• Dual G4s @500Mhz (rev 2.9 CPUs)
• Firmware version 4.1.8
  (v4.1.9 was released on 9/25, available via OS 9's Software Updates)
• 512MB RAM (one CL2 OWC dimm)
• 20GB OEM Maxtor 5400 RPM drive (original drive w/OS X)
• 45GB IBM 75GXP 7200 RPM drive (slave w/OS 9.2.1)
• Toshiba SD-R1102 Combo DVD/CDRW drive
• Retail Radeon AGP (tests also done in Q3 w/OEM GeForce3)
• Quicktime 5.02 (Pro)

Although there is an almost unlimited range of tests possible, here's those I have been able to run so far:

• iMovie2 Export to QT Movie
• Photoshop 5.5 Performance (Classic)
• "Cold" Boot time comparisons
• Loading Apps
• Mounting Firewire drive
• IDE and Firewire Disk Performance
• Quake3 tests with Radeon and GeForce3

For comments on DVD playback, improved Sony FW900 monitor support, disc burning, etc. see my OS X 10.1 first impressions page. Overall performance seems faster, especially launching applications. The interface in general seems faster also and the Dock can now be positioned on the sides as well as the bottom of the screen. The follow are some measured tests I've run so far.

iMovie2 QuickTime Export Performance:
I used my standard test of taking the Tutorial clips stacked end to end and exporting with the default "CDROM Medium" settings. The graph shows the time in seconds (shorter bars are faster) for OS 9.2.1, OS X 10.0.4 and OS X 10.1:

Note: iMovie 2.1.1 for OS 9 was released last night. Running the tests 3 times and averaging the results showed there was literally no difference in performance over the previous version on this system.

Note - an owner of a Dual G4/800 sent his iMovie2 test results:

• OS 9.2.1: 89 Seconds
• OS X 10.1: 77 Seconds

Photoshop 5.5 Performance:
I ran the 21 filter PSbench test (10MB image size) in OS 9.2.1, OS X 10.0.4 and OS X 10.1 Classic modes. The difference in performance (total time for all 21 filters) was very close. This was to test Classic performance vs OS 9 native and compare the two OS X versions.

Note - an owner of a Dual G4/800 sent his results:

• OS 9.2.1: 56 seconds
• OS X 10.1: 60 seconds

As is standard for PSBench tests, screen mode was 1024x768/millions colors. 128MB of RAM was allocated to Photoshop (enough to ensure that the swap file was not used, although in classic I'm not positive how this is allocated).

General/Disk Tests:
Disks used for these tests were the original 20GB Maxtor OEM drive (75% free) and an OWC external oxford 911 bridge based AC powered case with IBM 60GB 75GXP drive that was 20% full and not optimized. (See the Firewire articles page for a past review of the Elite FW drive as well as a build guide.) There are faster drives than the Maxtor 20GB OEM drive I used of course, but this is just a comparison of how it performed with the two OS versions.
First some general tests like booting, starting classic, mounting a AC powered Firwire drive and loading IE 5 browser. (Browser times were from clicking the icon to both the main browser window and download manager window appearing. Due to the fact it's impossible to get repeatable web page loading from the net (server/connection loads vary constantly), I did not include time to load the selected web page.)

Note: There was literally no difference in the time to launch Photoshop in Classic between the two versions. (Photoshop 5.5/OS 9.2.1 was on a separate IBM 75GXP 45GB slave drive.)

I also timed how long it took to duplicate a single large file (457.2MB Q3 pak file) and a folder with 642.8MB of Files inside with the OEM 20GB Maxtor drive and the OWC Elite 60GB Firewire drive.

The big improvement is seen in OS X's apps loading, but these disk bound tests (isolating only the drive tested) didn't show any significant gains really. And again the OEM 20GB Maxtor is far from the fastest IDE drive available. The next graph shows tests with a AC powered, 60GB (7200 rpm) fast Firewire drive:

The Firewire drive tests showed measurable benefits from OS X 10.1.

Sherlock Searching: I also ran tests of how long it took to search both the OS X and OS 9.2.1 drives using Sherlock. OS X 10.1 was slower in this test, but since OS X 10.1 installed more files than were present with 10.0.4 I didn't consider this a fair test. Searches on the Firewire drive that had the same file count for both tests showed OS X 10.1 was faster (1.5 seconds vs 1.9 seconds).

Quake3 Performance:
I used the 1.29f-2 beta of Quake3 for all tests. Since the main purpose was to see if the base OS was more efficient (CPU/driver wise), to eliminate the graphics card fill rate, etc. tests were primarily run at 640x480 mode (except where noted). I used the standard "fastest", "normal" and "high quality" options in Quake3. (Making sure the resolution was 640x480 except as noted since fastest defaults to a lower res and high quality to a higher one).
In addition to the tests at fastest/normal/high quality, I also tested using Oliver's "240" custom config file (linked in a previous article here) with both sound off and on. All tests had the default "game options" page settings. The standard demo in Quake3 1.29 was run, all tests had SMP support enabled (r_smp 1). Results are in Frames Per Second (longer bars are faster).

As a FYI, at 1600x1200/HQ settings, the Radeon scored 35.1 FPS.

As a FYI, at 1600x1200/HQ settings, the GeForce3 in the DP/500 scored 58.3 FPS.

Note: Apple does not recommend (or warranty) the GeForce3 card in a 2xAGP mac such as my G4/500 DP. I have seen no problems using it so far however.

For comments on DVD playback, improved Sony FW900 monitor support (with the Radeon), disc burning, etc. see my OS X 10.1 first impressions page. I've also posted articles on Optimizing 10.1 with SpeedDisk and tests of Firewire drive mount/sleep issues with 10.1.
-Mike