• Simon Portegies Zwart rests his right hand on a prototype of the GRAPE-6, a computer he'll use to simulate star clusters (seen on monitor) and galactic nuclei.

    Simon Portegies Zwart rests his right hand on a prototype of the GRAPE-6, a computer he'll use to simulate star clusters (seen on monitor) and galactic nuclei.

    Photo / Donna Coveney

    Full Screen

Space researcher is using prototype of the world's fastest computer

Simon Portegies Zwart rests his right hand on a prototype of the GRAPE-6, a computer he'll use to simulate star clusters (seen on monitor) and galactic nuclei.


A Hubble Fellow in MIT's Center for Space Research (CSR) has received the small prototype of the fastest computer on the planet. The "baby" GRAPE-6 has a peak speed of 21.6 gigaflops, which is more than 100 times faster than the sustained speed of a workstation. The special-purpose machine, about the size of a shoebox, calculates the gravitational attractions between stars.

In collaboration with Professor Jun Makino of Tokyo University, a GRAPE-6 system (short for GRAvity PipE; see the web site for more information) was given to Simon Portegies Zwart at the CSR. He received the computer as part of a larger collaboration including astronomers Piet Hut at the Institute for Advanced Study in Princeton, NJ and Steven McMillan at Drexel University in Philadelphia.

The GRAPE-6 is a dedicated parallel computer that was recently awarded the Gordon Bell prize for performance. The calculation it performed was a simulation of a triple black hole system embedded in the center of a galaxy simulated by 768,000 stars. The sustained performance achieved was 1.349 teraflops on a prototype 8-board GRAPE-6 system. The full GRAPE-6 configuration is expected to become available in fall 2001, with a peak speed of 100 teraflops.

Dr. Portegies Zwart will use the computer for simulating star clusters and galactic nuclei. His calculations combine the interplay between the nuclear evolution of individual stars and binaries with the dynamical evolution of the star cluster using the Starlab software environment. For more information, go to the web site.

A version of this article appeared in MIT Tech Talk on June 6, 2001.


Topics: Space, astronomy and planetary science

Back to the top