A Click on this Banner shows your site support to my Sponsors

Accelerate Your Mac! logo
The Source for Mac Performance News and Reviews
Review: PowerBook Firewire G3/500
By Mike
Review date:3/22/2000
(Updated Quake2 scores 3/30/2000)
Real World Applications and Game Performance
Intro | Apps/Game Performance | Benchmarks | Compatibility | DVD/Movies | Specs | Summary
How Does the Powerbook G3/500 Perform in the Real World?
This page lists test results in common Mac applications like Bryce 2, Infini-D 4.01, After Effects 3.1, Premiere 4.2.1, Photoshop 5.5 and popular 3D games like Quake1, Quake2, Quake3 and Unreal Tournament. I've included results from my PowerBook G3/250 (1998 Wallstreet 1 previously reviewed), an iMac DV SE (previously reviewed), B&W G3/400 and G4/450 AGP (Sawtooth) system for comparisons.

Infini-D Tests

I use the Infini-D 4.0 tutorial file 'Chapter 7 completed' as a test of performance. Rendering quality was set to Ray Trace, medium anti-aliasing, shadows on, patch detail low. I leave the output movie file at the default settings. Infini-D was allocated 40MB of RAM for the tests. The graph below shows the total time to complete the 150 frame rendering (3D animation to appx. 10MB Quicktime movie). Shorter bars indicate faster performance.

Inifini-D results

Bryce 2 Tests

Tests in Bryce 2 were run using the sample scenes in the KAI folder called "Magical Knight I" and "Alexandria II". Bryce 2 was left at the default memory allocation of 16980k, antialiasing was on, screen resolution was set to 1024x768, thousands colors. Shorter bars indicate faster performance.

Bryce 2 Results

Bryce performance (like all rendering apps) is primarily CPU speed dependent and won't show any rendering speed benefit from faster video cards for instance (a common misconception).

After Effects 3.1 Tests

The following is a graph showing the time to render a short special effects movie (10MB file size) in Adobe's After Effects 3.1. Shorter bars indicate faster performance.

After Effects Render Test

Premiere 4.2.1 Tests

The following is a graph showing the time to create the 'sample project' in Adobe Premiere 4.2.1 (set to full duration of the project, 320x240, 22KHz/16-bit stereo, 15FPS keyframe every 5 frames). Final movie size is appx. 8MB. Shorter bars indicate faster performance.

Premiere 4.2.1 results

Photoshop 5.5 Tests:
I also ran a complete PS5Bench (21 filter test) series using Photoshop v5.5. PSBench settings are 1024x768, millions colors, VM off, Interpolation set to bicubic (better) and Photoshop should be allocated enough RAM to avoid any swap file activity from the 10MB test image filter actions. (140MB was allocated to Photoshop 5.5 for all the other systems except the Pismo, which had 90MB allocated with VM ON in the default 128MB as-shipped configuration.) Photoshop 5.5's History' settings from the default 20 to 1 and unchecked the 'automatically create snapshot' option<.

The G4/450 AGP (Sawtooth) test used the current (as of this review date) OS 9 Altivec extensions (4) as well as the Adobe current release of the Altivec Core Plug-in and Lighting Effects Filter (available here). Only a few of the 21 filters in the test are accelerated by Altivec however (Lighting Effects benefits most, about 4x faster). Since the filters that took longest to run don't benefit from Altivec currently, the total test time for this 21 filter test is similar to a G3 of the same clock speed.

[I've included results with an AMD Athlon 600MHz CPU based system running Windows 98. Unlike the Pentium III and G4, there is no support in Photoshop 5.5 for AMD's Altivec/PIII-like enhanced instruction set (called 3DNow!), but it still does well in many filter tests using brute force alone (triple superscalar FPU engine, highly pipelined design and 600MHz clock speed).]

Photoshop 5.5 test results

With larger image sizes than the 10MB test file, or if the 21 filter tests used primarily Altivec enhanced ones of course the G4 would show a much larger benefit. Without Altivec support, a G3/500 beats a G4/450 at most any CPU bound application. (The stronger FPU of the G4 is primarily for double precision math.) For more info or to download the PSBench action file, see PSBench Home Page.

Game Performance Tests and Image Quality:

For 3D gaming, I'd suggest 192MB or more of installed RAM if you can afford it. Even with 128MB installed, Virtual Memory will still be needed for many of the latest games (some have 90MB or more preferred memory allocations). Remember the game, OS and OpenGL all need RAM. The PB G3/500 comes with 128MB installed, but I added another 128MB PC100 SODIMM from Other World Computing for a total of 256MB and disabled virtual memory.

In some cases the 8MB (fixed) video ram was a handicap (no 32bit mode 1024x768 gaming at least at the default settings), but considering the Rage128 chip really doesn't have the fill rate for 32bit mode at that resolution it's not a real issue. Frankly I was glad to finally see a notebook with enough graphics horsepower to run Quake3 and Unreal Tournament in 32Bit mode at 640x480.

For those that are not aware, the AGP Rage128 Mobility chip in the new Powerbooks has a provision for an 8MB Vram upgrade (which also increases the memory bus width to 128Bit), but there are no signs the video RAM is upgradeable in the PB. Like the VR128 chip in the new iMac, the Rage128 Mobility is a 64-bit design running at a lower clock speed than the latest Rage128Pro models found in the G4/AGP systems. It is however based on the Rage128 Pro core, but runs at 105/105MHz (Chip/Memory clock), at least the reference design does. Most owners of the new Powerbook will be more than pleased with the overall performance, especially compared to previous Powerbooks.

How I tested: I ran my usual array of game demo framerate tests with game settings the same as I use for all my video card and CPU upgrade reviews. In the case of Unreal Tournament tests on the PowerBook G3/500, I used OpenGL mode and medium skin/world detail, unlike the previous UT tests (Rage128 vs Rage128 Pro) which used RAVE mode and high detail settings. Consider that factor when evaluating the performance scores. (To date I have not retested the other systems in OpenGL/Medium detail modes.)

Game Performance Results: All scores in the tables below are in frames per second (higher is better). For information on how to test framerates in these games, see the FAQ's Game topic area.

Quake1 RAVE v1.09
Shadows, Flames, Filtering ON
(iMac had VM ON 128MB Ram, others 256MB/VM OFF)
G4/450 scores in ( ) are for Rage128 Pro card
PB G3/500
iMac DV
B&W G3/400
Rage128 r2
Rage128 AGP
64.0 (83.7)
27.9 (37.3)

(Results for my Wallstreet 1 are not shown, since with ATI v4.2 drivers/OS 8.6
it recorded only 5.3 FPS at 640x480 with all options enabled in Rave Quake 1.09.)

Quake2 (16Bit mode)
(iMac had VM ON 128MB Ram, others 256MB/VM OFF)
G4/450 scores in ( ) are for Rage128 Pro card
[Scores in red are PB G/500 with cache @250MHz & Tweaked Config file]
PB G3/500
iMac DV
B&W G3/400
Rage128 r2
Rage128 AGP

PB G3/500 red scores used an edited config.cfg file with the following variable changes:

  • set gl_force16bit "1"
  • set gl_swapinterval "0"
  • set gl_macmultitexture "0"
  • set gl_bigmem "1".

Quake2 was allocated 100MB of RAM (which didn't help FPS before the config file changes). Not sure which made the biggest difference but the increase was dramatic! (Perhaps swapinterval & disabling multitexture but I ran these tests at 3AM and am too tired to retest to isolate which setting helped most). [Note - I think the force16bit setting is the real factor.] The tweaks should help other systems, but I've not had time to retest them other than the G4/450 Rage128 Pro -- no change at 640x480, but 1024x768 ran 40 FPS with the edits.

Quake3 Arena Tests

Demo 1 framerates in the retail version of Quake3 Arena which had the first beta point release patch applied. Dynamic lighting and lightmap settings were used (the default settings for Quake 3 - only resolution and color depth was changed). All results are in frames-per-second, higher is better.

Quake3 Arena 16Bit 640x480 800x600 1024x768
PB G3/500 35.2 24.8 18.1
PB G3/333 15.1 (no data) (no data)
iMac DV SE 25.0 19.8 14.0
B&W G3/400 32.6 27.8 20.1
G4/450 Rage128 AGP 36.40 (not tested) 20.40
G4/450 Rage128 PRO AGP 38.90 (not tested) 25.60

The same tests were run in 32Bit mode.
(PBs and iMac can't run 32Bit/1024x768 due to only 8MB vram)

Quake3 Arena 32Bit 640x480 800x600 1024x768
PB G3/500 24.0 14.7 N/A
PB G3/333 (no data) (no data) N/A
iMac DV SE 19.4 12.5 N/A
B&W G3/400 30.3 22.9 15.4
G4/450 Rage128 AGP 32.50 (not tested) 15.30
G4/450 Rage128 PRO AGP 36.80 (not tested) 20.30

Cursor movement in 32Bit mode felt sluggish (even in the main game menus),
however I did not tweak with the control settings/sensitivity to see if that would help.

Unreal Tournament 16Bit & 32Bit Tests

The 'cityintro' timedemo was run in Unreal Tournament full version with the 405 update applied. All settings were at their defaults (high detail, low audio) except the min desired framerate was set to 30 (not 20 as the default). All results are in frames-per-second, higher is better. Since the UT's timedemo stats reports min, max and average framerates, all are listed below. (Rage128/Rage128 Pro results from my previous peformance comparison article and therefore have no 800x600 results.)

The PowerBook G3/500 ran OpenGL mode and used GameLauncher (UT OpenGL allocated 142MB of RAM). Also important to note is that the PB G3/500 (OpenGL mode) tests used 'Medium' detail skin & world detail - all other systems used Rave mode and high detail settings.

Unreal Tournament 16Bit 640x480
PB G3/500
(OpenGL mode)
G4/450 Rage128 AGP
(Rave mode)
G4/450 Rage128 PRO AGP
(Rave mode)

The same tests were run in 32Bit mode:

Unreal Tournament 32Bit 640x480
PB G3/500
(OpenGL mode)
(too little vram)
G4/450 Rage128 AGP
(Rave mode)
G4/450 Rage128 PRO AGP
(Rave mode)

Note: Remember the UT cityintro is has no firing or controller inputs, multiple enemies, etc.
- actual framerates in the game in heated action would often be lower.

Image Quality Notes: I've included screenshots of Quake3 in 32-bit color mode. The thumbnails below are linked to full size images for closer examination. The LCD in use is not as sharp at 640x480 as its native 1024x768 resolution (and there's no optimize function as with desktop displays to sharpen the image). However the game still looks very good in 32Bit mode.

Quake3 32bitQuake3 32Bit SS 2

(Click to see 640x480 high quality jpeg images appx. 65KB)

Comparing 16Bit to 32Bit OpenGL Image Quality
16Bit Image32Bit Image

Suggested Resolutions by Game Title:

  • Quake1: 800x600 or 640x480
  • Quake2: 640x480
  • Quake3: 640x480 (32bit color looks best)
  • Unreal Tournament: 640x480 (if 32Bit mode)

Performance Tips: Reducing game settings (quality/detail levels) can help with higher resolution modes' performance but don't expect miracles. For network/deathmatch play, responsiveness is more important than looks to stay alive. Setting Vertex instead of Lightmap lighting in Quake3 for instance can help quite a bit. You may find that 640x480, 32Bit runs fine however with most default settings.

Quake 3 Performance Tips: To boost your game framerate without sacrificing too much in image quality, here are some Game Options I often disable in Quake3 that improved framerates:

  • Dynamic Lighting: Off
  • High Quality Sky: Off
  • Marks on Walls: Off
  • Ejecting Brass: Off
  • Simple Items: On
  • set cg_gibs 0 (in console or autoexec.cfg file)
    (helps as much as 5 FPS some readers said, reducing gore but otherwise no real negative image quality impact)

This make the game far smoother in 32Bit mode and it still looked good. Reducing geometric detail and texture quality can help was well. For even more FPS boost, selecting Vertex lighting instead of Lightmap, but that really hurts the realism in the game.

About RAM: Performance may have improved a bit with Virtual Memory off - however even with 128MB of installed RAM, many games were unplayable with VM disabled (remember many games now want 70MB to 100MB of RAM and the system uses 20+MB, and OpenGL needs available RAM as well).

I highly recommend serious gamers run 192MB or more of RAM, especially of you want to run Unreal Tournament in OpenGL mode (UT was set to 142MB of RAM, plus the OS and OpenGL need available RAM as well). Adding RAM is the first upgrade to make to any Mac.

Summary: Applications and Game performance was the best I've ever seen on a portable as of spring 2000. Overall I was pleased with game performance and it's dramatically improved compared to previous notebooks. (Most all of the games listed on this page are not playable at all on my Wallstreet RageII based graphics chip). I'd estimate the Rage128 Mobility has twice the 3D horsepower of the previous RageProLT in the Lombard and Wallstreet II models. However no portable graphics chip can compare to the performance of the desktop models in demanding 3d games like Quake3 and Unreal Tournament.

Although I wish there was a 8MB VRAM upgrade option, I still think the PB G3 Firewire is a well rounded machine and the first notebook to have near the 3d graphics power of the Rage128 Mobility. (There will be faster chips in the future of course, which will make the Rage128 mobility seem anemic I'm sure.) Powerbook Firewire owners (with adequate RAM) will now be able to enjoy the latest games for the Mac anywhere they go. For the most demanding 3D games, reduced settings will be required however and they won't be as smooth as modern desktop Macs with better graphics cards and more video memory.

Other Sources of Game Performance Comparisons:
  • My Video card reviews page lists articles and reviews with comparisons of literally every popular Mac 3D video card currently on the market as well as reviews of cards from the past.
  • Articles in the Game News page sidebar.
  • For 3Dfx video card specific information, see my new site.

The next page shows the results of benchmark tests (CPU/Disk/Video/CDrom).
Or you can use the links below to jump to other parts of the review.

Index of PowerBook Firewire G3/500 Review

Intro | Apps/Game Performance | Benchmarks | Compatibility | DVD/Movies | Specs | Summary

- or -

Copyright © 2000, All Rights Reserved.
All brand or product names mentioned here are properties of their respective companies.