Washington Apple Pi

A Community of Apple iPad, iPhone and Mac Users

iMac Benchmarks: Quick Comparisons

© 1998 Washington Apple Pi Labs

Related articles: iMac Memory, iMac Review

Benchmarks

Benchmarks are wishy-washy things. When you are comparing apples to oranges, do you compare color? Texture? Vitamins? Taste? Ballistic characteristics when you throw them at bad mimes? Do you use the same standards all the time, even when the standards become outdated or inappropriate?

The most widely quoted benchmark in the Macintosh world is MacBench 5.0, by Ziff-Davis Publishing Co. MacBench 5.0 is more-or-less OK, but Washington Apple Pi Labs prefers our own Mac Bench (see "Introducing: Washington Apple Pi Mac Bench," Washington Apple Pi Journal, January/February 1998, pp. 25-26). Among many other flaws, Ziff-Davis keeps "renormalizing" the benchmark with every release, making it impossible to do longitudinal comparisons. In Version 5.0, for example, the Power Macintosh G3/300 is given a score of 100 as the "standard," making it very difficult to make comparisons with older machines.

In short, we didn't use MacBench 5.0. Instead, we used Symantec's System Info 4.0, part of Norton Utilities for Macintosh 4.0. System Info 4.0 has several things to recommend it, but the biggest one is that it covers older machines. It does have a few quirks: it suggests the disk cache should be set to 128K, that AppleTalk should be off, and that the video should be set to 256 colors. Since these are not ideal for an iMac, we dubbed these "Norton settings." We then created another set of settings with the disk cache set to Default (by pressing the Default button in the Memory Control Panel), AppleTalk to On, and video to Millions of colors. We also decided to retest the iMac after adding 64 megabytes of memory (for a total of 96 megabytes) and installing Mac OS 8.5 (instead of the "fresh out of the box" combination of 32 megabytes of memory and Mac OS 8.1). We then ran several benchmarks, under differing conditions:

Although we had rejected it earlier, we were curious, and decided to try MacBench 5.0 as well, and got these results:.

From this, we drew a few conclusions:

• As the MacBench 5.0 disk test shows, it really does pay to have more memory, and use the Default setting in the Memory Control Panel to set an optimum disk cache;
• An iMac may or may not be three-quarters as fast as a Power Macintosh G3/300. Despite our best efforts, neither Apple nor any of our so-called "friends" would loan us one for testing. However, three-quarters as fast as a G3/300 is still very, very fast;
• The System Info 4.0 benchmarking demonstrates that, yes, you can get higher benchmark scores by turning off networking and setting an iMac screen to 256 colors. However, if you didn't want networking and you were satisfied with 256 colors, you could have purchased a Pentium;
• Yes, an iMac really is between 500 and 600 times faster than a Mac Plus;
• Yes, it really is five to six times faster than a Power Mac 6100, too.

Subjective Benchmarks

Subjectively speaking, we also, unanimously, concluded that we'd rather be working on one iMac than 500 to 600 Mac Plus computers. In theory, 500 to 600 people working at Mac Plus computers might actually get more work done than one person working on an iMac, especially since the Mac Plus didn't come with Nanosaur. We have no desire to test this theory.

There are, of course, other things an iMac can do that a Mac Plus cannot, such as run Gerbils! Gerbils!, created by the same people who wrote Nanosaur, was designed as a demonstration of what can be done with Apple's QuickDraw 3D technology. Unlike Nanosaur, Gerbils! really isn't a game; no matter what you do, you can't win or lose. Instead, you watch oddly-shaped "gerbils" run around on a track and occasionally bump into one another while music plays in the background. You can vary the track texture, the shape of the "gerbil," and radically alter the track layout.

You can also run, if you have enough memory, multiple copies of Gerbils! This is far, far more fun than running any "normal" benchmarking utility so, after installing a total of 96 megabytes of memory and Mac OS 8.5, we installed Gerbils! on the iMac. Then duplicated the Gerbils! folder. Several times.

We discovered that you could run up to 12 copies of Gerbils! simultaneously, each with its own customized, textured track. This serves as a good test of the multi-tasking abilities of Mac OS, and nicely demonstrates the speed and power of the iMac's video circuitry. But mostly it is funny: when a gerbil is bumped off a track, it says "Bye-bye." When twelve tracks are going at once, it sounds like a preschool when the kids are leaving for home.

Economic Benchmarks

"The news media" has expressed interest in who is buying the iMac. The answer is: almost everyone. There have been many Mac loyalists, finally moving from their Mac Plus machines to the PowerPC age. There have been government agencies (NASA's headquarters purchased 600), die-hard Windows users (some of whom claim they are "just testing it out"), and game players (the iMac is a great game machine).

For Mac loyalists with older machines, what kind of value does the iMac offer?

Compared to a Mac Plus of twelve years ago, an iMac is almost half the cost. It comes with 32 times as much memory, 5,000 times as much storage, and (as noted earlier) is between 500 and 600 times faster. If you set the iMac screen to 800 x 600 pixels (the most common setting), the built-in display shows almost three times as much information, and in millions of colors instead of black and white.

It is entirely possible that none of the software you ran on your Mac Plus will run on an iMac, and it won't talk to your old ImageWriter dot matrix printer, either. This is a blessing. Trust us on this.

While we would still like to do some comparison tests against a Power Mac G3/300 or G3/333, all our tests do show one thing: this is a fast, powerful machine, and an excellent value. Moreover, as two members of the Washington Apple Pi Labs team insists, "it's cute."

Revised December 13, 1998 Lawrence I. Charters
Washington Apple Pi
URL: http://www.wap.org/journal/