[Buff]

Originally posted by Serge_Oprano@Nov 7 2004, 12:49 PM
Mike,
time for you to start travel the world BEYOND 5 stars hotels in 3rd world countries...
we already LOST the technological edge...
Cell phones with camera hitting the shelves in USA this year were in Europe/Japan for 2-3 years

Need more examples?

Didn't Take Long: IBM Regains Supercomputing Crown

Fri Nov 5, 3:13 PM ET

Jason Lopez, www.newsfactor.com

IBM's (NYSE: IBM - news) Blue Gene/L, which is being installed at the Lawrence Livermore National Laboratory, recaptured the supercomputer speed record that Silicon Graphics' (NYSE: SGI - news) Columbia system at NASA (news - web sites) held for barely a week.

The Department of Energy (news - web sites)'s National Nuclear Security Administration says Blue Gene/L hit 70.72 teraflops, trumping the 60.1 teraflops of data crunched by the SGI machine.

Only a quarter of Blue Gene/L has been set up. Kim Cupps, the project leader for the Blue Gene/L installation at Livermore, says the machine will operate at 280 teraflops -- or 280 trillion floating point operations per second -- when it is fully installed.

"Codes that used to take 30 days to run can now take 3 days on just a quarter of the machine," she told NewsFactor.

It's About Science

The purpose for Blue Gene/L is to speed up research. "We really can solve very large problem sizes much more quickly and it advances our ability to look at results that much more quickly," Cupps added.

Unlike the user experience on a PC, the researchers relying on a supercomputer's performance immediately feel the upgrade. PC's are not often pushed to the limits of their capabilities. Supercomputers are, which is what makes the installation of Blue Gene/L, or any scientific supercomputer, cause for celebration.

Though Blue Gene/L can be partitioned to run multiple problems at once, it will probably be used for larger ones. "You can run the smaller problems on existing machines," Cupps noted.

Plenty Fast

Supercomputers do not all at once become obsolete. They can be used, essentially, for many years. But like PCs, it becomes cheaper to simply scrap old machines rather than maintain them.

"That's a pretty strong word -- obsolete," Cupps said. "The typical replacement for these computers is about three years. But that does not mean they don't continue to be useful."

Three years from now, Blue Gene/L will probably be in service. But the foundations of its design will most likely be set aside. "The architectures used in our Blue Gene/L won't scale to the petaflop," she commented. New designs and ideas for power, cooling and floor space will be necessary to create a computer that operates above a thousand teraflops.

"We'll need disruptive technology to get us there," Cupps acknowledged. For now, Blue Gene/L is plenty fast.