Are you ready for Multi-Core technologies?
April 8th, 2008
If you are, then you may be the only one
“Today we are at an inflection point in the computing landscape as we enter the multi-core era.”
- Mark D. Hill and Michael R., Marty, Professors of Computer Science at University of Wisconsin in an article to be printed in 2008 by the IEEE
The inflection point that Professors Hill and Marty refer to is no small event: It’s a major technological leap into the future. Microchip processors from Intel and IBM and others are going to radically outstrip software capabilities with launch of multi-core processors beginning this year and those scheduled through 2010. Over the course of these next several years, dual-core and quad-core chips, both of which are available in the market today, will be joined by six-core and octo-core chips, all of which will give way to multi-core processors by 2010-2011 with as many as 80 cores plus.
These multi-core chips are leapfrogging today’s software capabilities, most of which were written for single-processing “pipe” chip architectures. Multi-core chips will be best utilized with parallel-programming, a programming approach preferred by supercomputer programmers for use in supercomputing server clusters.
To meet the challenge of essentially re-writing and re-thinking computer software to accommodate for the processing power inherent in parallel programming, Google and IBM teamed up last year to fund software development education and research at several U.S. colleges. (Named colleges so far are: University of Washington, Carnegie Mellon, MIT, Stanford, University of California at Berkeley and University of Maryland.)
In March of this year, Intel and Microsoft announced much of the same sort of approach. The Wintel brothers pledged $20 million to the University of California at Berkeley and University of Illinois at Urbana to have those two institutions focus curriculum and research on this growing software need. Intel is going alone on initiating several study and research programs on parallel software development at various colleges and universities in India.
What will parallel programs operating on multi-core and multi-threaded processors do? Just for starters, here are some realistic applications to expect:
- The Health Sciences field may be the biggest winner. IBM is working with University of Pennsylvania to load onto a supercomputing platform all mammogram scans they can get their hands on, all of which will have all personal identities stripped from the records. This way, a doctor who just took a mammogram of a patient could compare it to other mammograms scanned and stored from across the country, if not worldwide. Using recognition software and other analytical tools, the doctor could then see time progressions of the patient’s diagnosed condition to help determine what is likely to happen to the patient over time under various treatment scenarios;
- Far more accurate weather forecasting and greater pinpoint satellite imaging;
- More realistic vehicle crash tests;
- More accurate geophysical subsurface mapping.
These and many other capabilities will ripple through the economy as the inflection point becomes more apparent in applications. For the private sector, more affordable processing power could revolutionize product design. Eventually, the long-predicted future of Artificial Intelligence in computers and machines will also become more feasible with the step-function increase in processor capacity, if the software can catch up to the advances in chip-level processing power. Maybe the question is: Time to learn parallel processing software?