Reviews and Daily News with a  Difference!
Select a Topic Area To See a List of Related Reviews & Articles
Systems  | CPU Upgrades  | SCSI | IDE | Firewire  | Video  | Audio  | Games  | Misc/OS  | Archives  | Search
Testing XLR8's Dual G4 for Beige G3/PowerMac G3Return to News Page

Click for SMOKING FAST 6G SSDs!
Click for SMOKING FAST 6G SSDs!

XLR8's MPe Dual G4/500 Upgrade for Beige G3
Performance Compared to Dual G4/500, G4/733 and 1GHz PC

Published: 3/19/2001
(Updated 3/23/2001 for Summary PShop 5.5 graph)

NOTE: This article was written in March 2001. In 2002 XLR8.com closed their doors and this upgrade is no longer available. Also in 2002 there were some problem reports from owners of this upgrade in the forums here, so considering the cost (which was very high), you could sell the Beige G3 and buy a used Dual G4 500 Mac w/AGP slot, onboard Firewire/USB for about the same total cost and have a better performer to boot.
(Original Article text from March 2001 follows)

XLR8 MPe Dual G4

XLR8 MPe G4/500 in Beige G3 MiniTower

Intro:
Thanks to a loaner from XLR8, I've been able to test the performance of their new Dual Processor G4/500 CPU "MPe" upgrade for Beige G3s. The MPe upgrade arrived installed in a "bare-bones" Beige G3/233 Minitower (original 4GB IDE drive, onboard video, etc. - no faster IDE or SCSI controller card or drives, etc.) Since the 2MB onboard video could not run 1024x768/millions colors, I installed an ATI Radeon PCI card running the latest 1.1.1 driver update for these tests. I also installed a Firewire PCI card to allow transferring test files and apps. (The full system details of the Beige loaner and other systems used in this article are listed at the bottom of the page.)

Lest you think this upgrade isn't rugged by its appearance, during shipping the side cover was broken (dropped package I assume). I was leery that the shock from this might have affected operation, but the system powered up and ran right out of the box. Note the XLR8 dual CPU module extends to the edge of the case and requires disconnecting the floppy drive cable.
(Note: The clearance from the top of the floppy drive connector and the bottom of the module is very close - I saw no problems with this, but a reader suspected a failure he had may have been due to the bottom of the module's circuits contacting the pins in the floppy connector. Although this may be a rare occurance (and not proven to be floppy pin contact), due to the close proximity, applying some electrical tape as insulation over the floppy connector's pins can't hurt.)

XLR8 MPe Info:
First a few words on the XLR8 MPe Dual CPU capable upgrade. Here are comments from Jack Kolk, VP and General Manager of XLR8 on the MPe's features and compatibility:

  1. This card currently supports Mac OS version 8.6, 9.0, 9.01, 9.04 and 9.1.

    OS X multiprocessing support will be added in Q2, obviously after Apple ships the golden master. However, the card will work with OS X as a single processor card at the speed of the processor in the LIF socket ( there are two sockets, one LIF and one ZIF in which the G4 processors are installed).

  2. The 2 ZIFs on this card are MPe, Multiprocessing Enabled. Only XLR8ís ZIFs are MPe. Non MPe G4 processors will not work anywhere near as well as they do not support the full MESI protocol as required for high performance multiprocessing. Interested individuals can learn more from our white paper at: www.XLR8.com/News/Press/000705-MPWhitePaper.html

  3. The processors on the card do not have to be symmetrical, meaning you could start with dual 400's and later add a 500 and even later replace a 400 with a 733 and finally upgrade to dual 733's. We have not seen any performance penalty for having 2 different speed processors.

    (Question) So you are saying you can mix a 7400 and a 7450? (667 and 733's are 7450 CPUs with on-die L2 cache)

    (Answer) Yes, you will be able to mix them. The beige will only work at 733, because on the 7450 the multiplier steps are 1, not 1/2 like the current 7400.

  4. XLR8's ZIF technology is modular, upgradable and configurable. You can start with one MACh Velocity G4 MPe ( a "solo" card) and upgrade it to a MACh Velocity G4 MPe dual processor card by adding a second processor. Conversely you could start with a MACh Speed G4 ZIF MPe and later buy a MACh Velocity G4 MPe and add your current processor to it. The XLR8 explanation of the options can be viewed at : http://xlr8.com/mpe/MPeInfo.html
    -Jack Kolk, XLR8"

Applications Tested:
The system arrived on Saturday morning and so far I've run MPEG video compression tests in Cleaner 5.02, iMovie2 exports, Photoshop 5.5/6.0 filter tests and CineBench2000 (3D/OpenGL/Rendering tests). All tests compare at least the Beige G3 w/MPe to my dual G4/500 AGP system. The Photoshop 5.5 filter test results include comparisons to the performance of my dual G4/500 AGP, a G4/500 AGP (single CPU), a G4/733Mhz (single CPU) and 1GHz Pentium III system.

XLR8 targets the MPe upgrades for those in the print and video production business. They indicated their survey noted the highest interest in dual-processor upgrades was from Beige G3 owners and that it was a common machine in many professional printing/publishing companies. The MPe upgraded system was surprisingly stable, with no problems seen during any of the tests. More information on the XLR8 MPe upgrades for Beige G3s is available at www.XLR8.com. (Note: XLR8.com is not affiliated with, nor a sponsor of this site.)

Time willing, I may be adding results from other tests later. Although this upgrade is not targeted towards gamers, I ran some quick tests in Quake3 just to satisfy those that are curious - the results are listed near the bottom of this page. (At over $1000 for this upgrade, hard-core gamers should consider selling their Beige G3 and pooling the money towards a new 4xAGP/PC133 based Mac with faster 3d card if they're after the best 3d game performance possible.)

FYI: I've asked for permission from XLR8 to replace the MPe in this loaner Beige G3 system to test with other ZIFs for comparison, but have not received a reply yet. I've added a note below under the Photoshop 5.5 tests that include Beige results with a stock G3/233 system and a Beige G3 with G3/450 ZIF upgrade from a previous review.

Cleaner 5.02 MPEG Tests:
Cleaner 5.02 is the latest version of the popular Media Cleaner Pro utility by Terran (now owned by Media100). For a list of features and formats it supports (Mac and Windows) and why it's so popular, see this 2pop Cleaner 5 review and the Terran Cleaner 5 product page. This test used a 320x240 Sorensen QT movie converted to MPEG. (Using wizard settings of CDROM, 4x, High-End, MPEG settings. Data rate avg 2.0Mbits/sec.) The graph below shows the total time for the Beige G3 w/XLR8 MPe and my Dual G4/500 AGP system. (Timings were from Cleaner so there's no human stopwatch error.)

Cleaner 5.02 results

iMovie2 Export Tests:
iMovie2 was used for an easily repeatable test of movie export performance. I take the Tutorial project and stack the clips end-to-end (no transitions to reduce the number of variables). I export to CDROM Medium size movie, with the default settings (H263 codec, 320x240, 44KHz audio, etc.) The graph below shows the total time for the Beige G3 w/XLR8 MPe and my Dual G4/500 AGP system. (Timings were from a stopwatch so there's some chance of human error.)

Tests were done with both Quicktime 4.12 and Quicktime 5 (beta) preview 3. As noted in my previous tests, QT5 shows dramatically higher performance for this test, even on a single CPU system.

iMovie2 H263 export

It's amazing to see that QT 5 cut times nearly in half. (Note: iMovie 2.01 was used since both systems were running OS 9.04. (iMovie 2.03 requires OS 9.1.)


Photoshop Filter Tests:
PSBench's 21 filter action script was created as a cross-platform test of Photoshop performance, long before the G4 CPU based Macs were available. Only a few of the 21 filters in the test are accelerated by Altivec (Lighting Effects benefits most, about 4x faster). The overall score is heavily weighted towards non-Altivec supported filter functions (the filters that took longest to run don't benefit from Altivec), so the total test times don't show a large difference between a G3 or G4 of the same clock speed. However since users don't always use Altivec aware filters all the time, the results are still a valid comparison of system performance for the filter functions listed.

Rather than just a simple graph showning total times for the filter series, as in my other reviews, I've listed each filter, its description and the time each system took to complete it. (Each filter is run 3 times, the avg. is displayed in the table below. PhotoShop's timing function is used, not a stopwatch.)

The graph below shows a comparison of the Beige G3 with XLR8's G4/500 MPe (dual CPU) in yellow, compared to several other systems.

PS 5.5 summary

Filter-by-Filter Performance Comparisons:
A web page isn't wide enough to show every system tested in the past here, so I selected several for a filter-by-filter comparison including my G4/500 Dual Processor system and a reader's G4/733 system. (This also allows you to compare performance filter-by-filter to see the difference by function, not just total time.) The table below shows the times in seconds to perform each of the 21 filter operations on a 10MB image. (Fastest filter times are in bold.)

Photoshop 5.5 Tests

PS5Bench Test
(10MB Image File)

HP Pavilion 1G
1GHz PIII CPU

256KB L2 at 1GHz
133MHz bus
256MB PC800 RAMBUS
Geforce 2 AGP (64MB)
Win98 SE
VM On
Apple G4/500 AGP
500MHz G4 7400 CPU
1MB L2 at 250MHz
100MHz Bus
256MB SDRAM (222)
Rage128 Pro AGP (16MB)
OS 9.04
VM OFF
Apple G4/500 MP
Dual 500MHz 7400 G4 CPUs
1MB L2 at 250MHz x2
100MHz Bus
1GB SDRAM (222)
Radeon AGP (32MB)
OS 9.04
VM OFF
Beige G3
w/XLR8 MPe

Dual 500MHz G4 7400 CPU
1MB L2 at 250MHz
66MHz Bus
320MB SDRAM
Radeon PCI (32MB)
OS 9.04
VM OFF
G4/733 w/Radeon AGP
733 MHz G4 7450 CPU
256KB L2 at 733MHz
1MB L3@244MHz
133MHz Bus
256MB SDRAM (222)
ATI Radeon AGP (32MB)
OS 9.1
VM OFF

Rotate 90

0.2

0.2

0.2

0.3

0.2

Rotate 9

1.6

2.0

1.2

1.8

1.2

Rotate .9

1.5

1.9

1.2

1.6

1.2

Gaussian Blur 1

0.5

0.5

0.4

0.8

0.3

Gaussian Blur 3.7

1.3

1.4

0.9

1.6

1.1

Gaussian Blur 85

3.2

1.6

1.1

2.0

1.3

Unsharp 50/1/0

0.6

0.7

0.5

0.9

0.6

Unsharp 50/3/7/0

1.4

1.7

1.0

1.8

1.3

Unsharp 50/10/5

1.6

1.7

1.1

1.9

1.4

Despeckle

1.7

0.8

0.5

0.8

0.6

RGB-CMYK

2.6

4.0

4.2

4.1

2.3

Reduce Size 60%

0.6

0.4

0.3

0.5

0.2

Lens Flare

2.3

3.3

2.2

3.2

2.6

Color Halftone

3.5

3.0

3.0

3.1

2.4

NTSC Colors

3.2

3.4

3.5

3.4

2.7

Accented Edges

7.4

8.9

9.1

8.8

7.2

Pointillize

8.7

12.5

7.4

7.6

9.6

Water Color

15.8

18.7

19.3

18.6

15.9

Polar Coordinates

5.7

2.9

2.2

2.8

2.3

Radial Blur

21.2

29.9

16.2

17.4

15.9

Lighting Effects

1.2

1.7

1.2

1.4

1.3

PS5Bench Index
(time to complete)

85.8
Seconds

101.2
Seconds

76.7
Seconds

84.4
Seconds

76.4
Seconds

System

HP Pavilion 1G
1GHz PIII CPU

256KB L2 at 1GHz
133MHz bus
256MB PC800 RAMBUS
Geforce 2GTS AGP (64MB)
Win98 SE
VM On
Apple G4/500 AGP
500MHz G4 7400 CPU
1MB L2 at 250MHz
100MHz Bus
256MB SDRAM (222)
Rage128 Pro AGP (16MB)
OS 9.04
VM OFF
Apple G4/500 MP
Dual 500MHz 7400 G4 CPUs
1MB L2 at 250MHz x2
100MHz Bus
1GB SDRAM (222)
Radeon AGP (32MB)
OS 9.04
VM OFF
Beige G3
w/XLR8 MPe

Dual 500MHz G4 7400 CPU
1MB L2 at 250MHz
66MHz Bus
320MB SDRAM
Radeon PCI (32MB)
OS 9.04
VM OFF
G4/733 w/Radeon AGP
733 MHz G4 7450 CPU
256KB L2 at 733MHz
1MB L3@244MHz
133MHz Bus
256MB SDRAM (222)
ATI Radeon AGP (32MB)
OS 9.1
VM OFF

Other Beige G3 Results: Based on past Photoshop 5.02 tests from my Formac G3/450 upgrade review, a stock Beige G3/233 took 254.0 seconds for this test. With the G3/450 ZIF upgrade installed and a VR128 graphics card, the total time was 130.0 seconds.



Photoshop 6.0 Tests:
I also compared the Beige G3 with XLR8 MPe Dual G4/500 to my G4/500 Dual processor AGP system using Photoshop 6.0. (The PShop 6.01 update was not applied.)

Photoshop 6.0 Tests

PS5Bench Test
(10MB Image File)

Apple G4/500 MP
Dual 500MHz 7400 G4 CPUs
1MB L2 at 250MHz x2
100MHz Bus
1GB SDRAM (222)
Radeon AGP (32MB)
OS 9.04
VM OFF
Beige G3 w/XLR8 MPe
Dual 500MHz G4 7400 CPU
1MB L2 at 250MHz x2
66MHz Bus
320MB SDRAM
Radeon PCI (32MB)
OS 9.04
VM OFF

Rotate 90

0.2

0.4

Rotate 9

1.2

1.8

Rotate .9

1.1

1.6

Gaussian Blur 1

0.3

0.6

Gaussian Blur 3.7

0.6

1.1

Gaussian Blur 85

1.0

1.8

Unsharp 50/1/0

0.4

0.7

Unsharp 50/3/7/0

0.9

1.3

Unsharp 50/10/5

1.0

1.6

Despeckle

0.3

0.5

RGB-CMYK

5.9

5.7

Reduce Size 60%

0.3

0.5

Lens Flare

1.9

2.7

Color Halftone

2.5

3.0

NTSC Colors

2.6

2.6

Accented Edges

7.8

7.6

Pointillize

7.3

7.4

Water Color

17.4

16.8

Polar Coordinates

1.6

2.1

Radial Blur

15.9

16.9

Lighting Effects

3.9

4.3

PS5Bench Index
(time to complete)

74.1
Seconds

81.0
Seconds

System

Apple G4/500 MP
Dual 500MHz 7400 G4 CPUs
1MB L2 at 250MHz x2
100MHz Bus
1GB SDRAM (222)
GeForce2MX AGP (32MB)
OS 9.04
VM OFFVM OFF
Beige G3 w/XLR8 MPe
Dual 500MHz G4 7400 CPU
1MB L2 at 250MHz
66MHz Bus
320MB SDRAM
Radeon PCI (32MB)
OS 9.04
VM OFF

PhotoShop Summary:
The dual G4/500 (Gigabit Ethernet) system has a faster bus/memory speed (1/3 faster), faster graphics card/interface, hard drive/controller ect. so some may say this isn't really a fair test. However I compared them since as with any expensive upgrade, you should consider the upgrade cost vs. getting a new system. (I didn't have a dual G4/533 system to test, but for these filter tests the difference is very small I suspect.) Although it trailed the dual G4/500 AGP system a bit in most filters, I was surprised to see the XLR8 upgraded Beige was faster in a few filters than the Dual G4/500 AGP system. I can only guess the XLR8 MP plugin and control software was the reason why.

With the amount of RAM allocated to PhotoShop (with a 10MB image size), in these filter tests the hard drive and graphics card should have a minimal effect. I suspect the faster RAM/System bus speed is primarily responsible for the differences in performance for the filter tests. To be honest, I never thought I'd see a Beige G3 be able to come close to my dual G4/500 in Photoshop performance.

(For gamers, the faster graphics card port and other features of the later G4 systems would make them a better choice than upgrading a Beige G3. The onboard firewire ports and faster IDE interface of the later G4s is also a plus, especially if you've not upgraded your Beige G3 with a faster IDE controller or Firewire card.)

CineBench 2000 Tests:
As regular readers know (from past reviews and the Video card/System performance database here) Maxon, makers of Cinema 4D (a 3D modelling/rendering/animation program) has a benchmark program to gauge/compare performance in 3D applications called CineBench 2000. The CineBench test consists of the following tests:

  • A wireframe and shaded terrain spin (using Cinema software and then OpenGL acceleration)
  • A "Fly-through" of a fairly detailed scene in both wireframe and shaded mode (using Cinema software and then OpenGL acceleration)
  • Rendering of a sample scene (using both processors if present)

The scores below compare the Beige G3 w/XLR8 MPe upgrade (w/Radeon PCI card), my Dual G4/500 AGP system (w/Radeon AGP card) and a reader's G4/533 AGP system (w/Radeon AGP card). Scores are in 3 categories: Shading (Cinema's software 3d), OpenGL (hardware accelerated 3d), and Rendering (time to render a sample scene. The software is MP aware and the Rendering scores reflect the dual CPU test results. (Higher numbers are better.)

Cinebench 2000 results

There were some dual G4/533 entries in the database with lower scores than the ones I used, (about the same as my dual G4/500), so results can vary depending on graphics card drivers used, etc. (The dual G4/533 results used OS 9.1, other systems used OS 9.04 which could also be a factor.) Graphics mode for all systems in the graph was 1024x768, millions colors. Each used an ATI Radeon card with 1.1.1 drivers. (The Beige used a PCI Radeon since it has no AGP slot.)
To download Cinebench for your own tests, see http://www.maxon.net/pages/download/benchmarks.html.


Quake3 v117 Tests: Since this upgrade is targeted at publishing/video professionals (not gamers), I only ran a simple series of Quake3 v117 tests at 32bit mode, high geometric detail, texture quality one notch from max and all game options on (HQ sky, ejecting brass, marks on wall, etc. were all enabled). As expected the G4/500 with 2x AGP Radeon card and faster system bus outperformed the Radeon PCI card. (Both cards uses the latest 1.1.1 drivers and OpenGL 1.2.1 under OS 9.04.)

Quake3 results

For other video card/system tests see the Video articles page.


XLR8's Dual Processor Control Panel:
This page is already too heavy to show all the screens, but here's the main one from the XLR8 MPe Control Panel which shows details on each CPU including speeds (CPU, Cache, RAM), temperatures of each CPU and more. (Each tab is a page with more info/details.)

XLR8 Control panel

Availability & Pricing:
(Update - as noted above, in 2002 XLR8.com closed their doors. I don't know of any sources of this model anymore. The review loaner was returned in March 2001.)


System Configurations:
Other than the G4/733 reader's system (detailed in the table), here are the details on the systems I used for testing.

  • Apple Beige G3/233 MiniTower (XLR8 supplied system)
  • XLR8 MPe Dual 500MHz G4 CPUs (rev 2.9)
  • 1MB of L2 Cache at 250MHz (x2)
  • 66MHz system bus
  • OS 9.04
  • Virtual Memory Off
  • 320MB RAM
  • OEM 4GB IDE Hard Drive (ATA/3 onboard IDE)
  • OEM CD-ROM Drive (IDE)
  • ATI Radeon PCI w/32MB DDR SDRAM
    (Using OpenGL 1.2.1 and latest 1.1.1 Radeon drivers)

  • Apple Dual G4/500 AGP
  • Dual 500MHz G4 CPUs (rev 2.9)
  • 1MB of L2 Cache at 250MHz x2
  • 100MHz system bus
  • OS 9.04
  • Virtual Memory Off
  • 1GB RAM (222)
  • OEM 20GB Maxtor IDE Hard Drive (ATA/66 onboard IDE)
  • OEM DVD-ROM Drive (IDE)
  • ATI Radeon AGP (retail) w/32MB DDR SDRAM
    (Using OpenGL 1.2.1 and latest 1.1.1 Radeon drivers)

  • Apple G4/500 AGP (fall 1999 system)
  • 500MHz G4 CPU (rev 2.8)
  • 1MB of L2 Cache at 250MHz
  • 100MHz system bus
  • OS 9.04 with Firmware update
  • Virtual Memory Off
  • 256MB PC100 RAM (222)
  • 45GB IBM 75GXP IDE Hard Drive (ATA/66 onboard IDE)
  • 6X DVD-ROM drive (IDE)
  • ATI Rage128 PRO AGP w/16MB SDRAM

  • HP Pavilion 1G
  • 1GHz Pentium III (Coppermine)
  • 256KB of on-die L2 Cache at 1GHz
  • 133Mhz system bus
  • Windows 98 Second Edition
  • Virtual Memory Enabled
  • Intel VC820 Motherboard
  • 256MB PC800 RAMBUS RAM
  • 40GB Maxtor IDE Hard Drive
  • 10X Pioneer DVD ROM (IDE)
  • 4x/4x/24x CDRW (IDE)
  • Hercules Prophet II 64MB DDR (Nvidia GeForce 2 GTS)
    (owner replaced original GeForce 32MB DDR card)
  • Intel 10/100 PCI network card
  • SoundBlaster Live PCI audio card

Test Settings:
All tests used display settings of 1024x768, millions colors. Virtual Memory was off for all Macs tested. Photoshop had Interpolation set to bicubic (better) and Photoshop was allocated enough RAM to avoid any swap file activity from the 10MB test image filter actions. (128MB was allocated to Photoshop for the systems I tested.) All tests with Photoshop 5.5 used the latest (5.5.2) Adobe Altivec Core Extension, MP extension and Lighting Effects filter.

On the 1GHz PC, Photoshop 5.5 tests included SSE support. (I did not have PShop 6 for Windows to test.) Windows background tasks like McAfee's anti-virus and the task scheduler were disabled. The Windows swap file was set to the same min/max size (384MB) - a common tip to improve performance by preventing Windows from resizing the swap file during apps use).

For more info or to download the PSBench action file, see PSBench Home Page. (There may be a later version of the action script there now, I've used the same one for about a year now to make sure all tests used the exact same script.)



OWC XYM SPECIALS!
(Ad/Sale items)

= UPGRADES by Mac =
Upgrades just for
YOUR Mac!

= Refurb Mac Pros =
(Click for List)

SSDs from under $50!
Fast SSDs for Most Macs/PCs

= ThunderBolt =
Drives, Docks & More

= HARD DRIVES =
Up to 12TB HDDs
HGST, WD, Seagate, Toshiba

= 2.5in HDs & SSDs =
Notebook Hard Drives and DIY drive/case kit bundles

= MEMORY =
Lifetime warranty RAM Upgrades!

= OPTICAL DRIVES =
Internal and External Superdrives/Blu-Ray drives

= VIDEO / DISPLAY =
Graphics cards, Displays, Adapters, Cables & more

= AUDIO ITEMS =
Interfaces, Cables, Software, Speakers, Headphones & more

= SOFTWARE =
Apps, Utilities, OS, VM, Games and more

= WIRELESS =
WiFi and Bluetooth Devices/Adapters/More

= Repair Service =
for iPhone, iPad, Macs

= iPad/iPhone/iPod =
Accessories, Cases, Repairs & More

NuGuard iPhone Case *Extreme* Drop Tested!


XLR8YourMac T-Shirts


 
= back to www.XLR8YOURMAC.com =


= Other Site Topic Areas =
Mac Mods/Upgrades | CPU Upgrades | Storage | Video | Audio/HT | OS Updates/Network | Recent


Copyright © 1997-2017. All Rights Reserved.
All brand or product names mentioned are properties of their respective companies.

Legal: Site Privacy and terms/conditions of use.