Virginia Tech racing to deploy Power Mac G5 supercomputer to make Top 500 list

“Virginia Polytechnic and State University is racing to deploy its new supercomputer cluster based on Apple Computer Inc.’s G5 desktops by Oct. 1 in order to make the next ranking of the Top 500 list,” reports Tom Krazit for IDG News Service. “Faculty and students hope to set up the 1,100-node Terascale Computing Facility by the end of this week, said Lynn Nystrom, a Virginia Tech spokeswoman. The university thinks the new cluster will rank among the top 10 fastest supercomputers deployed at academic institutions, and among the fastest overall in the world, she said.”

“Virginia Tech announced the cluster earlier this month. It will cost the university $5.2 million spread out over a five-year period, Nystrom said. Each node uses two 2.0GHz PowerPC 970 processors made by IBM Corp. and comes with 4G bytes of memory and 160G bytes of storage. The PowerPC 970 chip, like Advanced Micro Devices Inc.’s Opteron chip., is a 64-bit processor that can also run 32-bit applications. Virginia Tech chose the Apple systems for its cluster because of the price/performance benefits of a cluster based on the G5 desktop, Nystrom said,” Krazit reports.

“The university will present its benchmarking information to the Top 500 rankings by the Oct. 1 deadline, Nystrom said. The final scores will be released some time in November, she said,’ Krazit reports. Full article here.

Related MacDailyNews article:
Virginia Tech Power Mac G5 Supercomputer Cluster photos posted online – September 21, 2003

7 Comments

  1. “The PowerPC 970 chip, like Advanced Micro Devices Inc.’s Opteron chip., is a 64-bit processor that can also run 32-bit applications.”

    Wrong. The AMD chip does not run 64-bit apps and 32-bit apps simultaneously. In that sense, the PowerPC chip is NOT like the AMD chip. It is much better.

    On another note, though, I heard this cluster could rank as high as the top 5 of all supercomputers. Is that not true?

  2. A-Bomb,
    Sorry, but she does not say anything about running them simultaneously. Maybe she should be more specific, but she is not wrong.

    I believe someone did the theoretical peak computation in Slashdot and said it could get as high as no. 3. But then again, real performance does not match theoretical performance in virtually all cases.

  3. I don’t know how the Top 500 works, but software to best use this cluster won’t appear right away. Whatever rank it gets is likely to be lower than the performance it ends up delivering later.

Reader Feedback

This site uses Akismet to reduce spam. Learn how your comment data is processed.