Mac OS X 10.2.3--How Fast Is It?

The most frequent user gripe about Mac OS X when it was first released as version 10.0 nearly two years ago, aside from the lack of applications, was it's lack of speed, particularly in the Finder. We also noted that it took a very long time to boot up, about two minutes, as compared with other operating systems.

In response, we began benchmarking each version's speed starting with 10.0 vs 10.1 on April 17, 2001. This has been a little difficult because of the initial lack of applications and benchmarking tools and because those that became available were frequently updated. Our choice of benchmarks, though limited by availability, have been those generally available to anyone and useable by the general user. We think they do a fairly good job of reflecting any significant changes in speed among versions of OS X.

Historically, there were major speed improvements of some type in versions 10.0.3, 10.1, 10.2 and 10.2.1. Version 10.1 brought fast application launches in one to five seconds which have remained virtually unchanged through 10.2.3 so we excluded all but one (Classic Mode) from our benchmarks from 10.2.1 onwards. Version 10.2 significantly improved startup time and the launch time for Classic Mode. Version 10.2.1 significantly improved the the login time and the shutdown time which had inexplicably increased a lot in 10.2.

The computer used in the benchmarks has been the same, a 450 MHz G4 Cube, through version 10.2.1. It now has 800 MHz G4 CPU and 80 GB Seagate Barracuda drive upgrades installed so we were forced to use our other 450 MHz Cube which is our network server. Every "identical" computer gets slightly different scores and times on benchmarks due to slight differences in the performance of its components. Even the same computer will have minor differences when run at different times.

We re-ran 10.2.1 on the new Cube and noted that there are only minor differences with the previous Cube. Then we ran 10.2.2 and 10.2.3 with the stock ATI Rage 128 Pro and the Nvidia GeForce2 MX graphics cards. The latter was pulled from a Power Mac G4 tower. Our Cube has 708 MB of RAM but is otherwise stock. Here are the results:

Mac OS X Actions on a 450 MHz G4 Cube*
(Average Times/Scores)

Action	10.0.4	10.1	10.1.5	10.2	10.2.1	10.2.2	10.2.3
.
Chime to Login (sec)	96	68	79	56	50	50	50
Login to Desktop (sec)	20	9	19	20	11	11	12
Shutdown (sec)	-	4	5	20	10	9	9
Classic Mode (sec)	40	46	43	33	27	27	27
Scroll in Acrobat 5 (sec)	69	82	71	-	44	45	44
Scroll in AppleWorks (sec)	-	-	15	11	11	11	11
Let1kWindowsBloom (sec)	-	61	57	56	54	56	54
CocoaBench Image Score	-	-	-	11	11	11	11
Speed Run Graphics Score	-	-	-	-	282	288	291
Xbench Quartz Score	59	58	58
Xbench Open GL Score	84	82	83
Xbench User Int. Score	61	61	61
Quake III Arena (fps)	21	20	21

As we noted in our speed test of 10.2.1, it is significantly faster than 10.2 from login to the complete desktop and in shutting down. Classic Mode also launches faster. The speed of 10.2.1, 10.2.2 and 10.2.3 is virtually identical in all of the benchmarks. Only the Speed Run v1.1.2 graphics score shows a modest improvement from 10.2.1 through 10.2.3.

We also ran all of our speed benchmarks with an Nvidia Geforce2 MX graphics card pulled from a Power Mac G4 tower in place of the stock ATI Rage 128 Pro. The GeForce2 has 32 MB of video RAM (VRAM) while the Rage has only 16 MB.

Jaguar (v10.2) implements Quartz Extreme which offloads some graphics operations to the graphics card. Apple originally indicated that 32 MB of VRAM was needed to run Quartz Extreme in Mac OS X 10.2 and later, so many people with early G4s like our Cube replaced their stock Rage card with a GeForce2 MX or GeForce3 for $150-$300. But now, Apple states the minimum requirements include only 16 MB of VRAM.

"Quartz Extreme functionality is supported by the following video GPUs: NVIDIA GeForce2 MX, GeForce3, GeForce4 MX, or GeForce4 Ti or any AGP-based ATI RADEON GPU. A minimum of 16MB VRAM is required."

Here are the results of the two graphics cards:

Mac OS X Actions on a 450 MHz G4 Cube with ATI Rage 128 Pro and Nvidia GeForce2 MX Graphics (Average Times/Scores)*

Action	10.2.1		10.2.2		10.2.3
Graphics Card	Rage	GF2	Rage	GF2	Rage	GF2
Chime to Login (sec)	50	47	50	48	50	47
Scroll in Acrobat 5 (sec)	44	47	45	48	44	46
Scroll in AppleWorks (sec)	11	13	11	13	11	12
Let1kWindowsBloom (sec)	54	55	56	56	54	55
Speed Run Graphics Score	282	236	288	237	291	242
Xbench Quartz Score	59	56	58	56	58	56
Xbench Open GL Score	84	72	82	73	83	72
Xbench User Int. Score	61	59	61	59	61	60
Quake III Arena (fps)	21	39	20	40	21	40

Four of the benchmarks yielded identical results and are not tabled above. Except for the time from the startup chime to the login window and for the 3D game Quake III Area (1024x768), the stock ATI Rage 128 Pro gave faster results than the Nvidia GeForce2 MX or equalled it.

As we reported previously, there appears to be little reason to replace the ATI Rage 128 Pro with a GeForce2 MX card unless one plays 3D games at resolutions of 1024x768 or higher. But what about Quartz Extreme? While the results above appear to be disappointing, Quartz Extreme really only accelerates window compositing, e.g. moving or resizing windows, according to Apple. We have no test specifically for that but an impressive demonstration of Quartz Extreme's acceleration is shown in the Seybold SF 2002 keynote by Phil Schiller, Apple's VP for Worldwide Product Marketing at minutes 12-22.

In summary, version 10.2.3 does not appear to improve the speed of OSX, at least for our test computer.