
Click for External Drive Sales!
![]() The Source for Mac Performance News and Reviews |
|
|
Review: PowerForce G4 350MHz G4 CPU Upgrade for PCI Macs By Mike Published: 1/23/2000 |
|
|
|
|
|
|
|
| Cache Control Software | |
| Powerlogix supplies both a 'Emergency Boot' floppy (more later on this) and a CD, both with their G3/G4 Cache Profiler installer included. The installer places an extension in the System Folder and an alias to the Cache Profiler application in the Control Panels folder.
I suggest leaving the cache speed (set via the Cache Control tab) at the default 2:1 settings and verifying that Speculative Access is disabled (set via the Other tab). Unlike G3 CPU upgrades that generally only had issues with Speculative Access/Speculative Processing enabled with Retrospect backup softare and Adaptec wide-scsi cards, the G4 CPU design may not even boot in an older Mac if speculative access is enabled. I suggest before you replace your existing CPU card, insert the Powerlogix floppy and power-up the system to ensure their software fix for this issue is present in the Mac's Novram before booting with the G4 CPU card. (Booting from the floppy automatically writes the fix to non-volitile RAM and restarts the Mac.) More info is provided in the details on each page of the control below. Overall I like the Powerlogix v1.3 cache profiler; it's uncluttered and has all the options needed. However I have two suggestions for improvements in the next version:
PowerLogix G3/G4 Cache Profiler v1.3
About Cache Speeds: I recently revisited the issue of cache speed by testing an iMac DV at 400/200 vs the stock 400/160. I saw zero gain in 3D games (which move a lot of data over the bus), even running 640x480 game modes to prevent the video chip from being the bottleneck. I saw a very small gain in applications, with only a 2 second reduction in a 5 minute Bryce2 rendering test. Although some owners swore they saw dramatic gains (from MacBench I suspect), I think if they were to actually run timed comparisons they'd see the benefit is very small in most cases. (When G3 CPU cards first arrived in late 1997, I thought a 1:1 cache speed was the holy grail.) If faster cache speeds *prove reliable* in extended use, I have nothing against running them other than they increase heat and may shorten the life of components. However often owners set the cache too high, which can ruin reliability. MacBench's CPU test is influenced by cache speeds far more than real world applications. Remember - a fast unreliable computer is lot less useful (and hazardous to your data) than a slow reliable one. I've pushed and tweaked most every Mac and upgrade I've ever owned (when possible), but don't sacrifice reliability by running on the razor's edge of stability. Write-back (default) vs Write-through Cache Modes: I'm not sure about the current Apple G4 systems, but the previous B&W G3 models used 'Write-through' cache modes vs. the faster 'Write-back' mode. In my tests last year with a utility that would toggle these modes on an Apple G3, I saw very little gain from Write-back mode, at least in the Apple G3. Perhaps with the 66-100MHz memory bus speeds it's less of a performance issue than with older Macs that have a much slower memory bus. Remember data reads are not affected by this cache mode, only writes.
I'm not a Linux OS user, but Powerlogix's CD and control has some unique utilities for Linux. (More information on using the PF G4 with Linux is provided on the CD.)
By installing the software first, this ensures the 'virtual firmware' is present before booting with the new CPU card. I suggest before installing the card, simply insert the PowerLogix boot floppy and power up your Mac before installing the upgrade. This writes the virtual firmware fix to Novram and automatically reboots the Mac with no user intervention required. The G4 is said to be far more likely to cause problems in older Macs if Speculative Access is enabled. This issue affects all Macs made before the first Apple G4 model _except_ the 8600/250 and 300, 9600/300 and 350 models which are reported to have 'G3 clean' ROMs. I've run G4 and G3 CPU upgrades in my 9600/350 with no problems with Speculative Access/Speculative Processing enabled, but tested with it disabled for this review since that's the suggested setting that will be required for most Macs. Powerlogix includes an 'Emergency Boot Floppy' in case you replace the battery or clear the Novram (possible if Zapping the PRAM before the startup bong or from the Finder during a restart. I've not been quick enough to be able to do this personally, but wanted to warn readers of the potential issue. (With SA enabled in an older Mac, a G4 CPU upgrade may not be able to boot, or could cause data corruption if it was able to boot in this mode). If you replace the battery or clear the Novram, insert the PowerLogix boot floppy, power up the Mac and the floppy will re-write the Novram virtual firmware fix and restart the Mac automatically. Newer Tech claims the boot floppy may not be a 100% certain solution (that the G4 may not be able to boot from the floppy 100% of the time) - but tests so far shows it does work. PowerLogix responses to the issue of Speculative Access/Speculative Processing can be found at http://www.powerlogix.com/support/spec.html. What's interesting is that Apple's G4 Sawtooth system has speculative processing disabled in the current version. (Verified here with my G4/450 AGP system). In my older Mac tests, this resulted in a 4% drop in performance in most applications, but actually seemed to help Photoshop filters a bit based on PSbench tests. Dynamic Power Managment: Due to some reader comments recently, I tested putting the Mac to sleep with the upgrade installed. The PCI SCSI card connected hard drive never really went into sleep mode, but otherwise it awoke from sleep fine. (On the new G4 systems, most PCI cards you'll see a report upon waking from sleep with a PCI card installed that the card doesn't allow power to be removed from the PCI bus, so I doubt this ATTO PCI SCSI card allows that either.)
Info page (needs support phone number/email IMHO) Since cache speed is far from a major factor in applications performance and cache overclocking can hurt stability, I always suggest you leave the cache setting at the default (usually 2:1 or 1/2 CPU speed). I never needed to change any of the default settings in over a week of use while performing tests for this review. About Reported CPU Temperatures: Many readers frequently ask if a specific C reading is too high. Most commercial CPUs are rated for 65C operation and if you have a temperature problem you'll know it quickly in my experience - as errors in the OS will occur almost immediately when a CPU overheats. As a side note, my PowerBook G3/250 (Wallstreet 1) has run 76C reported temperatures since day one with no adverse effects. I suspect the PowerBook CPUs may be higher temperature rated parts however (I seem to remember reading 85C somewhere). Bottom Line is that if your CPU is running too hot the Mac OS will let you know very quickly. Photoshop Altivec Enablers: Until early January 2000, only Apple G4 system owners had the Photoshop 5.5 Altivec plug-ins, but Adobe publicly released an updated version of them fixed bugs and improved performance. I did see noticably better Lighting Effects filter performance as shown on the Application Tests page. |
|
|
Summary: The Powerlogix software was easy to install and use. The Linux OS software is a nice plus if you're a Linux fan. One note to those that are upgrading their Powerlogix software from a previous version. The current v1.3 installer does not automatically disable the previous extension (use the extensions manager to do so manually).
|
|
| The next page describes installation of this upgrade. Or you may use the links below to jump to a specific page. | |
|
|
|
|
Intro | Benchmarks | Appl. Tests | Software Controls | Installation | Specs/Design | Summary - or - |
|
|
Copyright © Mike, 2000. Users of the web site must read and are bound by the terms and conditions of use. |