The previous post on the PITAC report that states US prowess in computational science is on the decline doesn’t mean all aspects are on the decline. Though raw computing power is just one component of computational science, to some it is the most important. In that vein, the latest Top500 list of supercomputers suggests that US dominance in this sector isn’t waning. With IBM once again flexing enormous muscle, Big Blue not only holds the first and second positions but they command nearly 60% of the total installed performance with over 50% of the total systems on the list. Hewlett-Packard is second in both performance and number of systems. The number of US installed machines on the latest list increased 5% since November’s list going from 267 to 294 machines, which continues a recent trend.

The latest Top500 list can be found at the Top500 website - http://www.top500.org/

Cray announced today that the Cray XT3 and Cray XD1 have posted leading overall results on the HPC Challenge benchmark tests. With the XT3, they’re claiming victory on seven of the 10 tests. According to Cray:

In comparing customer-reported HPC Challenge results for three large-scale systems of about the same size, an 1,100-processor Cray XT3 supercomputer had the best scores on seven of the 10 “condensed results” tests, compared to an SGI Altix 3700 system with 1,008 processors and an IBM Blue Gene system with 1,024 processors. In the seven tests, the Cray XT3 typically outperformed the next-best system by a factor of two to five times, and was up to 17 times faster than the third-ranking system.

Presentations from Spring Intel Developer Forum, including a great deal of talk about multicore processor plans, are available for download from Intel. (Search for “Dual/Multicore” under “Technology Topics.”)

According to ExtremeTech, an Intel exec at the forum said the company “has no less than 15 dual-core development plans underway…Moreover, Intel intends to combine 64-bit processing and dual-core capabilities in its “Extreme Edition” processor family, offering the best of both worlds.”

New Scientist reports on an FPGA-based supercomputer being built at the Edinburgh Parallel Computing Centre. The designers say it will operate at one teraflop and will be up to 100 times more energy efficient than a conventional machine of the same power. Scotland’s FPGA High Performance Computing Alliance will “develop software tools to enable programmers to create code for FPGA chips more easily.” FHPCA is led by Nallatech.

(Hat tip to Slashdot.)

For those of you who stumbled here via Google and might not be in the know, an FPGAs are specialized chips that have transistors connected into the most basic functional blocks–logical ands, ors, and multipliers that are used to execute any computer code. Rather than being hard-wired in a predetermined configuration as they are in traditional processors, connections between the functional blocks can be set and reset in FPGAs.

Everyone has seen the IBM commercials that tout a computer’s ability to reconfigure itself depending on what it’s being asked to do. But such “on-the-fly” circuit changes aren’t just the vision of the corporate world. The academic sector is also a major partner in bringing such vision to reality by getting federal funding for just such research. In the latest edition of BusinessWeek Online, the article “Mighty Morphing Power Processors” points to the University of Texas and the Ohio Supercomputer Center as two institutions heavily involved. There is a lot at stake in this next generation, reconfigurable chip. The article states

IBM is hardly the only chipmaker chasing morphing semiconductors. Virtually every major supplier of so-called logic chips is working on some such notion, including Hewlett-Packard, Intel, NEC, Philips Electronics, and Texas Instruments. A dozen or more startups are in the race as well, including Velogix, picoChip Designs, and MathStar.

Need a supercomputer but can’t afford a Cray or time on IBM’s behemoth Blue/Gene? According to an article at New Scientist, you might have an option besides stringing together all of the neighborhood Dells. It seems a company in California is marketing a small supercomputer with a price tag of $100,000. Certainly not pocket change, but with a claimed peak power of 230 Gflops and about the size of a refrigerator, just slide the armoire out of the way. Sporting 96 1.2 GHz processors, this mini comes out at $435/Gflop (using simplistic math of course). Granted it wouldn’t make the last rung of the latest Top500 Supercomputer ladder and you wouldn’t use it to run a large, complex CFD simulation, but 230 Gflops is more than capable of crunching some solid numbers.

Such a machine could very well find its niche in smaller markets and bring computing power to those who need it but couldn’t really afford it. With IDC reporting that the high performance computing market grew roughly 23% from 2003 to 2004, which includes statistics of systems valued at less than $50K and more than $1M, the demand is clearly there.