The unveiling of the new Pittsburgh Supercomputing Center Cray XT3, called Big Ben, represents another step in the goal of increasing the rate of educational capacity and emerging technologies for computational science especially since its another resource added to the TeraGrid. Sporting 2090 processors, Big Ben is somewhat smaller than its brethren at ORNL. However the ORNL machine proves the XT3 has substantial scalability, having gone from 3700 processors to over 5000 in just a few weeks. CTWatch caught up with HPC Challenge benchmark suite co-developer Dr. Piotr Luszczek and asked him about the new Cray. While discussing differences between the higher end supercomputers and clusters, Luszczek pointed out Cray’s use of AMD processors with Hypertransport technology as an advantage over other machines that rely on some form of PCI and the resulting bottlenecks in processor data throughput. He further stated that

Big Ben’s bandwidth performance is an example of good use of existing technology by Cray since it’s a system that could not be easily replicated with other processors.

More detailed information about Big Ben and PSC can be found here.

This entry was posted on Wednesday, August 3rd, 2005 at 12:11 pm and is filed under Hardware. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.