces tensile strain on germanium which its etched on to slabs of silicon. Under strain, germanium’s physical properties change, making suitable for lasing.
“With a strain of three percent, the material emits around twenty-five times more photons than in a relaxed state,” explained Martin Süess. The technique used to create this strain at the atomic level is equivalent to the comparable forces exerted on a pencil as two lorries pull on it in opposite directions, the researchers claimed.
They now believe it will be possible to build tiny lasers using this technique, paving the way for future nanocomputers. The work was published in this week’s Nature Photonics
Forget external hard drives and more RAM; if you really want to upgrade your computing rig, it’s time to get your electrons spinning in the right direction. New research in the field of “spintronics” suggests that changing the spin of the electrons in an internal processor is the best way to get your computer running quickly.
According to Nitin Samarth, a professor of physics at Pennsylvania State University, advancements in computing speeds slowed down about five years ago because of limitations in the way computer processors are made.
Today’s processors rely on the density of transistors on a computer chip. The transistors act like switches, turning “on” or “off” – corresponding to a “one” or “zero” state – to change the flow of electrons through the chip. The more tightly packed the transistors are on a chip, the faster the flow of electrons through the processor. So tightly packed transistors make computers run faster. But all those moving electrons create a lot of heat.
“The very fundamental limitation that stops computer manufacturers from making these chips faster is that the transistors are reaching the density at which the heat they generate cannot dissipate fast enough to prevent the computer from melting,” Samarth said.
Samarth and his team have been working on giving computers a new way to get their ones and zeros. Instead of having electrons flow hotly through a transistor, the scientists have designed new kinds of materials that will let individual electrons be “cool” by just spinning in one of two opposite ways.
“In one of those mysterious aspects of quantum mechanics, Nature allows the electron only to either spin up or spin down in the presence of a magnetic field,” Samarth said.
Samarth and his team of researchers have been trying to cool down the process of creating super-fast transistors, and they’re doing it one electron at a time.
“One of the interesting phenomena that people discovered in recent years is that you can change the orientation of an electron’s spin just by using a voltage,” Samarth said. “You are not back to creating heat-generating resistance inside a transistor by making the electrons flow, but rather you are just changing the orientation of their spin.”
These electron-manipulating transistors could pave the way to faster, more energy-efficient computers, according to Samarth. But first, he and his team need to build them.
“It’s a little bit like playing atomic-scale Legos,” said Samarth of the process of creating “spintronic” transistors. “Like the components in modern computer chips, the highly specialized devices that we are fabricating here, which might function as spintronics transistors, typically are smaller than 100 times the width of one strand of your hair.”
Samarth uses two ultra-high-vacuum chambers to perform a process known as molecular-beam epitaxy.
“At first we create a very high-vacuum inside these chambers, then we deposit beams of selected kinds of elements in a very controlled way in order to deposit a layer of material that is the thickness of only one atom,” Samarth said.
“Because of the ultra-high vacuum, while one atomic layer is going down, nothing else is being deposited…We then can use a variety of different elements, plus the power of thermodynamics and chemistry, to engineer the crystal we want by building it up one atomic layer at a time.”
In addition to the creation of these “spintronic” transistors, the researchers are also working on manipulating electron spin in other semiconducting devices.
Yesterday was Pi Day, and to celebrate the yearly occasion, you no doubt tried your hardest to recite Pi
to as many decimal places as you could. Of course, most of us probably couldn’t
get past the first few decimal places, but there was one person who could,
thanks to a set of computers powered by a handful of NVIDIA graphics cards.
Santa Clara University researcher Ed Karrels ended up breaking the world record for computing digits
of Pi to eight quadrillion places to the right of the decimal point. Karrels
used graphics cards to do the work rather than CPUs, and he spread the work
across three different computers: one with four NVIDIA GTX 690 cards, one with
two NVIDIA GTX 680 cards, and 24 computers at the Santa Clara University Design
Center with one NVIDIA GTX 570 card each.
The calculation took 35 days to complete, from December 19 to January 22, beating out the previous held
by a team at Yahoo, who used 1,000 CPU-only computers, which took 23 days to
compute Pi to two-quadrillion places, just a quarter of what Karrels’s setup
achieved. After the 35-day run, Karrels conducted a second run to double-check
the math, which took just 26 days using newer versions of his programming
tools.
Karrels will speak at the GPU Technology Conference in San Jose, California next Tuesday, where he’ll be
explaining the math behind the Pi calculation achievement, as well as the
programming tricks he used, as well as the logistics of conducting
supercomputing tasks on a budget.
When it comes to data crunching, quantum computers will leave today’s fastest processors in the dust.
For starters, a quantum computer would be able to store more bits of information in its memory than there are particles in the universe. And where a conventional silicon-based computer handles one computation at a time in sequence, a quantum computer would work on millions at once.
That kind of staggering power would give a single quantum computer the ability to simulate a whole world in a holographic environment, replicate biological systems to understand diseases and find cures, solve the loads of equations necessary to create extremely accurate weather forecasting and simulate how subatomic particles interact, showing fundamentally how everything in the universe works.
NEWS: Quantum of Solar: Cells Get Mystery Power Boost
Several quantum computers linked together would make a quantum Internet so powerful that search engines would respond to queries almost like a human being, answering questions immediately and in any language.
In recent months, different groups of scientists and engineers have made important strides toward this amazing new world. They have built machines that can store quantum particles, control them, observe them and send them over fiber-optic cables.
For people in the field, it’s an exciting time. “We’re gradually removing the stumbling blocks,” said Bill Munro, a research scientist at Japanese phone giant NTT, who has done extensive research into quantum computing. “We’ve shown with the initial experiments that (quantum computing) can work.”
Some of the most recent work published in this area has come from scientists at Aalto University in Finland, who have found a way to store quantum particles, see them and change them.
Like conventional computers, quantum computers work by manipulating bits of information. In current computers and laptops, the bits are comprised of electrons, the magnetic fields of metal particles on a disk or the open and closed circuits on a microchip. They’re stored as “0s” or “1s” and long strings make the binary code that’s the essence of every program.
ANALYSIS: Magnetic Bacteria Could Build ‘Bio-Computers’
In quantum computers, the bits are actually not physical particles, but units of information called qubits that describe the state of particles, including atoms and subatomic particles, such as ions, electrons and photons. For example, a qubit might be represented by the direction in which an electron spins or the polarization of a photon of light – that is, how it’s oriented.
Qubits can be either a “0″ or a “1,” or both a “0″ and a “1″ simultaneously — a characteristic called superposition, which is what gives a quantum computer its extraordinary ability to process so much information at once. And like regular electronic bits, qubits need to be controlled and stored in order to get a desired input or output. You need some way to interface with them, just like you need a mouse or a keyboard to interface with the bits in a PC.
But there’s a major catch: qubits are easily disturbed by photons of light or heat or just about anything else in the natural environment. As soon as one tries to interact with a qubit, its value changes and it can even lose its crucial superposition characteristic.
In February, Mika Sillanpää and his colleagues at Aalto University reported that they had found a way to interact with a qubit without messing with its superposition.
NEWS: Quantum of Solar: Cells Get Mystery Power Boost
They built a tiny device that simulates the quantum state of a single atom. Sillanpää calls his device a kind of “artificial atom” (above). It consisted of a tiny piece of aluminum attached to a bit of sapphire. The scientists connected this component to a small piece of material capable of vibrating, called a resonator.
They put both components into a small cavity and cooled them to just above absolute zero. That turned the aluminium into a superconductor. Superconductors are known for conducting electricity with no resistance and can also behave as single atoms, entering a quantum state.
When the aluminium entered a quantum state, its energy made the resonator vibrate in a particular way. The vibration stored the quantum state information, or qubit. At the same time, it transferred energy into the cavity, which emitted a microwave photon that could be detected. It was the first time anyone had turned a bit of quantum information into a mechanical motion. It’s like an electron inside a conventional computer being converted into a pixel of text on a screen.
Silanpää told Discovery News that theoretically, by reversing the steps and firing a microwave photon at the component, the scientists would be able to change the quantum state of the artificial atom. A successful experiment demonstrating this — next on his list — would be similar to having a keyboard that entered new information into a computer.
ANALYSIS: Magnetic Bacteria Could Build ‘Bio-Computers’
This kind of link, or interface, between the quirky energy states of quantum particles and the macro world of tangible computers is necessary if we’re ever going to harness quantum power.
NTT scientist Bill Munro said since the device allowed for reading and writing qubits, it was a big step toward a useful computing device.
Meanwhile, at Yale, in January, a team of physicists found a way to observe qubits without ruining their superposition. Instead of interacting with the qubits directly, the team took partial measurements of the particle’s quantum state. They still disturbed the qubit, but it was in a known way, so they could correct for it.
This research goes some way to building quantum computers. But alone, these machines wouldn’t make an Internet; they need to be connected and exchange information. That’s where sending qubits over long distances comes into play.
NEWS: Quantum of Solar: Cells Get Mystery Power Boost
At the University of Innsbruck, Andreas Stute and his colleagues did this with ionized atoms. The researchers put a single calcium ion between two highly reflective mirrors. They hit the ion with a laser, which changed its quantum state, writing a single qubit of information onto it. They then hit the ion with a second laser. The ion emitted a photon, which carried the qubit they wrote down a fiber optic cable.
Last year, a similar experiment, with un-ionized atoms of rubidium, was conducted at the Max Planck Institute of Quantum Optics in Germany. Stephan Ritter, a physicist there, led a group that transmitted the rubidium atom’s quantum state from one “node” of a network to another.
Both sets of experiments are important to building a quantum Internet, as they demonstrate that qubits can travel long distances.
ANALYSIS: Magnetic Bacteria Could Build ‘Bio-Computers’
Making a quantum computer and full-on Internet is still hard — and still some years away. But even with the challenges, it’s clear that quantum computers that outperform the familiar electronic ones are coming. It’s just a question of when.
“Many, though not all, of the fundamental questions about whether such computer are possible in principle have been answered,” Munro said. “Now we can get to real R&D.”
Scientists hope that one day in the distant future,
miniature, medically-savvy computers will roam our bodies, detecting
early-stage diseases and treating them on the spot by releasing a suitable
drug, without any outside help.
To make this vision a reality, computers must be
sufficiently small to fit into body cells. Moreover, they must be able to
“talk” to various cellular systems. These challenges can be best
addressed by creating computers based on biological molecules such as DNA or
proteins. The idea is far from outrageous; after all, biological organisms are
capable of receiving and processing information, and of responding accordingly,
in a way that resembles a computer.
Researchers at the Weizmann Institute of Science have
recently made an important step in this direction: They have succeeded in
creating a genetic device that operates independently in bacterial cells. The
device has been programmed to identify certain parameters and mount an
appropriate response.
The device searches for transcription factors – proteins
that control the expression of genes in the cell. A malfunction of these
molecules can disrupt gene expression. In cancer cells, for example, the
transcription factors regulating cell growth and division do not function
properly, leading to increased cell division and the formation of a tumor. The
device, composed of a DNA sequence inserted into a bacterium, performs a
“roll call” of transcription factors. If the results match
preprogrammed parameters, it responds by creating a protein that emits green
light – supplying a visible sign of a “positive” diagnosis. In
follow-up research, the scientists – Prof. Ehud Shapiro and Dr. Tom Ran of the
Biological Chemistry and Computer Science, and Applied Mathematics Departments
- plan to replace the light-emitting protein with one that will affect the
cell’s fate, for example, a protein that can cause the cell to commit suicide.
In this manner, the device will cause only “positively” diagnosed
cells to self-destruct.
In the present study, published in Nature’s Scientific
Reports, the researchers first created a device that functioned like what is
known in computing as a NOR logical gate: It was programmed to check for the
presence of two transcription factors and respond by emitting a green light only
if both were missing. When the scientists inserted the device into four types
of genetically engineered bacteria – those making both transcription factors,
those making none of the transcription factors, and two types making one of the
transcription factors each – only the appropriate bacteria shone green. Next,
the research team – which also included graduate students Yehonatan Douek and
Lilach Milo – created more complex genetic devices, corresponding to additional
logical gates.
Following the success of the study in bacterial cells, the
researchers are planning to test ways of recruiting such bacteria as an
efficient system to be conveniently inserted into the human body for medical
purposes (which shouldn’t be a problem; recent research reveals there are
already 10 times more bacterial cells in the human body than human cells). Yet
another research goal is to operate a similar system inside human cells, which
are much more complex than bacteria.
With much progress being made in nanotechnology, the future of computers has two directions: nanotechnology and cells. Nanotechnology is the engineering of a system at the molecular scale. These processes are either “bottom-up” or “top-down”.
“Bottom -up” is the construction at the atomic level one atom at a time while “top-down” is using precise tools to achieve nanotechnological scales. Nanocomputing will give rise to four possible types of computers: electronic nanocomputers, biochemical and chemical nanocomputers, mechanical nanocomputers, and quantum nanocomputers. Nanotechnology allows for much smaller devices to be built without wasting space because it is built one atom at a time. For instance, silicon transmitters will be based on carbon nanofibers which are faster, smaller, and consumes less energy.
Advances in nanotechnology will have a significant impact on the environment, energy, healthcare, and medicine. There may be a future involving “nanobots” which would assemble products at the atomic scale and can turn one material into another, self-replicating, and being injected into the human body to repair disease at the cellular level.
Nanotechnology is projected to be generating trillions of dollars in the near future, with many companies already reaping the benefits. A patent moratorium may soon be in place to aid growth in the field. As wonderful as this new technology is, we must remember to keep in mind the pros and cons.
The pros of nanotechnology are that it will allow humans to create anything faster, smaller, and better. It will help stop disease and aid in energy. But, the cons are that a strong set of ethical standards will be needed to govern the new technology. For example, nanorobots can fall into the wrong hands and be used against us instead of for us, which must be taken into consideration.
So, when can we expect to see these advances? Well, we don’t know. But, we are currently on our way to the last and fourth generation which takes place 2015-2020 and that is molecular nanosystems: molecular systems by design, atomic design, and emerging functions. So in the very near future we will be reaping the benefits from nanotechnology and will begin taking ethics into much more consideration as it advances. With much progress being made in nanotechnology, the future of computers has two directions: nanotechnology and cells. Nanotechnology is the engineering of a system at the molecular scale. These processes are either “bottom-up” or “top-down”.
“Bottom -up” is the construction at the atomic level one atom at a time while “top-down” is using precise tools to achieve nanotechnological scales. Nanocomputing will give rise to four possible types of computers: electronic nanocomputers, biochemical and chemical nanocomputers, mechanical nanocomputers, and quantum nanocomputers. Nanotechnology allows for much smaller devices to be built without wasting space because it is built one atom at a time. For instance, silicon transmitters will be based on carbon nanofibers which are faster, smaller, and consumes less energy.
Advances in nanotechnology will have a significant impact on the environment, energy, healthcare, and medicine. There may be a future involving “nanobots” which would assemble products at the atomic scale and can turn one material into another, self-replicating, and being injected into the human body to repair disease at the cellular level.
Nanotechnology is projected to be generating trillions of dollars in the near future, with many companies already reaping the benefits. A patent moratorium may soon be in place to aid growth in the field. As wonderful as this new technology is, we must remember to keep in mind the pros and cons.
The pros of nanotechnology are that it will allow humans to create anything faster, smaller, and better. It will help stop disease and aid in energy. But, the cons are that a strong set of ethical standards will be needed to govern the new technology. For example, nanorobots can fall into the wrong hands and be used against us instead of for us, which must be taken into consideration.
So, when can we expect to see these advances? Well, we don’t know. But, we are currently on our way to the last and fourth generation which takes place 2015-2020 and that is molecular nanosystems: molecular systems by design, atomic design, and emerging functions. So in the very near future we will be reaping the benefits from nanotechnology and will begin taking ethics into much more consideration as it advances.
Piper Jaffray analyst Gene Munster notes in a research report this morning that there has been recent speculation from some tech blogs that Apple could launch a watch as a companion device to the iPhone. And he thinks that could be the start of a huge, important trend for Apple.
“While we are unsure of the ultimate launch timing (likely 2014 or later), we believe that Apple will eventually introduce some type of wearable computing product,” he writes. “As we have previously noted, we believe that wearable computers will ultimately be a major future trend. We expect Apple could profit from the trend in two ways. First, the company could create products for consumers, like the watch. Second, we believe the company could expand its MFi program that licenses hardware manufacturers the ability to make products that connect to iOS devices. While we don’t believe the watch itself is something that will excite investors, we believe the trend offers future revenue potential beyond the iPhone/iPad franchise.”
Munster actually argues that some time over the next 10-plus years “wearable computers could eventually replace the iPhone and smartphones in general.” His view is that “technology could progress to a point where consumers have a tablet plus wearable computers, like watches or glasses, that enable simple things like voice calls, texting, quick searches, navigation, etc. through voice control.”
Longer term, he adds, “screens in glasses or projectors could replace the necessity of a screen from a smartphone or tablet. These devices are likely to be cheaper than an iPhone and could ultimately be Apple’s best answer to addressing emerging markets.”
A new type of machine could rival quantum computers in exceeding the power of classical computers, researchers
say.
Quantum computers rely on the bizarre properties of atoms and the other construction blocks of the universe.
The world is a fuzzy place at its very smallest levels — in this realm where quantum
physics dominates, things can seemingly exist in two places at once or spin in
opposite directions at the same time.
The new computers rely on “boson” particles, and resemble quantum computers, which differ from
traditional computers in important ways. Normal computers represent data as
ones and zeroes, binary digits known as bits that are expressed by flicking
switch-like transistors on or off. Quantum computers, however, use quantum
bits, or qubits (pronouced “cue-bits”), that can be on and off at the
same time, a state known as “superposition.”
This allows the machines to carry out two calculations simultaneously. Quantum physics permits such behavior because it
allows for particles that can exist in two places at once or spin in opposite
directions at the same time.
In principle, quantum computers could solve certain problems much faster than can classical computers, because
the quantum machines could run through every possible combination at once. A
quantum computer with 300 qubits could run more calculations in an instant than
there are atoms in the universe.
However, keeping qubits in superposition is challenging, and the problem grows more difficult as more
qubits are involved. As such, building quantum computers that are more powerful
than classical computers has proven very difficult.
Now, though, two independent teams of scientists have built a novel kind of device known as a boson-sampling
computer. Described as a bridge between classical and quantum computers, these
machines also make use of the bizarre nature of quantum physics. Although
boson-sampling computers theoretically offer less power than quantum computers
are capable of producing, the machines should still, in principle, out-perform
classical computers in certain problems.
In addition, a boson-sampling computer does not require qubits. As such, “it’s technologically far
simpler to create than building a full-scale quantum computer,” said
researcher Matthew Broome, a quantum physicist at the University of Queensland
in Australia.
Boson-sampling computers are actually a specialized kind of quantum computer (which is known more formally
as a universal quantum computer).
“The main difference between boson-sampling computers and universal quantum computers is that boson-sampling
quantum computers can’t solve a universal set of problems like universal
quantum computers can,” Broome said. “But they are still conjectured
to be able to solve problems that would be massively intractable for classical
computers. Boson sampling computers are an intermediate model of a quantum
computer.”
Boson-sampling computers are not based on qubits, but on particles called bosons. “In our case, we use
photons,” said researcher Ian Walmsley, a quantum physicist at the
University of Oxford in England. Photons are the packets of energy that make up
light, and are one type of boson.
Broome and Walmsley were in separate groups that each devised a boson-sampling computer, based off concepts
first described by theoretical computer scientist Scott Aaronson at MIT. The
computers involve multiple devices that can each generate single photons. The
photons are inserted into a network where they can interact with one another.
They emerge from outputs equipped with sensors to analyze the particles.
The task of calculating which outputs these photons will emerge from, an operation known as boson sampling,
grows well beyond the capabilities of classical computers the more photons are
involved. The new computers accurately resolved what paths the photons would
take — three photons with Broome and his colleagues’ machine and four in
Walmsley and his collaborators’ device.
Since boson-sampling computing is in its infancy, it remains uncertain whether these computers can solve problems
beyond boson sampling. Still, this research suggests that computers based on
quantum physics could indeed tackle problems beyond the reach of classical
computers.
Previously, there was nothing to say “that anything you can do on a quantum computer you can’t do on a
normal computer, which leaves in question the necessity for quantum
computers,” Broome said. “Now, with boson sampling, we’re coming up
with machines based on quantum physics that can attack problems strongly
believed to be intractable for classical computers.”
In the future, “it would be great to push these computers toward more photons to tackle problems that would
be challenging to simulate on normal computers,” study coauthor Walmsley
added. Using about 20 to 30 photons would be a problem beyond the capabilities
of classical computers.
Both research teams detailed their findings online Dec. 20 in the journal Science.
The UK’s national museum of computing has restored a 61-year-old computer
to full clacking, blinking, and punch-card-reading working order. It’s noisy
and, as you can see in the video below, it’s awesome.
The computational machine is a 5500-pound
monstrosity with 828 Dekatron valves — think old-fashioned RAM in which you can
actually see the process of memory storage — 480 relays, which are
electromagnetic switches, and 199 lamps, which blink on and off as the computer
runs its calculations. Programs are input via punch cards, and results are
outputted to a teleprinter — sort of like an old-fashioned typewriter.
The Harwell Dekatron computer was used for atomic
energy research, automating calculations that previously had to be performed by
hand. Interestingly, it was a decimal computer, not binary.
Although very slow — it took five to ten seconds to
multiply two numbers — it was also very reliable, running an average of 80
hours a week and once, according to Wikipedia, running for ten days straight
over a Christmas/New Year’s holiday.
“In 1951 the Harwell Dekatron was one of perhaps a
dozen computers in the world, and since then it has led a charmed life
surviving intact while its contemporaries were recycled or destroyed,” museum
trustee Kevin Murrell said.
Check out this 50-second video — the Dekatron
sounds more like a printing press than a computer:
<iframe width=”560″ height=”315″ src=”https://www.youtube.com/embed/vVgc8ksstyg” frameborder=”0″ allowfullscreen></iframe>
The Harwell Dekatron was used until 1957, at which
point it was given to the Wolverhampton and Staffordshire Technical College,
which used it to help educate students until 1973. After going on display
temporarily in a museum, being dismantled, and stored, it was discovered by
volunteers from the National Museum of Computing four years ago.
The machine is now on display at the museum
As expected, this week’s Supercomputing 12 show in Salt Lake City, Utah, had lots
of a news. It announced the fastest system in the world and a different system
that is the most power-efficient of the large computers.
As usual, the important TOP 500 list is up and,
as anticipated, the new Titan system at the U.S. Department of Energy’s Oak
Ridge National Laboratory (ORNL) took the top spot. It shows sustained
performance of 17.5 petaflops (more than 17,500 trillion floating point
operations per second) and peak performance of over 27 petaflops on the LINPACK
benchmark. This system is based on a Cray XK7 system with 18,688 nodes, each
containing a 16-core AMD Opteron 6274 and an Nvidia Tesla K20 graphics
processing unit (GPU) accelerator.
In second place is the previous leader, the
Sequoia system at the Lawrence Livermore National Laboratory, based on IBM’s
BlueGene/Q system and its PowerPC CPUs. Third up is the “K computer”
at Japan’s RIKEN Advanced Institute for Computational Science based on Fujitsu
SPARC64 processors.
In addition to Titan, perhaps the most notable
change is that the Stampede system at the Texas Advanced Computing Center
(TACC) at the University of Texas in Austin is the first supercomputer in the
top 10 using Dell PowerEdge servers with Intel Xeon Phi processors. It comes in
at 2.7 sustained petaflops, but is supposed to get to 10 petaflops when fully
deployed next year.
Overall, there are now 23 systems on the list
with a performance above one petaflop; just over four years ago there was only
one. Sixty-two of the top 500 are now using accelerators or co-processors.
Intel provides processors for 379 systems on the list, AMD Opteron is used in
61 systems, and IBM Power processors are used in 53 systems.
Also updated this week was the Green 500 list,
which ranks the most efficient computers based on flops per watt running the
Linpack benchmark. This time, the top system is a relatively small cluster
(#253 on the TOP500 list), the Beacon system run by the National Institute of
Computational Science and the University of Tennessee. It is built using
Appro’s GreenBlades, each running 2.6GHz Xeon E5-2670 processors (8 cores/16
threads) and Xeon Phi 5110P co-processors, using the Infiniband FDR
interconnects. This results in 2.5 gigaflops per watt.
Second place went to Saudi Arabia’s Sanam system,
which uses Adtech blades, Xeon E5-2650′s and AMD FirePro S10000 accelerators.
Third place went to the Titan system and fourth to a smaller but similar Cray
system called Todi, based in Switzerland. It’s interesting to see that the next
24 spots all go to systems built on the more homogeneous IBM BlueGene/Q
architecture, running at 2.1GF/ watt. In general, it looks like performance and
power requirements are both going up, but performance is growing more quickly.
