Let's face it: Software development, for the most part, isn't all that it can be. Recently Adobe inadvertently underscored this problem when it posted an article on its site (using data taken from benchmark studies conducted by DMN's own Charlie White), showing how a single-processor 3.06 GHz Dell Precision Workstation dramatically outperformed a similarly equipped dual-processor 1.25 GHz Macintosh G4 at rendering compositions in After Effects 5.5. So what's the problem? Software that isn't written to take advantage of the hardware. That's a big problem.
The article in question, of course, is the now-infamous "PC Preferred" document that appears on Adobe's Web site at http://www.adobe.com/motion/pcpreferred.html. In it, Adobe uses benchmark results showing how After Effects 5.5 renders much, much more quickly on an Intel-based system than on a Motorola-based system. There's nothing particularly extraordinary in that, except for the inexplicably inflammatory anti-Macintosh tone (and nomenclature) of the Adobe article, especially given that Mac users make up 28 percent of Adobe's market and a whopping half of the total creative production market. What is extraordinary is how eager this article makes Adobe seem to advertise the shortcomings of its own software's performance.Before I go on, let me emphasize something. I'm not a Motorola apologist. I use a Mac because of the environment and workflow, and I don't care in the least what piece silicon is buried beneath a heat sink hidden inside my computer's case, as long as it gives me some decent performance for my money. It's irrelevant. What is relevant is how well software is written for any given chip, any given system architecture, any given OS.
Now, being familiar enough with Charlie's benchmark features, I have no doubt as to the accuracy of the data. Charlie--as much as I wish otherwise--conducts his tests evenly and bases them on common tasks professional users certainly do engage in. And, at times, I've double-checked his figures and even contributed my own benchmark tests for him to use because I didn't want to believe what I saw: that a Pentium 4-based machine could not only be faster than a Mac, but significantly faster ... with Adobe software.
But why is it faster? Yes, gigaHertz do mean something. But what happens when you look not at comparisons between Intel- and Motorola-based systems but between single- and dual-processor Mac systems? Well, take a look at the following chart showing relative performance between three applications running on single- and dual-processor Mac systems.
These tests were conducted using various built-in filter effects and image operations on a variety of Mac-based systems, and, as you can see, After Effects itself (Production Bundle, no less) is not exactly a model of multiprocessor efficiency. Where, on average, Final Cut Pro 3 rendered compositions in 58.25 percent the time on a dual-processor systems and Combustion 2 rendered at 63.73 percent the time, After Effects pulled off a mere 90.09 percent. (You can find more examples of these benchmarks here, here and here, along with more details of the elements involved in the tests.)
Now, let's say After Effects could handle dual processors on a Mac as well as Final Cut Pro--that is, that render times on a dual-processor system could be cut down to 58.25 percent of their current speeds. What would happen? In the tests Charlie White conducted and that Adobe cites in its article, the Mac would actually see slightly better render times than the Dell Precision Workstation 3.06 GHz P4 in some cases, while running neck and neck in others. But there certainly wouldn't be the disparity cited by Adobe and Charlie.
In other words, it isn't the processor or the operating system; it's the software.
1 2 Next