Once upon a time, chip makers made computer chips faster every year by increasing their processing speeds. But lately, the microprocessor industry has run into some fundamental limits to those speeds. So chip makers have had to switch directions and look elsewhere for better performance. The latest solution: Design chips with multiple processor cores. The result: Today’s big-brained chips that can do more processing than ever before— if the software is modified to take advantage of their design.
Beyond clock speed
Only a year ago, the processors that ran most Macs had a single core, the brain that does most of the computing. Only the CPUs in high-end Power Mac systems intended for the power-hungriest professionals sported more than one core. But today, every Mac model sold contains at least two processor cores—essentially, two independent brains jammed together on a single chip. Mac Pros contain two dual-core chips, for a total of four processor cores. And Intel recently announced a chip with four processor cores.
Traditional single-core computers worked straightforwardly. All the work your Mac needed to do was put in the pipeline, and the Mac’s processor would work on each item in turn, as fast as it could. The faster the chip, the faster the computer could work.
But when chip speeds started stalling, chip makers such as IBM, Intel, and AMD turned to a different speed-boosting trick: adding processor cores.
Is your software savvy?
By default, Mac OS X assigns every task it’s given to the processing core with the most power to spare. Also by default, every program you run is a single, self-contained task. So if you had a (purely hypothetical) 16-core Mac and were running 16 programs, each would essentially have one core all to itself.
Ilustration by Oliver Wolfson
But what if you had one program that required huge amounts of processing power? If OS X saw it as a single task, all the operating system would do is assign it to one core—so the program would be able to use only one-sixteenth of the Mac’s total processing power.
That’s unacceptable, of course. So most high-end programs (including Adobe Photoshop and Apple’s Final Cut Pro) have been written to split the work they do into smaller tasks, each of which can be farmed out to a different processor core.
Now that multicore Macs are nearly ubiquitous, programmers of all kinds of applications will need to modify their software to support multiple cores, breaking up functions into small sets of tasks that can be more easily farmed out to the next available processor core.
In addition, software that’s already been modified to work with multicore systems will need to become more efficient. In our tests of the first Intel iMacs, we found that some tasks, such as the 3-D Ren-der, iTunes MP3 Encode, and iPhoto Import, used both cores efficiently. Other tests revealed that certain iMovie filters were not at all efficient. Photoshop itself was scattered: for some tasks, it was extremely well optimized for multiple cores; for others, it wasn’t remotely multicore-savvy.
In any event, the future of computer performance no longer seems to be solely in the hands of the chip makers. Power has shifted, at least somewhat, to the software developers. Chip makers have found a way to keep increasing the amount of power inside a computer’s chassis, but programmers will have to find ways to take advantage of it.
What’s a Mac user to do? In one sense, there isn’t much you can do. The only kind of new Mac you can buy today is one with multiple processor cores. But in another, you should make your voice heard. Ask the companies that make your most-important software programs if their apps are taking full advantage of your multicore Mac. You can even download Apple’s CHUD tools and test your own Mac with one processor core disabled to see if your favorite programs are really taking advantage of all that Intel processing power you paid for. In the meantime, Intel and Apple will continue to release chips and computers with an increasing number of processor cores. So when Intel releases that four-core chip, and when Apple puts two of them into a Mac Pro (the Octopus?), let’s just hope that we’ll have plenty of Mac software that can tap into their power.
[ Jason Snell is Macworld ’s editorial director. ]