The following text is copyright 2003 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

A different Big Mac index

By Scott Bradner

Given my fondness for Apple computers you knew I would eventually write about the Virginia Tech. Mac-based supercomputer-on-the-cheap. I was waiting for some of the testing to be completed. Since the test numbers are beginning to come in I guess it's time. This is one heck of a system - not the sort of thing that you would want in your den but it looks quite useful for those of you who are into computational fluid dynamics.

For those of you who have not been following this story, Virginia Tech's Institute for Computational Science and Engineering (ICSE), already the home of more computing power than the average enterprise, decided to try to join the ranks of the supercomputer elite but they wanted to do so with a total budget smaller than the annual power bill for the existing big guys. After looking at various alternatives, including AMD and IBM based systems they had just about given up when Apple announced the new dual 2GHz G5 processor desktop last June. (http://www.apple.com/powermac/) A few days later, after a trip to Apple headquarters to convince Apple that they were serious, Virginia Tech ordered 1,100 systems via Apple's on-line store. Apple started delivering the machines starting September 5th and a small army of Virginia Tech students and teachers worked around the clock to install and interconnect the machines. A month later they were testing what may turn out to be one of the five most powerful supercomputers in the world. Virginia Tech spent $5.2 million on the Apple computers, Cisco gigabit Ethernet and Mellanox InfiBand switches and $2 million more on upgrading the building and installing a big Liebert UPS and a backup generator. This is rather more than I spent on my last Apple PowerBook but a pittance when compared to the $30 to $200 million that others have recently spent to join the same club. (See http://computing.vt.edu/research_computing/terascale/ for more information.)

Does this trick have any lasting meaning? I'm not sure. It does show that stock high-end Apple computers can be cost / performance competitive with the best that anyone else can offer and that the OSX operating system can be used for advanced clustering applications with minimal special software. I would expect that the next generation of Apple's X-Server machines would be even better for this sort of thing since they are much smaller and more can fit in a rack. Also, since Virginia Tech will be publishing all of the designs and (hopefully if they can deal with the patent issues) all of the software they used I would expect to see a bunch of these Mac clusters popping up in research institutions and Fortune 100 companies soon. Pundits have been calling the Virginia Tech machine "Big Mac" but this Big Mac has special but not secret sauce. It will be good fodder for Apple's and, since Apple uses IBM processor chips, IBM's marketing departments.

Economists use a "Big Mac Index" (http://www.econedlink.org/lessons/index.cfm?lesson=EM156) to compare prices in different countries. This Apple-based Big Mac may do the same for those comparing the costs of high-end computing. If so, it will not be good news for those companies building traditional supercomputers.

disclaimer: Being from Harvard I would hasten to add that low cost is not the whole story but the above is my own observation, not the university's.