The latest CPU upgrades take advantage of the 10X bus multiplier on the most recent revision of the PowerPC G3 chips.  Previous revisions of the G3 chips were limited to a bus multiplier of 8X.  In short the bus multiplier dictates how fast a given chip may be run with respect to its system bus.  For instance, if the system bus (the speed at which the motherboard runs) on your machine is 50 Mhz like on the Apple Power Macintosh 7300/166, 7300/200, 7500, 8500/150, 8600, 9500/150, 9500/200 and 9600, and you are running a CPU chip with a maximum bus multiplier of 8X, then the fastest that your system is rated to run reliably at is 400 Mhz.  (Of course you can tweak things a little bit with some systems and CPU upgrades from other manufacturers, but the previous rated speed in terms of Mhz for these machines with a G3 upgrade and a 50 Mhz bus was 400 Mhz).  With the latest revision of the G3 chip, the bus speed multiplier has been pushed to 10X giving a theoretical reliable limit of 500 Mhz.

For those of you that read my previous review of Newertech's G3 CPU cards, you will already be familiar with their reputation for quality that Newertech puts into their products.  The timing issues (speculative processing introduced with the G3) that plagued G3 upgrades with compatibility problems in the past have been taken care of by most manufacturers of G3 card upgrades, however the methods for resolving the speculative processing conundrum differ.  Most manufacturers of G3 CPU card upgrades take care of the bus timing incompatibility with a fix in a software extension that loads prior to all other extensions.  Newer has placed the fix into the hardware so that it is active all of the time, not just when the extension loads.  Granted, the software fix appears to work with few problems if any at all, and there are manufacturers that are producing great CPU card upgrade products like Powerlogix.  I just happen to prefer the fix being in hardware and working all of the time rather than waiting for the fix to load.  For those of you that are trying to decide which upgrade card to buy, there is one caveat with buying a Newer card that some will object to.  Newer does not support overclocking of their cards.  There are no adjustments for the core CPU speed available as on other manufacturers designs.  What you buy in terms of rated CPU speed is what you get, however there are some software adjustments that can be made with the Newertech control panels.  More on this later in the review.  The advantage of buying a Newer card is that they put much effort into designing a card that will have no conflicts, and they typically are the fastest cards on the block (without overclocking)  Speaking of overclocking, there are risks involved and possible long term reliability issues with data corruption that are not well known.
 

Review Tests/Ratings: This CPU card review is broken down into the following categories:

1. BenchMark Performance: MacBench 5.0. Other unaccelerated system scores are shown for comparison.
2. Applications Performance: How the card did in several CPU intensive tests like Photoshop, Bryce 2, JMP etc...
3. Software/Controls: Ease of use and features of the supplied software and hardware controls.
4. Documentation: How clear and complete the installation and setup instructions are in the supplied manual.
5. Specifications: Features and details on the hardware design. Includes compatibility information.
6. Summary: Final comments,  and pricing/availability.

Test System:

The test machine for the CPU card upgrades was a 9600/300 with 512 MB of RAM, and dual 9 GB Quantum Atlas 10k Ultra 160 SCSI hard drives. The hard drives are driven by an Initio Ultra 2 Wide SCSI accelerator in slot 1.  In addition, the machine has dual monitors, and a 3DFx Voodoo II card. The graphics card used for testing was the iXMicro Ultimate REZ with 8 MB of VRAM in slot 6 hooked up to an Applevision 750.  Slot 5 contained a iXMicro Twin Turbo with 8MB of VRAM.  Slot 4 contained the 12 MB Voodoo II card from Micro Conversions.  On both machines, I did not run with minimal extension sets as I wanted to know what the real world performance of these cards were, additionally, some of the applications that I was going to test included extensions of their own that were required in order for the application to run. I also did not run tests with Libmoto math library extension.  Newer includes the Libmoto math libraries with their cards as an optional part of the installation process.  The benefit of Libmoto is increased floating point benchmarks,  however there is no real world benefit from using it and with System 8.6 there is really no need as there are improved math libraries in System 8.6.  Additionally several applications conflict with the Libmoto extension such as Wavefunction's MacSpartan and MacSpartan plus molecular modeling programs and the game Hornet 3.0.  Most other manufactures also include Libmoto with their upgrades, so as to give a perceived "significant" speed up in floating point benchmarks.  So the inclusion of Libmoto is not unique to Newertech.
 

System Hardware Summary:

Apple PowerMac 9600/300
512 MB RAM (two stock 32 MB, 2 additional 32 MB DIMMS, and 6 64 MB DIMMS.
Dual 9GB Quantum Atlas 10k Hard Disks (40% full and defragmented)
Ultimate Rez graphics card set to 1152X870 in millions of colors.
OS 8.6 disk cache 8160 MB, VM off.