Update: According to the official press release today (3/1/99) on the PF 466 ZIF, the production models will have 233MHz rated Cache (466/233), not the faster IBM cache on the review sample (a 466/311). Although cache speed is not the major factor in performance, I wanted to make that clear.

PowerLogix is first to ship with the fastest G3 CPU speed ever, a copper-based 466MHz rated IBM CPU model with 1MB of fast IBM backside cache. Since everything but the CPU of this model is basically identical to the PowerForce G3/400 ZIF, this review focuses on the performance of the new model.

If you're new to CPU upgrades, or have not owned a pre-G3 Apple or Mac clone, the ZIF upgrades for Apple's G3 Macintosh models are generally free of the compatibility and bus speed issues that often complicated older Mac CPU upgrades (see the FAQ and CPU Reviews page for more details).

Installing the Zero Insertion Force (ZIF) socket upgrade is accomplished in a few minutes. Since bus speed is fixed by the Apple jumper block (G3-ZONE has details), there is only one adjustment setting needed - System Bus to CPU ratio. Select the proper ratio as shown in the PowerLogix manual according to your G3 card model and system. The different bus speeds of the Apple G3 systems require different CPU/Bus ratio settings.

For example, the beige/AIO G3 has a 66Mhz system bus, the B&W G3 has a 100MHz bus. This means that B&W G3 CPU speeds must be a multiple of 50Mhz, so the closest setting to the rated CPU speed would be 450MHz (a 4.5:1 Ratio). The 66MHz bus speed of the beige/AIO G3 allows selecting the rated 466MHz CPU speed (a 7:1 ratio). As readers (and I) demand, I experimented with higher speeds in both the beige and B&W G3 with great success on this particular sample.

Note: Although speeds shown here were 100% reliable during the product loan duration, I can't recommend overclocking to anyone else. The Powerlogix manual does not endorse or suggest overclocking the card, however my understanding is they honor the warranty on their products regardless of speeds used. Based on last summer's G3 CPU Card Survey, the current list of overclocking reports in the Rate Your G3 Card searchable database, and tens of thousands of mails over the last two years, I know that most owners of adjustable speed cards are risk takers and run them beyond the rated speeds.

For a guide to installing a ZIF upgrade, see the installation/docs page my PowerForce G3/400 ZIF review.

The PowerForce G3 466 ZIF contains the fastest PowerPC to date - the IBM 466MHz G3. The white rotary switch in the lower right corner is used to select the bus speed multiplier (ratio) that determines the CPU speed. The advantage to having this adjustment is twofold; it allows the upgrade to work at the proper CPU speed in both beige G3 and B&W G3s (different bus speeds, so different ratios required) and it also allows experimenting with higher CPU speeds (potentially risky, but rewarding in this case).

Performance Tests:

MacBench 5.0

All tests were done at 1024x768, thousands color graphics mode (note that the reference graphics score in MB 5 was done at 1152x870, millions colors). Disk cache size for the beige G3 was 4MB, for the B&W it was 6MB. Virtual Memory was off, full extension sets were used (not trimmed). No Speed Doubler or Libmoto active.

Later I was able to run tests in my B&W G3 at speeds of 550/275:

(Graphics score bar trimmed to fit image space )

Application Tests:

The following common/popular applications are primarily CPU/FPU speed bound but were run since I have a large series of comparison scores from the G3 Apps Tests page. These results show the total time to complete renderings of sample scenes, etc.. As shown in the results, CPU speed (not bus speed) is the primary performance factor, followed by Cache size and speed to a much lower extent. See the CPU Card Reviews, Yosemite and G3-ZONE pages for more info on the other cards/systems listed.

Photoshop 5 PSBench Results:

The following is a total time summary taken from the Photoshop 5 Performance page which lists all the details of the 21 filter PSBench test.

Test scenes are from the Sample Scenes folder, Kai subfolder (I used these std scenes to allow other Bryce2 owners to compare their results with mine). Default 16980k memory allocated to Bryce2. Antialiasing was enabled for all tests.

The Video card or driver/software has to be already maxed out with a 400Mhz G3 based on these results. Literally no gain from a 150MHz faster CPU. I'm not sure why the scores were lower than obtained with beta rage128 drivers on G3/400 as shown on the Yosemite Game Performance page. The Rage128 driver update and adding an IDE Slave drive is the only thing that has changed since the tests were run.

I didn't notice the pauses in outdoor areas, even at 1024x768 mode that I'd seen before with the Rage 128 beta drivers, at least in limited playtests (I loaded a saved game in the outside area where I'd seen the consistent pauses before). Possibly the Rage128 update fixed the problem, but not according to reader feedback.

Review Sample Hardware Specs:

• IBM 466MHz PowerPC 750 (Copper)
• 1MB L2 Backside Cache (2x IBM 043641WLAB 01L3242 3 Srams)
  Note: Production Versions to have 233MHz rated Cache
• Selectable Bus/CPU Ratios of 3:1 to 8:1 in .5x steps
• Fits: Apple beige G3/B&W G3/All-in-One G3
• Max Speedmeter Temp Seen: 27 degrees C
• Maximum Speeds Reached: 550/275
  (At 550MHz CPU speed, 2:1 cache ratio was max)

With the current design, there was just enough clearance in the B&W G3 to fit the ZIF in the socket with the ATI Rage128 card installed. Adjustments might not be possible with the ZIF in place due to the close fit between the rotary switch and the bottom edge of the ATI Rage128 card.

Summary:

As expected, performance was record setting in every category except Games, which showed little to no increase over a 400MHz G3 CPU. Apple G3 owners can now drop in an upgrade that will make their machines faster than any Macintosh ever made, in fact, probably faster than any personal computer ever made to date.

Pricing/Availability:

At the date of this review (Feb 1999), suggested retail pricing was $1399, but prices may be a bit lower at Bottom Line, Other World Computing or MacGurus. PowerLogix is expected to be shipping on March 1, 1999 I'm told (updated).

Test Systems:
The base systems used for this review were an Apple G3/300 and B&W G3/350/DVD with the standard array of extensions (no trimming done). System details are listed below.

System Hardware Summary:

• Apple PowerMac G3/300 (Rev 2):
• 160MB RAM (stock 32MB DIMM, one 128MB DIMM)
• Stock 6GB IDE Hard Disk (70% full, not defragmented)
• IDE/ATAPI ZIP drive
• Onboard ATI RagePro video chip with 6MB SGRAM
• Game Wizard 8MB Voodoo2 card (only PCI card installed)
• OS 8.1, 4MB Disk Cache, VM off, QT 2.5, QD3D 1.5.3, PC exchange, etc. [No Libmoto]

• Apple B&W G3/350/DVD (Yosemite):
• 192MB RAM (stock 64MB DIMM, one 128MB DIMM)
• Stock 6GB IDE Hard Disk (40% full, not defragmented)
• Maxtor 10GB IDE Slave Hard Disk (70% full, not defragmented)
• DVD/CDrom drive (DVD option)
• OEM IDE/ATAPI ZIP drive (slave on CDrom IDE channel)
• Stock ATI Rage128 card with 16MB SDRAM (driver update 1.0)
• Game Wizard 12MB Voodoo2 card (only PCI card installed)
• OS 8.5.1, 6MB Disk Cache, VM off, QT 3.0, QD3D 1.5.4, etc. [No Libmoto]