Accelerate Your Mac! - Review of the XLR8 MAChCarrier G4/450 CPU upgrade

The MAChCarrier G4/450 ships set to a 50MHz bus speed and 9x Bus/CPU ratio and I ran all tests with a 225MHz backside cache speed (controlled via software). I never touched the card's settings for this review. The S900's motherboard cache (512KB) was disabled unless otherwise noted, and the backside cache was set to copy-back (faster) mode. (See the Software Controls page of the review for info on how cache speeds/modes are set.)

Remember that Macbench Graphics scores will vary depending on resolution, color depth and the installed video card. This is not a reflection on the CPU card under review. Ditto for disk scores, as the hard drive type, speed and free space/fragmentation all affect scores. The baseline 1000 score is based on the results with a Beige G3/300/150/1MB (with extensions on of course).

Although I cannot recommend overclocking to others due to the risks involved, I did try the G4/450 ZIF module in a B&W G3 to see if it would run at 500MHz. Although it booted, there were errors running applications. This sample, at least without voltage tweaks, would not run reliably at 500MHz for those that are curious. Never buy a CPU upgrade assuming it will run overclocked. Each CPU chip, system/ram/cache/motherboard combination is somewhat unique and results can vary. Never assume someone else's results will be reliable or repeatable in your system.

MacBench 5.0 Tests: The graphs below compare the follow configurations and systems

The first graph shows the results of the S900 before and after the G4/450 upgrade
(Video card, OS, Disk, installed RAM, etc. were exactly the same for both tests):

I didn't include G4timedemo results in the S900 due to the fact the benchmark would not run with the Voodoo3 card installed (I did not install the Voodoo2 RAVE extension, which may have helped). Consider the scores below for reference only -- to illustrate the benefit of the G4's Velocity Engine/Altivec with software that supports it.

The graph shows performance from an XLR8 G4 ZIF upgrade in a B&W G3 (overclocked to 450MHz) compared to the same system with a G3/400 CPU and also a G4/450 AGP Mac. Note the improvement a G4 makes over a G3 in this benchmark which takes advantage of Altivec/Velocity Engine instructions.

ATTO Tools Disk Benchmark Results:

I used ATTO's Tools benchmark since it tests the drive and interface, and doesn't use the OS's disk cache by default. I discovered some very interesting results during testing of hard drive performance between the 604e/200 and the G4/450 in the S900. FYI: These tests were done after updating the S900 to OS 9.04 and adding an XLR8 PCI USB card (w/USB drivers v1.4.1).

What I discovered (repeatable) was with the G4 CPU upgrade installed, if the S900's 512K motherboard cache was enabled read *peak* rates were dramatically lower - almost 1/3 lower as shown in the graphs below. I repeated these tests 3 times (rebooting clean before each test as I do in all cases). Note: I do no recommend running with the motherboard cache enabled. As noted in all the site's troubleshooting guides, remove or disable the original L2 cache when installing a G3/G4 upgrade card is usually recommended or required for stability.

Also note that with the Acard IDE controller/Maxtor drive, a 604e/200MHz CPU delivers nearly the same pure disk performance in this benchmark as the 450MHz G4 CPU (250MHz faster CPU).

BTW: Many people still think IDE drives consume far more CPU cycles than SCSI drives. Before the era of IDE busmastering/DMA drivers this was true. Since there is no current utility that will report CPU cycle usage on the Mac, I can't say what CPU usage is on IDE in the MacOS. Recent tests by a respected PC magazine showed the Maxtor 40GB plus40 drive using a mere 3% of CPU cycles in a PIII 500 system running Windows 98. That's as low or lower than any SCSI drive I've seen. The 40GB Maxtor was given a 'kick ass' rating, besting the Seagate Barracuda II and Quantum EX drive. The S900 in this review used the lower performance 27GB model, but you can read my review of the Plus40 Maxtor in a B&W G3 and other hard drive/controller related articles at the storage topics page.

Newer Tech's Gauge Pro 1.01 Results:
Newer Tech's Gauge Pro utility also includes a memory bandwidth test feature. For this review I used GaugePro v1.01. Previous reviews used a beta version. [Note: As this review was being posted Newer released v1.02 of GaugePro, which is said to report much higher rates with G4 CPUs according to a report just received from a G4/AGP system owner running OS 9.04, which is said to have improved memory manager performance via Altivec. I haven't tested with the new version of GaugePro yet.]

The GaugePro results below are from the exact same S900 configuration except for the CPU card as noted in the images. Both upgrades run at the same bus speed.

Reported memory bandwidth was almost double with the 250MHz faster G4 CPU. BTW - a 604e CPU card running at 211MHz and 60.3MHz bus speed reported appx. 38MB/sec rates. I had previous set the PowerBoost Pro 604e card to 60.3MHz bus speed for tests in a PowerCenter Pro some time back and was impressed it even booted at that speed in the S900. I did not run any apps other than GaugePro due to time constraints, so I can't say if that bus speed would have been reliable in long term use. Still impressive for an onboard cache motherboard with interleaved RAM.

FYI: As noted in past G4 ZIF upgrade reviews, my G4/450 AGP (Sawtooth) system reported just over 115MB/sec using a beta version of GuagePro. The v2.6 G4 in the Sawtooth has errata (L1 cache wait states I'm told), so current G4s may have higher results.

The Source for Mac Performance News and Reviews
Review: XLR8's MAChCarrier^TM G4/450 By Mike Breeden Published: 4/7/2000
MacBench 5.0 and Memory Benchmark Test Results
Intro \| Benchmarks \| Appl. Tests \| Software Controls \| Installation \| Specs/Design \| Summary
Benchmark Tests
Benchmark tests were run with MacBench 5.0 as it is the accepted Mac standard. I've also included results of memory bandwidth tests using Newer Tech's free GaugePro utility and ATTO Tools' disk benchmark. The MAChCarrier G4/450 ships set to a 50MHz bus speed and 9x Bus/CPU ratio and I ran all tests with a 225MHz backside cache speed (controlled via software). I never touched the card's settings for this review. The S900's motherboard cache (512KB) was disabled unless otherwise noted, and the backside cache was set to copy-back (faster) mode. (See the Software Controls page of the review for info on how cache speeds/modes are set.) Remember that Macbench Graphics scores will vary depending on resolution, color depth and the installed video card. This is not a reflection on the CPU card under review. Ditto for disk scores, as the hard drive type, speed and free space/fragmentation all affect scores. The baseline 1000 score is based on the results with a Beige G3/300/150/1MB (with extensions on of course). Although I cannot recommend overclocking to others due to the risks involved, I did try the G4/450 ZIF module in a B&W G3 to see if it would run at 500MHz. Although it booted, there were errors running applications. This sample, at least without voltage tweaks, would not run reliably at 500MHz for those that are curious. Never buy a CPU upgrade assuming it will run overclocked. Each CPU chip, system/ram/cache/motherboard combination is somewhat unique and results can vary. Never assume someone else's results will be reliable or repeatable in your system. MacBench 5.0 Tests: The graphs below compare the follow configurations and systems The first graph compares the S900 system with a 200MHz 604e CPU card running a 50MHz bus speed against the same system with the XLR8 G4/450 card installed. The second graph compares the G4 upgraded S900 to the following more systems/upgrades: A stock B&W G3/400 A B&W G3 with XLR8 G4 upgrade set to 450MHz (450/225) with OS 9 & Altivec extensions An Apple G4/450 AGP rev 1 system (v2.6 CPU with errata) All systems were run with as similar an extension set as possible. The first graph shows the results of the S900 before and after the G4/450 upgrade (Video card, OS, Disk, installed RAM, etc. were exactly the same for both tests): MacBench 5.0 Performance - S900 604e/200 vs S900 w/XLR8 G4/450 Upgrade How the S900 w/450MHz G4 Upgrade Compares to More Modern Macs Notes: Some explanation of the MacBench graph and test components. Disk Scores: The S900 tests used an ATA/66 Acard PCI IDE controller and Maxtor 27GB (7200 RPM/2MB cache) drive with Softraid 2.2.1 drivers. G4/450 system used a WD Expert ATA/66 drive (7200 RPM/2MB Cache) with the onboard ATA/66 controller using Apple OS 9 drivers. B&W G3 (and XLR8 G4/400, Newer G4/400) used onboard ATA/33 controller with OEM 6GB Quantum drive using Apple OS 9 drivers. All disks were approximately 40% full and not optimized. Graphics Scores: All systems other than the S900 used an ATI Rage128 (not Pro) graphics card in a 66MHz PCI or AGP slot. The S900 used a Voodoo3 3000 PCI card in a standard 33MHz PCI slot. Due to the Apple 15" Studio Display LCD limit of 1024x768 mode, higher resolutions were not possible. See my video cards page for reviews of alternative graphics cards for these systems. G4Timedemo Benchmark: I didn't include G4timedemo results in the S900 due to the fact the benchmark would not run with the Voodoo3 card installed (I did not install the Voodoo2 RAVE extension, which may have helped). Consider the scores below for reference only -- to illustrate the benefit of the G4's Velocity Engine/Altivec with software that supports it. The graph shows performance from an XLR8 G4 ZIF upgrade in a B&W G3 (overclocked to 450MHz) compared to the same system with a G3/400 CPU and also a G4/450 AGP Mac. Note the improvement a G4 makes over a G3 in this benchmark which takes advantage of Altivec/Velocity Engine instructions. ATTO Tools Disk Benchmark Results: I used ATTO's Tools benchmark since it tests the drive and interface, and doesn't use the OS's disk cache by default. I discovered some very interesting results during testing of hard drive performance between the 604e/200 and the G4/450 in the S900. FYI: These tests were done after updating the S900 to OS 9.04 and adding an XLR8 PCI USB card (w/USB drivers v1.4.1). What I discovered (repeatable) was with the G4 CPU upgrade installed, if the S900's 512K motherboard cache was enabled read peak rates were dramatically lower - almost 1/3 lower as shown in the graphs below. I repeated these tests 3 times (rebooting clean before each test as I do in all cases). Note: I do no recommend running with the motherboard cache enabled. As noted in all the site's troubleshooting guides, remove or disable the original L2 cache when installing a G3/G4 upgrade card is usually recommended or required for stability. Also note that with the Acard IDE controller/Maxtor drive, a 604e/200MHz CPU delivers nearly the same pure disk performance in this benchmark as the 450MHz G4 CPU (250MHz faster CPU). Now look at the peak read rates with the motherboard cache enabled: Now the results with the same system, but a 200MHz 604e card (Since the 604e cards have no onboard cache, the MB cache is used) BTW: Many people still think IDE drives consume far more CPU cycles than SCSI drives. Before the era of IDE busmastering/DMA drivers this was true. Since there is no current utility that will report CPU cycle usage on the Mac, I can't say what CPU usage is on IDE in the MacOS. Recent tests by a respected PC magazine showed the Maxtor 40GB plus40 drive using a mere 3% of CPU cycles in a PIII 500 system running Windows 98. That's as low or lower than any SCSI drive I've seen. The 40GB Maxtor was given a 'kick ass' rating, besting the Seagate Barracuda II and Quantum EX drive. The S900 in this review used the lower performance 27GB model, but you can read my review of the Plus40 Maxtor in a B&W G3 and other hard drive/controller related articles at the storage topics page. Memory Bandwidth Tests: Although 100MHz bus speed Macs have generally 100MB/sec or higher memory performance, note that as shown on the Applications Tests page, the huge advantage in memory bandwidth is often not a factor in many applications tests. The efficiency of the L1/L2 caches isolate the system bus speed in most cases to where it is often not a factor in real world application performance. For this review I've limited bandwidth tests to just GaugePro (past reviews often used Stream and another memory benchmark). Newer Tech's Gauge Pro 1.01 Results: Newer Tech's Gauge Pro utility also includes a memory bandwidth test feature. For this review I used GaugePro v1.01. Previous reviews used a beta version. [Note: As this review was being posted Newer released v1.02 of GaugePro, which is said to report much higher rates with G4 CPUs according to a report just received from a G4/AGP system owner running OS 9.04, which is said to have improved memory manager performance via Altivec. I haven't tested with the new version of GaugePro yet.] The GaugePro results below are from the exact same S900 configuration except for the CPU card as noted in the images. Both upgrades run at the same bus speed. Now the same test with the XLR8 G4/450 CPU card installed: Reported memory bandwidth was almost double with the 250MHz faster G4 CPU. BTW - a 604e CPU card running at 211MHz and 60.3MHz bus speed reported appx. 38MB/sec rates. I had previous set the PowerBoost Pro 604e card to 60.3MHz bus speed for tests in a PowerCenter Pro some time back and was impressed it even booted at that speed in the S900. I did not run any apps other than GaugePro due to time constraints, so I can't say if that bus speed would have been reliable in long term use. Still impressive for an onboard cache motherboard with interleaved RAM. FYI: As noted in past G4 ZIF upgrade reviews, my G4/450 AGP (Sawtooth) system reported just over 115MB/sec using a beta version of GuagePro. The v2.6 G4 in the Sawtooth has errata (L1 cache wait states I'm told), so current G4s may have higher results.
Summary: In other than memory bandwidth tests, benchmark performance of G4 upgrades are as good or better than my G4/450 (v2.6 CPU) Sawtooth system in most cases. Benchmark performance is interesting for technical reasons, however I prefer to rely on real world application performance since that's what really counts. Application and 3D Game tests are covered in the next page of this review. 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.