The Source for Mac Performance News and Reviews

Review: XLR8's MAChCarrierTM G4/450 By Mike Breeden Published: 4/7/2000

*Real World* Performance Tests

Intro | Benchmarks | Appl. Tests | Software Controls | Installation | Specs/Design | Summary

Applications and Game Performance

This page lists test results in common Mac applications like Photoshop 5.5, Bryce 2, Infini-D 4.01, After Effects 3.1, Premiere 4.2.1 and popular 3D games like Quake2, Quake3 and Unreal Tournament. Although I normally include Soundjam MP tests in G4 upgrade reviews since it takes advantage of Altivec/Velocity Engine instructions, due to issues with the performance of the Imation CDR drive audio extraction, results are not listed for this review. (The Imation was the only CD drive I had to spare for this built-from-scratch system). Intech's CDTools drivers was not optimized for the Imation drive (even though it is a Plextor mechanism, the drive's firmware reports Imation) and actually had lower audio extraction than the Toast reader extension. With Toast's CDR reader extension, Soundjam would run separate extraction and encoding phases, making comparisons to past results impossible. For examples of SoundJam performance with G4 vs G3 CPUs, see the past G4 reviews on the CPU Upgrades page.

Photoshop 5.5 Tests:

I ran a complete PS5Bench (21 filter test) series using Photoshop v5.5 with the Altivec extensions (604e CPU tests had the Altivec plugins removed of course). Standard PSBench settings are 1024x768, millions colors, VM off, Interpolation set to bicubic (better) and Photoshop allocated enough RAM to avoid any swap file activity from the 10MB test image filter actions. (I allocated 140MB to Photoshop 5.5 for past reviews, but 80MB for this review due to the fact the S900 had only 144MB of RAM installed.) To improve consistency of times on the filter tests (each filter is run 3 times), I set Photoshop 5.5's 'History' settings from the default 20 to 1 and unchecked the 'automatically create snapshot' option. This dramatically lowered the filter times and removed all signs of disk activity during the filter tests (each filter is run 3 times during the test). Note: All G4 CPU tests had the 4 OS 9 Altivec extensions active. All G4 tests except the XLR8 G4/450z in the B&W G3 (from previous review) used the Adobe current Altivec Core (v5.5.1) and Lighting Effects (v5.5.2) Filter. The G4z/450 results used the first version of the Altivec Plugins and core. The v5.5.2 Lighting Effects plugin is faster than the original v5.5.1 (appx 15% based on tests with the G4/450 AGP). The G4/450 upgrade in the S900 had the current v2.9 G4 CPU revision, whereas the 100MHz bus speed B&W G3 and G4/AGP systems had a v2.6 G4 CPU revision (with errata). The "Gain" column indicates how many times faster the S900 with G4/450 upgrade ran the filters than the S900 with a 604e/200MHz CPU card. Filter GAIN (w/Upgd) S900 +XLR8 G4/450 S900 604/200 B&W G3/400 Apple G4/450 AGP XLR8 G4/450Z in B&W G3 Rotate 90° CW 4.40x 0.5 2.2 0.8 0.3 0.3 Rotate 9° CW 2.07x 2.9 6.0 3 2.3 2.3 Rotate .9° CW 2.12x 2.6 5.5 2.8 2.2 2.2 1 pix Gaus. Blur 2.88x 0.8 2.3 1.4 0.5 0.7 3.7 pix Gaus. Blur 2.67x 2.4 6.4 3.3 1.6 1.9 85 pix Gaus. Blur 3.17x 2.9 9.2 4.3 1.9 2.4 Unsharp Mask 50%/1pix/0 level 2.54x 1.3 3.3 1.7 0.8 0.9 Unsharp Mask 50%/3.7pix/0 level 2.68x 2.8 7.5 3.7 1.9 2.2 Unsharp Mask 50%/10pix/5 level 2.81x 2.7 7.6 3.7 1.9 2.3 Despeckle 6x 0.8 4.8 2.5 0.8 0.9 RGB-CYMK 2.35x 4.6 10.8 5.1 4.7 4.5 Reduce 60% 3.43x 0.7 2.4 1.3 0.5 0.5 Lens Flare 2.42x 4.3 10.4 4.4 3.4 4.0 Color Halftone 2.36x 4.2 9.9 3.9 3.4 3.5 NTSC Colors 2.88x 4.1 11.8 4.2 3.9 3.9 Accented Edges 2.37x 10.4 24.6 11.3 10.2 9.9 Pointillize 2.37x 14.3 33.9 15.9 14.5 14.1 Water Colors 2.36x 21.4 50.4 23.6 21.7 20.8 Polar Coordinates 3.33x 3.9 13.0 5.4 3.3 3.5 Radial Blur 2.21x 32.6 72.2 36.2 34.6 32.1 Lighting Effects 8.33x 2.4 20.0 9.3 2.0 2.5 Total Time 2.56X 122.6 314.2 147.8 116.8 115.4 System GAIN (w/Upgd) S900 +XLR8 G4/450 S900 604/200 B&W G3/400 Apple G4/450 AGP *XLR8 G4/450Z in B&W G3 * XLR8 G4/450Z upgrade used initial release of the Altivec Plugins, current versions are faster. FYI: A 600MHz AMD Athlon PC with 256MB 222 SDRAM running Windows 98 took 138.7 seconds to complete these tests.

Infini-D 4.01 Tests

Infini-D 4.01 does not use Altivec extensions, but is a common application I have used for comparisons of CPU/FPU performance. I used the same 'Chapter 7 completed' tutorial scene file from my past reviews. Rendering quality was set to Ray Trace, medium anti-aliasing, shadows on, patch detail low. I didn't change the default QT movie output file options. The graph below shows times to complete the 150 frame movie rendering with the S900 with a 604e/200MHz, the same S900 w/XLR8 G4/450 CPU, a 9600/350 (350MHz 604ev), a 9500 with XLR8 G3/450, an Apple B&W G3/400, and my rev 1 Apple G4/450 AGP (Sawtooth) system. All systems were running OS 9.0. Infini-D was allocated 40MB of RAM for the tests. OS 9 and its 4 Altivec extensions were used for all G4 tests. Since Infini-D draws each frame as it renders the file, I wanted to note that the S900's Voodoo3 has no Quicktime acceleration, unlike the other graphics cards used in the comparison systems. I doubt this really affected the scores, but wanted to mention it. The other systems also had more RAM (256MB to 320MB installed), the S900 had 144MB. However the application was allocated the same amount of RAM and Virtual Memory was off for all systems.

Bryce 2 Tests

I also ran tests in Bryce 2, using the samples 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. Again a G3 CPU of the same speed does just as well as a G4 CPU in this old and non-Altivec aware application. (Remember a second or two difference in timing is within normal run/run variation) . Bryce2 does not take advantage of Altivec extensions of course (but neither does most applications you're using today). Perhaps there is something in the Bryce2 and Infini-D 4 code that is better suited to a G3 rather than a G4.

After Effects 3.1 Tests

The following is a chart showing the time to render a special effects movie (appx. 8MB file size) in After Effects 3.1. Resolution was set to 1024x768, thousands colors. (The test project is from KPT Final Effects AE addon package, the Fire Effects demo project.) Interesting that the S900 with G4/450 upgrade was slightly faster than the G4/450 AGP system in this particular 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. Again I'm not sure it matters in this test, but the S900's Voodoo3 has no Quicktime acceleration, unlike the other graphics cards used in the comparison systems.

Game Performance Tests:

Unlike many 'serious' applications, 3D games are one of the best tests to show a system's overall performance. Very few applications stress the total system more than modern 3D/OpenGL games. The CPU, Memory Bus and Video subsystems are pushed very hard in these games, therefore they are a great way to show the benefits of more modern PCI/Memory controllers, faster memory bus and more advanced motherboard designs. Tests were done in Quake2, Quake3 (retail) and Unreal Tournament. As I've said in most all my G3/G4 CPU upgrade reviews (and video card reviews), the graphics card is a major factor in the speed and quality of 3D games. The first gamer's purchase should be a fast 3D card if you don't already own one. The S900 used a retail Voodoo3 3000 PCI card (standard 33MHz PCI bus of course). The B&W G3/400 and Apple G4/450 systems used a Rage128 PCI and AGP card respectively (both running in 66MHz slots). All systems had OS 9.0 and OpenGL 1.1.2 installed. The Rage128 cards used Univ. Driver release 4.2. The Voodoo3 (S900) used beta 11 drivers (current as of this review date). All G4 tests had OS 9's Altivec extensions active. Since the Voodoo3 in the S900 does not support 32Bit mode OpenGL, all tests were done in 16Bit mode. As you'll see below and shown in past video card reviews here, the Voodoo 3 is a higher performance 3D gaming card than any of the current Rage128 cards. This is why the S900 with G4/450 upgrade delivered higher framerates in most cases than the G4/450 AGP and the B&W G3/400 systems, even though they have faster memory busses (twice as fast) and more modern motherboard chipsets. Quake 2 Tests: OpenGL mode. All tests used OpenGL 1.1.2. Graphics settings in the game were : 16-bit color Max Texture Quality No Sync Every Frame No CD Audio Full Screen Quake 2 1024x768 framerates are typically twice as high with a Voodoo3 as the Rage128, but the new Rage128 PRO in the latest Apple G4 models performs better, but still no match for the Voodoo3 at higher resolutions. Despite a faster motherboard and 2x faster system bus, the fill rate of the Rage128 chip limits performance at higher resolutions. For comparisons of other video cards in a G4/450 AGP system, see my Nov 1999 Mac 3D Card Roundup. Note: By applying the Quake2 v1.03 config tweaks noted in my update to the PowerBook Firewire review to the G4/450 AGP system, the Rage128 Pro delivered 40FPS at 1024x768. Quake 3 Tests: Graphics settings at the defaults except where noted otherwise: 16-bit color Lightmap lighting Geometric detail high Texture quality 16-bit Texture detail 1 notch from max Texture filter - bilinear The tables below show Quake3 16-bit color mode performance with each of the system/upgrade combos in frames-per-second. In this game, the G4/450 AGP with Rage128 Pro card compared better at higher resolutions against the S900/Voodoo3. I suspect Quake3 benefits from the faster memory bus of the G4/AGP system more than Quake2 (it's a more resource intensive game, illustrated by the lower framerates on all systems compared to Quake2). For comparisons results with other systems/video cards see my Mac 3D Card Roundup or Mac Game/FPS Database. Unreal Tournament 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. (G4/450 AGP Rage128 and Rage128 Pro results are from my previous performance comparison article which didn't test 800x600 resolution. Since that test used min desired FPS=30, I used the same setting in the S900. Setting this number to 0 would have been preferred, but I didn't have time to retest the other system/cards at that setting.) The S900 results were in 3dfx mode of course since it used a Voodoo3 3000 card. Based on tests in the Powerbook Firewire review, OpenGL mode may perform better with the Rage128 cards than RAVE mode. Unreal Tournament 16Bit 640x480 AVG FPS (Min/Max) 800x600 AVG FPS (Min/Max) 1024x768 AVG FPS (Min/Max) S900 w/604e 200MHz (3dfx mode) 15.21 (7.46/36.53) 14.75 (7.38/35.87) 12.2 (7.13/34.54) S900 w/XLR8 G4/450 (3dfx mode) 34.56 (16.76/76.44) 33.56 (16.60/74.72) 32.32 (16.2/72.08) G4/450 AGP Rage128 PRO AGP (Rave mode) 39.20 (21.9/84.25) Not Tested 25.57 (13.08/54.54) G4/450 AGP Rage128 AGP (Rave mode) 33.37 (16.67/78.28) Not Tested 20.86 (12.44/43.68)

Summary: Overall applications and game performance of the upgraded S900 was surprisingly good, even compared to an Apple G4/450 'Sawtooth' system. The Voodoo3 card certainly helped game performance as well. As shown in the results, most non-Altivec enhanced applications show no real benefit from a G4 CPU compared to a G3 of the same speed, but for those applications that do support Altivec, significant gains are possible. (Applications that use double-precision floating point math should show a benefit with a G4 however.) However in cases like Photoshop, only specific filters benefit. More programs will support Altivec (Velocity Engine) in the future, but consider your specific needs when deciding on a G3 or G4 CPU upgrade.

The next page describes the software cache control supplied with the card. Or you may use the links below to jump to a specific page.

Index of XLR8 MAChCarrier G4 450MHz Review Intro | Benchmarks | Appl. Tests | Software Controls | Installation | Specs/Design | Summary - or - Back to WWW.XLR8YOURMAC.COM

Copyright © 2000. All Rights Reserved. All brand or product names mentioned here are properties of their respective companies. Users of the web site must read and are bound by the terms and conditions of use.