New Benchmark for Fastest Supercomputer Share your comment!

The way the power and speed of supercomputers are measured is about to change.

A new benchmark will be introduced in November to rank the world’s fastest supercomputers, and will eventually replace the Linpack metric that is currently the basis for the Top500 list.  Called the High Performance Conjugate Gradient (HPCG), the new benchmark is being introduced by Jack Dongarra, distinguished professor of computer science at the University of Tennessee, and the developer of the Linpack system.

Dongarra said the Linpack metric is out of date and expects the new HPCG benchmark will provide a better measure of the performance that can be expected for the types of real-world applications that new supercomputers are running.

“Linpack measures the speed and efficiency of linear equation calculations,” Dongarra said in a recent statement. “Over time, applications requiring more complex computations have become more common. These calculations require high bandwidth and low latency, and access data at irregular patterns. Linpack is unable to measure these more complex calculations.”

The Linpack ranking was developed in the 1970s and has been used by the Top500 list since 1993. The Top500 list is updated each June and November.

There will probably only be a few supercomputers ranked under the HPCG this coming November, Dongarra said. It will take a few years to convert the current Top500 list to the new system.

Posted on by David Weldon, Contributing Editor, Slashdot Media
3 comments
Dc Z
Dc Z

Does the Operating System affect the speed & performance of a Supercomputer?