Benchmark tests were run in MacBench 4.0 and the Motorola compiled version of BYTEMARK. As stated previously, I list CPU and Cache (in Mhz) in the format of CPU Speed/Cache speed, so 240/120 would mean a CPU speed of 240 Mhz with a backside cache speed of 120 Mhz. For reference, the maximum bus speed I ran was 41.6 Mhz (the card maximum is 45 Mhz). Without getting bogged down in all the math, ratios and switch settings let me just say that I kept bus speeds low to minimize any potential problems with the motherboard cache in the PowerTower Pro.

I could probably have gone faster, but I was working without any switch setting tables and in the end, I think you'll see that the results I obtained leave little to complain about in either benchmark or applications performance. Owners of the card will have far more time to explore the limits than I did. There are a lot of combinations possible. I'm fairly certain that many owners will report better performance with combinations I did not try. I just caution them to ensure that it is reliable.

One thing I have discovered - ByteMark is a extremely good tool to expose a CPU running at it's limits. I saw some combinations that would run overnight fine, but could not complete one run with ByteMark without an error. The fact it runs a fairly long, extensive set of math operations apparently raises the CPU temperature more than any standard application. For instance one setting would pass a 57 minute Infini-D rendering test but would fail the ByteMark test 1/2 way through. Keep that in mind when you're verifying your maximum speeds.

MacBench and BYTEMARK tests were run at several combinations found to be reliable in two days of extensive use. Remember that each card and system has some tolerance variation, so these speeds may not be attainable in every card or system. The specific CPU chip on each card and your motherboard components and RAM can affect maximum reliable speeds. For this reason I've listed MacBench scores for several speeds that were reliable with this particular card/system combination.

Why didn't I run the card at the stock 220/110 speed?
I used 240/120 as the low end for two reasons - if you buy a adjustable speed card I doubt you'll run it at the rated speed and if the card won't run reliably at 10% over the rated speed there is something wrong in the components and/or design. Some may complain about this but the site is not called "Accelerate Your Mac!" because I settle for the status quo.

The MacBench graph below shows how the card scored at 3 different speeds and shows its performance relative to a twice as expensive model with a faster rated CPU and cache. As the comparison shows, the PowerForce 220/110 provides a lot of bang for the buck.

The best speed I obtained with the card was assigned a 100% rating, and the other cards/settings are normalized to that figure for comparison.

As the scores above show - the best performance of the settings I tried was obtained at a CPU speed of 291.5 Mhz with a 116.6 Mhz cache speed (2.5:1 ratio). It is possible that a higher cache speed and/or bus speed would be reliable and provide better performance, but deadlines, a 24 hour burn in verification of each setting and a review backlog prevented me from further exploration. Since my tests show that CPU speed is paramount to performance, I concentrated on that. The fact the cache topped out at 140 Mhz or so limited things a bit, but you've got to admit the performance is impressive for a $500 card.

Scores with and w/o Speed Doubler 8.1

In previous OS versions Speed Doubler was said to boost scores by a few percent. Evidently the OS 8.1 code executed by the test is 100% native. I'd still not give up SD 8.1 for faster finder copies, non-native code boosts and more.

Scores with and without Motherboard 1 MB cache Dimm

Although MacBench showed some benefit to having the 1 MB cache installed, both Bryce 2 rendering tests finished in exactly the same time. The Infini-D test was slightly faster with the 1 MB cache, but percentage wise the boost was small - less than one minute on over an hour of total time ( 1:06:24 vs 1:07:18).

I ran the Motorola complied version of BYTEMARK test at the fastest speed I attained (291.5/116.6). Since I just started using BYTEMARK, I currently only have one other CPU card to compare it with, the XLR8 G3 266/177 card at various speeds (as reported in the XLR8 card review):

Keep in mind I have to rate benchmark performance relative to all cards I've tested (not just bang for the buck) so I rated Benchmark performance a 8. (This card will make up for it in value)

The next page has comments and some results from real world applications. Or you may use the links below to jump to a specific page.