Coming out of the Vision Lab at the University of Antwerp, the FASTRA II is based on gaming hardware and designed with tomography, "a technique used in medical scanners to create three-dimensional images of the internal organs of patients," in mind. The Vision Lab team began building the FASTRA and the FASTRA II because tomographical image reconstruction can take up to weeks on an ordinary computer, but in the end they produced this reasonably small 6000 Euro supercomputer. Not too shabby for something that was mainly intended to reduce their processing times. [Fastra—Thanks, Kjoost!]

This isn't the first time that the DoD is using PS3 consoles for supercomputing. In fact, these 2,200 units are going to be added to an existing Linux cluster of 336 PS3s used by the United States Air Force. According to Justification Review Documents, the purchase is all about getting the best value out the DoD's budget:

With respect to cell processors, a single 1U server configured with two 3.2GHz cell processors can cost up to $8K while two Sony PS3s cost approximately $600. Though a single 3.2 GHz cell processor can deliver over 200 GFLOPS, whereas the Sony PS3 configuration delivers approximately 150 GFLOPS, the approximately tenfold cost difference per GFLOP makes the Sony PS3 the only viable technology for HPC applications.

I'm all for balancing cost and features, but isn't it just a bit curious that someone thought to save on upgrading the supercomputer just after Call of Duty: Modern Warfare 2 was released? [Ars Technica via Boing Boing]

Cats: they're kinda dumb. They only seem smarter than dogs because they're not so friendly, and our society judges kindness harshly. It's true! an interesting theory! Which is why, after mice, simulating a feline-sized brain on a BlueGene/P supercomputer was next on IBM's to-do list. But for all the kitty talk here, this project wasn't specifically about creating a computerized house pet; it's part of a larger, ongoing project to eventually simulate a full human brain. The cat equivalency, derived from the number of virtual neurons and synapses the simulation can manage, at 1.6 billion and 9 trillion, respectively, just gives a sense of how far along the project is: today, despite being the biggest simulated brain ever, it's only capable of simulating the human visual cortex, or as PopMech so delicately puts it, "the wrinkly outer layer" of the human brain.

So how long before a supercomputer can simulate (roughly—since these computer simulations don't have the same neural patterning and learning capabilities of a real brain, among other things) an entire human cortex? Weirdly soon, says the project's lead scientist:

To [simulate a human cortex], he'll need to find 1000 times more computing power. At the rate that supercomputers have expanded over the last 20 years, that super-super computer could exist by 2019. "This is not just possible, it's inevitable," Modha says. "This will happen."

People need to stop getting worked up about the future, honestly: Before we have to worry self-aware robot uprisings, we're going to have to deal with decades of extremely dumb, extremely expensive fake pets. Enforced caution, I believe this is called. [Popular Mechanics]

Then it surpassed IBM's 1.04 petaflop/s supercomputer, achieving its 1.74 quadrillion floating points operation according to the Top500 Linpack benchmark. I mean, it was probably some extra CPUs coming online, but I'm pretty sure the main reason for the boost was that drawing.

Roadrunner took a dip from June's 2009 test, which gave it 1.105 petaflops. That's probably Wile E. Coyote's fault. My recommendation to IBM: Paint flames on it to win the #1 spot back.

This is the current top ten:

1. Jaguar, Cray, Oak Ridge National Laboratory, USA (1.75 petaflop/s)
2. Roadrunner, IBM, Los Alamos National Laboratory, USA (1.04 petaflop/s)
3. Kraken XT5, Cray, National Institute for Computational Sciences, USA (832 teraflop/s)
4. JUGENE, IBM, Forschungszentrum Juelich, Germany (825.5 teraflop/s)
5. Tianhe-1, NUDT, National SuperComputer Center, Tianjin, China (563.1 teraflop/s)
6. Pleiades, SGI, NASA Ames Research Center, USA (544.3 teraflop/s)
7. BlueGeneL, IBM, Lawrence Livermore National Laboratory, USA (478.2 teraflop/s)
8. BlueGene/P, IBM, Argonne National Laboratory, USA (458.61 teraflop/s)
9. Ranger, Sun, Texas Advanced Computing Center, USA (433.20 teraflop/s)
10. Red Sky, Sun, Sandia National Laboratories, USA (423.9 teraflop/s)

[Top500 Supercomputers via Cnet]

The supercomputer, dubbed simply as a "Question Answering" system, is named Watson. Designers believe it will have the speed and "understanding" necessary to research, buzz in, and then answer questions fast enough to compete on the popular game show. But it's not Google, engineers say. Don't you dare call it Google.

And as you can see in the video, Alex Trebek is totally for it, going so far as to appear as though all his lines were prepared by IBM PR. Anything to avoid having to face off against your mortal enemy Sean Connery, eh Trebek? [YouTube - Thanks, blam!]

So what's all that actually mean? IBM offered us some more tangible ways to wrap your mind around 20 quadrillion mathematical processes per second.

• If each of the 6.7 billion people on earth had a hand calculator and worked together on a calculation 24 hours per day, 365 days a year, it would take 320 years to do what Sequoia will do in one hour.

• 20 petaflops could offer a 50x improvement in our capability to predict earthquakes, allowing scientists to predict an earthquake's effects on a building-by-building basis across an area as large as Los Angeles County.

• 20 petaflops could also provide a 40x improvement in our capability to monitor and forecast weather. This would allow forecasters to predict local weather events that affect areas 100 meters to one kilometer in size, down from their current ten-kilometer ability.

The Sequoia will be powered by 1.6 million cores (specific 45-nanometer chips in development) and 1.6 petabytes of memory. It will be housed in 96 refrigerators spanning roughly 3,000 square feet.

It's for the U.S. Government who will use the system for "uncertainty quantification (UQ) studies" and weapon science calculations. [IBM Supercomputing]

The Mysterious Case of the Disappearing Atom Mass

The night that the Frenchy called me I didn't have any plans. Susan took the day off for shopping. Something about new stockings. I said yes. She never seemed to have enough of those. I never had enough of her in them either. Taking her down to the club for the usual bourbon and dancing was out of the question. Maybe that's why I said yes to Lellouch. I never was fond of the froggies. Not even while I was shooting Nazis in Normandy.

Laurent Lellouch. That was the name. I liked it as much as the sound of the case he wanted me to take: Nothing at all. Something about a war between gangs of Prussian gangsters, the Neutrons and the Protons. I didn't know them. It was all weird and related to that stuff they did at Los Alamos and then dropped in Japan. I knew Uncle Sam wasn't going to be far behind this one, but Louis said he was ok to trust him. A bit. I didn't have anything better to do, anyway. Pork chili down at George's while listening to what Lellouch had to tell me was a better plan than going with the boys to the 42nd. I looked out the window and saw it was still raining nails. Hot chili was it.

When I arrived, Lola nodded behind the bar and looked to the table where the guy was waiting. She rolled her eyes and shouted the usual order to George at the kitchen. The Frenchman was nervous, mumbling something about international conspiracies and computers and that guy from Germany who turned everything inside out with his theories. That equation. E=mc2. The told me about the protons and the neutrons. While I was downing my chili he went on and on about it. Inside those families there were quarks, which are bound by gluons. I didn't have a clue what he was talking about. The mass of a gluon is zero, he said, while the mass of the quarks is only five percent. So, where is the missing 95 percent?

Maybe he was onto something. I finished my chili, dropped a couple of Washingtons, and went on to see Janos the Hungarian. He wasn't going to talk. Fortunately for him, I'm a reasonable man. It was nothing that a simple knuckle kiss couldn't fix. Ten minutes and three teeth later he spilled. The key is in the quantum chromodynamics, something about equations running at the sub-atomic level. More gibberish, but I know he was telling the truth. I left him trying to fix his bloody nose and went to meet the Germans. I knew that if anyone had the answer, it was going to be Otto.

I was right. He knew about Janos, so I didn't have to get nasty again. Too bad. I was thinking about how much I wanted to see Susan in her new stockings. Wasting my time listening to this was making me angry. Otto said that the unaccounted mass came from the energy from the movements and interactions of quarks and gluons. The computations involved envisioning space and time as part of a four-dimensional crystal lattice, with discrete points spaced along columns and rows.

I still didn't know what the hell he was talking about, but I crossed the street to call the Frenchy. I had his answer. When he picked the phone he was excited like a little girl in her first date at the back of the movie theater. He wanted to meet right away. Get all the details. I just wanted to get my money and go meet Susan at her place. I told him to meet me at the park, on the corner of Fifth and 64th.

He was there when I arrived, sitting on a bench with a stupid smile in his face. He'd had a lead overdose. Someone got him before I could tell him that Einstein was right. E=mc2 was corroborated for the first time thanks to those computers they stole from the Germans and the Hungarians. I don't know who killed him. Probably the CIA. Or the KGB. Maybe the Italians. Or all of them. I knew it was time for some silk and alcohol. I took the envelope he still had in his coat and I closed his eyes. There are things that mere mortals don't need to know. And none of them were Susan's legs. [AFP]

Both the IBM and Cray systems break the petaflop processing barrier according to Top500 measurements (1.45 petaflops vs 1.38 petaflops, respectively). Heck, even IBM admitted to us that the two computers "run neck and neck." But there's a huge difference between them.

The Roadrunner uses roughly half the power of the Jaguar XT.

It assembles 12,960 IBM PowerXCell 8i Cell Broadband Engine processors and an additional 6,948 AMD Opteron Dual-Core processors. The AMD equipment handles "basic" functions while the IBM chips handle the intense number crunching. (Read all about the Roadrunner here.)

Seeing as the Cray XT5 uses 45,000 quad-core AMD Opteron processors to get the same job done, you've gotta be at least a little impressed. [Top500]

The system is powered by 45,000 quad-core AMD Opteron processors that take advantage of 362 terabytes of memory. This and other underlaying architecture allows processors to chew on 284 gigabytes of data per second with its impressive I/O bandwidth, which has apparently been a major bottleneck in supercomputers of yesteryear. Information is stored on 750 terabytes of hard drives.

The Cray XT Jaguar can be found at the Department of Energy's Oak Ridge National Laboratory where it will create scientific breakthroughs during the day, and succumb to Crysis at night. [ORNL and EurkeAlert]

The network uses fiber optic transmission to send information to 11 primary data centers in Europe, North America and Asia. From these centers, the data is passed to 140 secondary centers globally.

The processing architecture not only distributes the heavy processing load to computers across the world, but it allows 7,000 scientists to share access to LHC data, to sift through the mountains of information for a nugget of valuable data. And when it comes to understanding the fundamental nature of our Universe, we'll take all the eyes we can get. [China View]

The "cool stuff" will take about three weeks to get up and running, with the more visual demos coming at the tail-end of that. The one benchmark they currently have is that it hits 768 Gigaflops, which they hope to bump over 800 with some fine-tuning. Moving from Nvidia's Quadro 4600 to their newer Tesla cards should give the system a jolt as well, since they're explicitly designed for parallel computing applications, like what the CX1 is designed for.

The CX1 can hold up to eight computing blades—though the storage and visual blade each take up two slots, so the model they were showing had four computing blades, and one of each. While each blade is highly customizable, the cheapest one they had configured was about $4,000, and a fully spec'd out CX1 goes for about $85,000 (slightly higher than they originally announced). While it's not actually designed for gaming at all, for that much I'd want it to burn Crysis directly into my brain. [Cray]

You almost missed it there right next to the desk, didn't you? When it's not badly Photoshopped in a Cray brochure, the CX1 packs 16 Intel Xeon procs, either dual- or quad-core (you choose), with 8 supercomputing nodes that can accommodate 64GB of memory per node. Internal storage tops out at 4TB. You can custom-configure and purchase one today, on ranges from $25,000 to $6000.

Microsoft and Cray Team Up to Drive High Productivity Computing Into the Mainstream
Cray CX1 Supercomputer With Windows HPC Server 2008 and Intel Xeon Processors Starts at $25,000 and Provides "Ease-of-Everything" for New Users of HPC
SEATTLE, WA and REDMOND, WA, Sep 16, 2008 (MARKET WIRE via COMTEX News Network) — Supercomputer leader Cray Inc. (NASDAQ: CRAY) and Microsoft Corporation (NASDAQ: MSFT) today introduced the new Cray CX1 supercomputer pre-installed with Windows HPC Server 2008. With U.S. list prices starting at $25,000 to over $60,000, "ease-of-everything" features and the ability to fit into standard office environments and workflows, the new product reflects Microsoft and Cray's shared goal to drive high productivity computing farther into the mainstream in a broad array of markets including financial services, aerospace, automotive, petroleum, life sciences, government, academic and digital media.

Studies released by the Council on Competitiveness and the Defense Advanced Research Projects Agency (DARPA) found that nearly all large firms using HPC consider it indispensable for their ability to compete and survive, but smaller companies, as well as workgroups and departments in larger firms, have been hampered by the cost of HPC systems and a lack of access to in-house experts to help them use these systems.

The Cray CX1 supercomputer was designed from the ground up to address these barriers. It is the most affordable supercomputer ever offered by Cray and is designed to be easy to purchase, deploy, operate and upgrade. Purpose-built for offices, laboratories and university departments, the Cray CX1 is the world's highest-performing computer that uses standard office power.

The Cray CX1 product incorporates up to 8 nodes and 16 Intel Xeon processors, either dual or quad core; delivers up to 64 gigabytes of memory per node; and provides up to 4 terabytes of internal storage. Systems can be configured with a mix of compute, storage and visualization blades to meet customers' individual requirements. The quiet, deskside supercomputer features Windows HPC Server 2008 and interoperates with Linux. A three-year warranty with next-day, on-site Cray-certified support is standard.

"Windows HPC Server 2008, in combination with the Cray CX1 supercomputer, will provide outstanding sustained performance on applications," said Vince Mendillo, director, HPC at Microsoft Corp. "This combined solution will enable companies in various sectors to unify their Windows desktop and server workflows. Many Microsoft financial services customers, for example, want to unify back-office modeling and simulation with the work of front-office trading desks."

"IDC research shows that HPC has been one of the highest-growth IT markets during the past five years and the segment for HPC systems priced below $100,000 is headed for continued growth," said Earl Joseph, IDC's HPC program vice president. "The Cray HPC brand name and experience, combined with Microsoft's strategy of extending the familiar Windows environment upward to the server level, gives the Cray CX1 solution strong potential for exploiting the anticipated growth of this market segment."

"Cray sees Microsoft Windows becoming an increasingly important force in the HPC market," said Ian Miller, senior vice president of sales and marketing at Cray. "With the Cray CX1 high productivity system and Windows HPC Server 2008, we're bringing the power of Cray supercomputing to a much wider range of new users with an affordable and adaptable system that provides incredible value and is easy to install, program and use with a broad array of applications from independent software vendors (ISVs)."

The Cray CX1 high productivity system is also the first Cray product to incorporate Intel processors and the first milestone of the unique collaboration Cray and Intel announced in April to develop a range of HPC systems and technologies over the next several years to address various segments of the HPC industry.

"Taking advantage of the energy-efficient performance of the Intel Xeon processor 5400 series, Cray's CX1 system will bring many HPC capabilities to the office that were previously confined to the datacenter, enabling more users to employ supercomputing to help them solve some of their most difficult computational problems," said Richard Dracott, Intel's General Manager of High Performance Computing, "In addition, we continue to collaborate with Cray on developing the supercomputing technologies of the future, aimed at all segments of the HPC market."

Scientists at the Laboratory of Neuro Imaging at UCLA plan to use a Cray CX1 with Microsoft HPC Server 2008 for mathematical modeling and visualization. This will support their development of advanced computational algorithms and scientific approaches for the comprehensive and quantitative mapping of brain structure and function.

"We are very excited about utilizing the Cray CX1 to support our research activities," said Rico Magsipoc, Chief Technology Officer for the Laboratory of Neuro Imaging. "The work that we do in brain research is computationally intensive but will ultimately have a huge impact on our understanding of the relationship between brain structure and function, in both health and disease. Having the power of a Cray supercomputer that is simple and compact is very attractive and necessary, considering the physical constraints we face in our data centers today."

[Cray]

Designed from video game components, and costing $133 million, Roadrunner contains 12,960 chips redesigned from an I.B.M. Cell microprocessor that was originally created for Sony’s PlayStation 3 video-game machine. Add to that a bunch of Opteron processors from Advance Micro Devices, which are commonly found in corporate servers, and there's your supercomputer.

It runs on around three megawatts of power—around the amount that a shopping mall needs if it is to function properly—and needs three separate programming tools to run the trio of different processors. The complicated bit for programmers is to keep all 116,640 processor cores occupied simultaneously, or else the supercomputer does not run effectively. [New York Times]

This kind of setup works really well for tomography because the number-crunching can be done in parallel and is highly vectorized—the same kind of stuff the medical community and Air Force were eye-balling the PS3 for, since the Cell uses a similar kind of architecture.

On the other hand, it wouldn't be so great for more general computing stuffs that can't be crunched in parallel (multiple processors working at once). Either way, watch the video, gigaflops to terabytes, it's the nerdiest thing you'll see this week. [FASTRA, Thanks Toji]

So it can be easily transported to Area 51 to help with extraterrestrial communication efforts. OK, we lied about that, but we do love a good conspiracy theory. Did you hear the one about the Freemasons wanting Steve Jobs to design an exclusive iPod with subtle Freemason markings? Apparently, the FCC label on your iPod is an acronym for "Freemasons Capturing Consumers." Shocking. Be less gullible. [Akihabara]

[TMC Net via The Raw Feed]

To be located in Chicago, the computer would reportedly cost not only $200 million to build, but over $400 million to maintain over its five-year lifetime. It'll be used for only a small number of "Grand Challenge" science projects, such as simulating global warming and playing Crysis at 60fps. Your quad-core XPS tower just got a lot less impressive. [Boston via The Raw Feed]