Application Launch Times
The following chart displays the time required for the first launch after logging in. My methodology was the same on both machines - delete the cache files, restart, and load each application once, timing how long it took from double-click until a usable workspace was presented to the user. If I were to show timings for second launches, all of the times for OS X would be substantially better (I don't think OS 9 second launches were any different than first launches, but I may be wrong).
I also think there was something wrong with my 10.2 install, given the launch time disparities from 10.1, but I didn't bother to do any troubleshooting of the problem (I tend to launch applications about once a week, so it doesn't really bother me as to how long it may take to load any given app). Hey, I warned you these were unofficial benchmarks and should be used with a large grain of salt!.
Update: Fixed a couple of incorrect highlights.

As you can see, the G5 has finally eclipsed the G4 in OS 9 as the application launch speed king. OS X haters will point to the bloat of the operating system and laugh -- "Ha! It takes a Dual G5 to do what an older G4 could do in OS 9!" While that's certainly true when it comes to launching applications, from my perspective there's a heck of a lot more to an OS than opening applications. As noted above, with the stability of OS X, I launch apps very rarely, and don't really consider launch time as a differentiating factor. Nonetheless, the speed with which OS X launches many of these most basic applications is quite impressive.
Perl Test
This Perl benchmark, which I found on the net, is a very simple little test that reflects the speed of the Perl interpreter, as well as the speed of the CPU driving the interpreter. Since there's no GUI to worry about, video speed has no impact on these results. Also, since OS 9 lacks a native Perl interpreter (see previous paragraph about "more to an OS than opening applications"), its results are quite slow. In cases where multiple OS versions had the same low time for running a certain portion of the test, I didn't bother to highlight the ties.
Update: Fixed a couple of incorrect highlights.
Once again, the raw power of the Dual G5s, coupled with the OS X improvements through 10.2, leads to some truly amazing speeds in this basic benchmark.
Real World Tests
The last set of benchmarks involves an independent collection of tasks that was designed to simulate things that I occasionally had to do with my machine. At the time of 10.0's release, the only graphics application available was Corel PhotoPaint, so it was used for the graphics portion of these results. Following the table is a key that explains what each test entailed.
Scroll Word doc: A 16-page text-and-table document with minimal graphics was scrolled by clicking and holding the mouse button on a scroll thumb. The time is the time required from start until the scroll thumb hit the end of the scroll bar.
Recalc Excel sheet: This test is based on a real sales analysis spreadsheet with thousands of cells. It's a complex worksheet containing detailed lookup and validation formulas that cross reference one another in many ways. Recalculation is set to manual, and the time starts with the press of the recalc key and ends when the "Recalculating..." message disappears.
Image load: Grp Cnv: Dropped a folder containing 63 images, totaling 2.9mb, onto Graphic Converter. Timed from release of mouse button until last image was fully displayed.
Image load: Preview: Dropped the same folder of 63 images onto Preview. Timed from release of mouse button until it was possible to move through the images in Preview.
5-Point mesh warp: Using Corel PhotoPaint, open an18.9mb, 16.7 million color, 3072x2048 pixel image, and apply a 5-point mesh warp to the image.
5-pix gaussian blur: The same image has a five-pixel Gaussian blur applied.
Glass block effect : The same image has a glass-block effect applied.
Once again, although most of the G4 OS X times matched or beat their OS 9 counterparts, the power of the G5 shines through in the final column -- this machine loves to crunch numbers and push pixels!
Conclusion
I'll be the first to admit these tests are far from scientific, and I ran them mainly for my own interest. Any one of the numbers on any of the charts could clearly be questioned, though it's harder to argue that there's not an overwhelming amount of evidence that suggests the Dual G5 is one very, very, fast computer.
Review home page | General
Overview | Photoshop Elements
Games | Rob's
Benchmarks | Other Benchmarks | Conclusion
[Feedback is welcome -- please use this
thread on our forum site to discuss this article.]