Multicore madness

The new issue of American Scientist is now available both on the web and on paper. The subtitle of my “Computing Science” column makes the following rash assertion:

Multicore chips could bring about the biggest change in computing since the microprocessor

It’s always wise to be a little skeptical of such superlative claims. In Redmond, Washington, last week, I asked a foreman at a construction site what his big hole in the ground was going to be. He told me that a local software company was building the world’s largest underground parking garage. I have no particular reason to doubt that statement, but I would have been more firmly convinced if he had been able to answer my followup question about where I could find the second-largest underground parking garage.

The assertion in my subtitle doesn’t stand up particularly well to this kind of scrutiny. I really don’t want to be asked about the second-biggest change in computing in the past 30 years. But I hope my readers will forgive or overlook my moment of impetuous hyperbole and accept the broader point that the shift to dual-core and quad-core (and eventually many-core) processors really is going to make a difference in how computers work. We’ve seen decades of research on parallel processing and concurrency, but through it all the mainstream of computing has remained steadfastly single-minded. It looks like that’s finally going to change.

This entry was posted in computing.

2 Responses to Multicore madness

  1. randomwalker says:

    I have only once in recent memory asked a construction worker what he was building. The answer? the world’s largest hotel. Maybe all these construction guys are trained to say they’re building the world’s largest something-or-other :)

    Your subtitle is definitely hyperbole. But meh, hyperbole makes good copy. We all do it.

    I don’t see multi-core as a big deal myself. Most end-users already have way more CPU cycles than they need. Video playback takes up what, 5% of CPU?

    The real success story of parallelism is mapreduce and other server-farm techniques, which you mention; and those have nothing to do with multi-core.

  2. realmode says:

    I agree with randomwalker. Distributed computing will be a big deal. Multi-core, not so much.