Crave

Intel Xeon processors--particularly the "Harpertown" variety--dominated the top 500 supercomputer list. But IBM's Power chips made a strong showing as usual at the very top of the list. AMD's Opteron processor landed in the No. 1 and No. 4 ranked systems.

Top 10 processors in Top500 supercomputer list

(Credit: Top500.org)

The Top500 List--updated twice a year--of supercomputers was released Wednesday. Intel's Xeon, AMD's Opteron, and IBM's Power chips vied for most of the spots in the list.

The most dominant chip was the Intel Xeon E54xx series "Harpertown" processor. Appearing in 116 systems for 23.2 percent of the total. The largest for any single processor model.

The Xeon 53xx series "Clovertown" processor was next, appearing in 92 systems for 18.4 percent of the total. Following Clovertown was the Xeon 51xx series "Woodcrest" processor with 18.2 percent of the total.

Harpertown and Clovertown are quad-core processors, Woodcrest is dual-core.

In the No. 4 slot was the AMD Opteron dual-core chip (8.4 percent), followed by the X54xx series of Intel Harpertown processors (7.8 percent), then by the PowerPC 440 (4.22 percent).

(Note: Combining the Intel Harpertown E54xx series and X54xx series boosts the total for this chip model to 31 percent.)

The IBM Power processors passed the AMD Opteron family and "are now (again) the second most common processor family with 68 systems (13.6 percent), up from 61 systems (12.2 percent) six months ago," Top500.org said.

AMD's strongest showing was in the top five supercomputers. Opteron processors played a major role in the No. 1 IBM Roadrunner system, which connects 6,562 dual-core AMD Opteron chips as well as 12,240 IBM Cell chips (on IBM Model QS22 blade servers).

See: IBM's Roadrunner breaks petaflop barrier, tops supercomputer list.

The No. 4 Sun Microsystems' SunBlade system uses over 62,000 cores running inside AMD Opteron quad-core processors running at 2.0GHz.

The No. 2 and No. 3 systems were based on IBM PowerPC 450 chips.

Other Top500 processor highlights:

• A total of 375 systems (75 percent) are now using Intel processors. This is up from six months ago (354 systems, 70.8 percent) and represents the largest share for Intel chips in the Top500 ever.

• 56 systems (11 percent) are using AMD Opteron processors, down from 78 systems (15.6 percent) six months ago.

• 283 systems are using quad-core processor based systems.

IBM's Roadrunner supercomputer was named the fastest supercomputer in the world Wednesday after breaking the petaflop barrier earlier this month.

(Credit: IBM)

Good news for green tech: The fastest supercomputer in the world is also one of the most energy efficient. That's according to the Top500 supercomputers list, to be released Wednesday at the International Supercomputing Conference in Dresden, Germany.

Twice yearly, the list measures the 500 most powerful computer systems available commercially. This year, the 31st time the list has been put together, the honor of top supercomputer goes to IBM's Roadrunner, which is housed at the U.S. Department of Energy's Los Alamos National Laboratory. It's the first system to reach 1.026 petaflops (1 petaflop is equal to a quadrillion, or one thousand trillion, calculations per second).

For perspective, last year's most powerful computer, Lawrence Livermore National Laboratory's BlueGene/L--also made by IBM--reached 208.6 teraflops. This year that computer ranked No. 2, reaching a max processing speed of 478.2 teraflops.

Fun fact: the fastest supercomputer in the world--used to monitor the U.S. nuclear weapons stockpile--is really just a PlayStation 3 on steroids. Roadrunner is based on the IBM QS22 blades, which are built using advanced versions of the Cell processor in Sony's PS3. It also runs using x86 chips from Advanced Micro Devices, making it the world's first hybrid supercomputer.

In total, Roadrunner takes up 278 refrigerator-size server racks, and connects 6,562 dual-core AMD Opteron and 12,240 Cell chips.

IBM, which continues its dominance of supercomputing, makes 210 of the 500 systems, including 5 of the top 10. Hewlett-Packard is close behind, however. HP makes 183 of the fastest computers, including the No. 8 fastest system known as EKA, located in Computational Research Laboratories' data center in Pune, India.

Rounding out the top 10 is Sun Microsystem's Ranger at No. 4, Cray's Jaguar at No. 5, SGI's Encanto at No. 7, and SGI's Altix at No. 10.

On the processor side, Intel dominates the high-end market with 75 percent of all systems on the list and 90 percent of the quad-core based systems that were ranked.

Supercomputing, which pits the highest-end machines against challenges such as forecasting the global climate in coming decades or finding oil reservoirs underground, is a fast-changing field. The Top500 list once again had the most turnover compared with the preceding list, according to the researchers who compile it.

The main measurement used in compiling the list is the Linpack measurement, which puts each system through its paces by having to solve a dense system of linear equations.

The Top500 acknowledges that Linpack isn't a complete test of system performance, but it's a way to test for performance on a similar problem across each system. The need for a more complete benchmarking system has been under discussion for several years.

Some additional interesting statistics about the June 2008 list:

* Quad-core processors are used in just over half of the systems.

* The bulk of the systems (208 of the 500) contain between 2,049 and 4,096 processors. That's more than double the systems that used that amount just six months ago.

* Four of the top five computers (Nos. 1, 2, 3, and 5) are located in U.S. Department of Energy labs.

* The U.S. continues to be home to the most computing power in the world. Just over half of the systems (257) are located in the U.S. The U.K. is next with 53, followed by Germany with 46, France with 34, Japan with 22, and China with 12.

After "not specified," the most popular application area for these superfast computers is finance (15.2 percent of the list), followed by research (10 percent), geophysics (9.8 percent), information service (6.2 percent), and service (5.2 percent).

Scientists from the IBM Zurich Research Lab and the Fraunhofer Institute in Berlin are working on a microchip that uses micropipes of water to cool itself, IBM announced Thursday.

The chip's components are built in a 3D stack instead of side by side on a silicon wafer.

This diagram illustrates the chip-cooling concept. Water in a cooling container (purple) is pumped through integrated spaces between the chip's layers (orange).

(Credit: IBM)

Chips built in a three-dimensional stack formation offer more pathways for info to be processed and can shorten the distance chip information needs to travel by as much as 1,000 times, according to Thomas Brunschwiler, a senior engineer in the Advanced Thermal Packaging Group at the IBM Zurich Research Lab who has been working on the chip for almost two years.

The trouble is that this type of experimental chip structure also generates a large amount of heat.

To address that problem, the team has developed a cooling system consisting of micropipes of water as thin as a human hair (50 microns) that are interspersed between each chip layer.

To prevent an electrical short, the hairlike water pipes are hermetically sealed from the chip's other components first with a silicon wall and then with a layer of silicon oxide, according to Brunschwiler.

To bond the individual pipes from layer to layer without damaging other chip components, the scientists used a solder consisting of a mixture of gold and tin, which has a low melting point.

"This process enabled us to completely seal off the joints. Then we can use water, which is superior to other coolants," Brunschwiler said.

The water-cooled chip, which is intended for use in supercomputers, is 5 to 10 years away from being commercially available. "But before that, one would probably see chips with one core layer and a memory layer sitting on top that can still be cooled with (an) outside system," Brunschwiler said.

While unique in its microscopic scale, IBM's use of water to cool down the heat generated by computer processing is not novel.

Companies like IBM and Hewlett-Packard sell server racks with liquid cooling systems. Researchers at Ireland's Tyndall Institute and University of Limerick announced in March that they are working on a liquid cooling system incorporated into the packaging that encases chips. And in April, IBM announced a supercomputer that uses water alongside its processors to cool them.

Chips in 2009 will run faster but not necessarily hotter. That's the gist of what IBM, along with its joint development partners such as Samsung Electronics and Toshiba, announced Monday.

IBM-fabricated 32nm SRAM chip

(Credit: IBM)

The IBM alliance is using "high-k/metal gate" technology to achieve this, the same category of process technology that Intel currently uses in its 45-nanometer processors. The alliance says it is seeing performance improvements of up to 35 percent over 45nm technology at the same "operating voltage" or power levels.

This allows alliance chipmakers such as Samsung, Toshiba, and Freescale Semiconductor (formerly an arm of Motorola) to build more powerful chips that don't necessarily generate more heat. This is a necessary advancement for small devices such as cell phones, as well as data centers that use a large number of servers. The power reduction compared to 45nm technology can range as high as 30 percent to 50 percent, depending on the operating voltage, according to IBM.

The announcement serves as more of a progress update than a breakthrough achievement. At the beginning of last year, IBM announced technology to "speed the implementation of...high-k/metal gate in next-generation 32-nanometer computer chips."

A gate is a basic building block of a digital circuit, while high-k/metal is the material used. Intel, for example, used a high-k material called hafnium to replace the transistor's silicon dioxide gate dielectric for its 45nm processors. As transistors shrink, leakage current can increase. For chipmakers, it is crucial to minimize leakage. This is where high-k/metal gates come into play.

The technology will be available to alliance partners in the second half of 2009, though the design process for devices that use this technology can start now, IBM said. AMD was not included as part of the alliance because technically, it is a member only of a separate alliance, SOI (Silicon-On-Insulator).

IBM researchers have reportedly demonstrated technology that will increase hard drive capacity 100-fold, as well as offer major improvements in energy consumption (leading to much longer battery life) and better reliability. Production is estimated in seven to ten years.

In seven years, we'll be measuring hard drive capacity for portable devices in terabytes.

(Credit: Apple)

The reports summarizing the researchers' findings, which were published in Science (subscription required), use the shorthand "500,000 songs on a portable MP3 player" to describe the advance.

Today's iPod lineup contains no product advertised to hold 5,000 songs, so I'm not sure where the 500,000 figure came from. In fact, the current highest-capacity iPod is 160GB, and is advertised as being able to hold 40,000 songs. So this shorthand would imply a hard drive size of just under 2TB--only 12.5 times bigger than today's largest iPod.

That's actually well short of what Kryder's Law predicts--if hard drive capacity continues to double every year, then the hard drives of 2015 should be 128 times larger than today's. So the IBM researchers' claims of up to 100x capacity, while impressive, are not particularly surprising given the trends of the past decade. According to my calculations, 100x would mean the biggest iPod would have a 16,000 GB hard drive, which would be enough to hold more than four million songs at the current advertised compression rates. Or if you assume that Apple's lossless codec compresses the typical song to about 25MB, it could hold about 650,000 songs--with no loss in audio quality.

Of course, few people would use a portable hard drive of that size solely to store music--movies, games, and applications will probably take up most of that space. Still the idea that we'll be carrying terabytes of data in our pocket in a few short years explains why Apple, Microsoft, Google, and the rest of the industry are focusing so much attention on mobile computing.

IBM's latest supercomputer is hooked up to the watercooler.

Big Blue has come out with a new version of its high-end supercomputer, the Power 575, which can provide five times the performance of its predecessor on 40 percent of the power. A fully stocked Power 575 rack contains 448 processing cores.

An IBM technician inserts the copper piping.

(Credit: IBM)

A substantial part of the decrease in power consumption is due to a water cooling system that brings in chilled water from the outside, runs it through copper plates located above individual processors to absorb heat, and then draws the water out so it can expel the heat outside of the computer.

By getting rid of heat in this manner, the air conditioning requirements are greatly reduced for the "hydro cluster" 575. Air conditioning can account for roughly half of the power consumed by data centers. Conversely, instead of cutting electricity consumption, IBM, or one of its customers, could squeeze in more computing power into the same room and keep the air conditioning constant.

Computer makers have employed liquid cooling in various ways over the decades. Many liquids, and particularly water, can hold far more heat than air. Similarly, architects and building owners are experimenting more with liquid cooling and heating systems as energy prices rise.

"Water is about 4,000 times as efficient as air to cool a system," said Ross Mauri, general manager of Power systems at IBM.

The effectiveness of a water cooling system, however, depends largely on two parameters: how close you can get the fluid to the hot component and how cool you can get the liquid. In general, the closer the fluid to the chip and cooler the initial temperature, the better it works.

IBM, Hewlett-Packard and others have created blade server racks with integrated chilled water tubes. Chilled water circulates through the pipes but can't get as close to the hottest components.

The company has also created liquid cooling systems that fit inside computers and sit directly above hot components. These systems, however, have consisted of self-contained liquid vessels. The fluid heats up, rises, and then sinks again, but it stays moderately warm.

With the hydro cluster, "there is always chilled water in the system," Mauri said.

IBM isn't alone in its pursuit of brining liquid close to the chip. In February, the Tyndall Institute, a government-funded lab and incubator in Ireland, showed us an silicon impeller that can bring cooled liquids in close contact with chips. The impeller measures a few millimeters across.

The computer, along with a new Power 595 Unix server, sports a 5GHz chip, a speed bump over the existing 4.7GHz versions that have been on the market.

Unix and RISC servers, IBM wants you to know, aren't dead. In 2007, industrywide Unix server revenue grew 1 percent, Mauri said, the first time the market has grown in six years.

IBM also has been aggressively taking share from competitors Sun Microsystems and Hewlett-Packard, he added. In five years, IBM has gained 11 percent in market share, according to IDC numbers cited by Mauri, while HP and Sun have lost share.

AMD's quad-core Barcelona chip for servers is ready to launch--really launch this time.

AMD is shipping the quad-core Barcelona Opteron to channel and distribution partners this week, according to an AMD representative on Thursday. System vendors such as Dell and Hewlett-Packard are readying systems for shipment in the second quarter.

(Correction: AMD is shipping samples of the Barcelona processor to channel and distribution partners not production versions of the chip.)

AMD is shipping the B3 version of the processor that fixes a TLB bug. Large vendors such as Dell, HP, and IBM have been waiting for this version of Barcelona to arrive before they begin shipping systems.

HP ProLiant DL585 G2 Server

(Credit: Hewlett-Packard)

"We expect to start shipping systems in early Q2," HP spokesman Eric Krueger said Thursday. "We are anxious to get these systems moving soon," he added. HP has had documentation on its site for almost a month describing the HP ProLiant DL585, which Krueger confirms will carry the Barcelona processor.

"Look for us to expand our portfolio (of systems) too," Krueger added. A Dell representative confirmed that his company is also on track for shipment in the second quarter.

Barcelona was launched back in September and has faced repeated production and bug-related delays. Earlier "B2" versions of Barcelona have been going to sophisticated, high-performance computing (HPC) customers who know how to work around the bug.

IBM has come up with a technology that could one day let different cores on a processor exchange signals with pulses of light, rather than electrons, a change that could lead to faster and far more energy efficient chips.

The device, known as a silicon Mach-Zehnder electro-optic modulator--converts electrical signals into pulses of light. The trick is that IBM's modulator is 100 or more times smaller than other small modulators produced by other labs. Eventually, IBM hopes the modulator could be integrated into chips.

Electrons in, photons out.

(Credit: IBM)

Here's how it works. Electric pulses, the yellow dots, hit the modulator, which is also being hit with a constant beam of light from a laser. The modulator emits light pulses to correspond to the electrical pulses. In a sense, the modulator is substituting photons for electrons.

Since the beginning of the decade, several companies--Intel, Primarion, Luxtera, IBM--have been coming up with components that, ideally, will let chip designers replace wires in computers and ultimately chips with optical fiber. Wires radiate heat, a big problem, and the signals don't travel as fast as light pulses. (The research in this area is known as silicon photonics and optoelectronics.)

The problem with optical technology, however, is making it small. Optical components historically have been tricky to produce and tend to be fairly large. Computer makers need components that measure only a few millimeters on a side. The idea is to come up with a way to produce modulators, lasers, waveguides and other devices on silicon manufacturing lines.

Right now, it remains an open question when these products will come to market. Still, the plethora of prototypes is a strong indication that progress is moving along well.

If you haven't already, say goodbye to that little IBM sticker on your ThinkPad.

Turns out, Lenovo doesn't need the reputable computer brand to sell its notebooks and desktops anymore. Lenovo Chief Executive Bill Amelio said as much following the company's most recent earnings results, which was noted by E-Channel Line.

Now you see it, soon you won't: The IBM logo like the one on this ThinkPad will be no more.

(Credit: Lenovo)

"By making substantial progress on all of our critical priorities over the past few quarters, we're now a stronger, healthier company," Amelio said. "One important sign of this progress is our decision to completely transition our Think products from the IBM brand to the Lenovo brand two years earlier than planned." Previously there had been a deal made that allowed Lenovo to use the IBM logo on ThinkPads and some desktops for several years. The Chinese PC maker purchased IBM's computer business in 2005 for $1.75 billion.

Lenovo is currently dueling with Taiwan's Acer for the mantle of third-largest PC vendor in the world. Both are trying to build brand-name recognition globally, but also here in the U.S.