In the weeks since this evaluation was posted, a number of visitors have emailed and asked how I was able to achieve a 133MHz FSB setting, and how to enable the SuperBypass feature of the chipset. Those who asked about the 133MHz FSB indicated that because AMD did not implement an independent memory/host bus this shouldn’t be possible. To be quite frank, I had simply assumed that Soyo had worked with AMD to enable this feature without checking further. As for the SuperBypass feature, that should have been easily found in the BIOS, however…
Recently, I spoke with Soyo on this matter and found that the BIOS program on the motherboard I received was not intended for public consumption. The BIOS provided by Soyo for this motherboard has the 133MHz Host Bus option grayed out, so the user cannot select it. The person I spoke with was also quite surprised that my benchmarks scored higher with this setting, since it theoretically does not work on the AMD chipset. In addition, the BIOS that Soyo had available for this motherboard at the time I spoke did not allow SuperBypass to be enabled and it was disabled by default. The spokesperson indicated that they are in the process of testing a BIOS update that will allow SuperBypass to be enabled or disabled by the user via the BIOS.
Thinking that I must have hallucinated the benchmark results, I decided to run additional tests and either prove or disprove my initial results. The first test was using Content Creation 2000 under Windows 2000. The following table shows the results of the first three runs at each bus speed:
|Host Bus = >||100MHz||133MHz||Default|
The first three runs were with the 133MHz FSB setting (24.7, 24.9 and 24.5), followed by the 100MHz FSB setting (24.0, 23.9 and 24.7). Until the last run of the 100MHz setting, the results paralleled what I had seen during my initial tests, however the last run at 100MHz showed a result that was about the same as the 133MHz bus setting. I then ran three more tests using the ‘Default’ setting (which would be 100MHz in this case), and was even more surprised when the last two runs scored higher than any of the previous seven tests. In fact, the difference between the worst score and best was just short of 10% (23.1 vs 25.3) – far beyond the margin of error advertized by Ziff-Davis for their benchmark.
I then decided to run the Membench benchmark from Intelligent Firmware, and found that these also varied widely. Since every other motherboard I have tested with gave exactly the same results between test runs using Membench (assuming nothing changed between runs), I must conclude that for some unknown reason the motherboard itself is responsible for the variations. I ran both CC2000 and Membench at least 10 times each, and rarely got the same results twice. The variation with Membench was in the write and transfer tests (read test results were always the same).
What this means is that:
- The 133MHz FSB setting does not provide any benefit at all
- Users cannot enable the 133MHz FSB setting in any case, since the BIOS I have is not available to the public
- SuperBypass is not an option currently, but will be offered in a soon-to-be-released BIOS update.
To all of those readers who were confused or convinced I was on drugs, I must offer my sincere apologies for not investigating this issue more completely for the original review and also give my thanks to those who contacted me to find out what I had been taking while doing the evaluation, allowing me to follow up. This does not, in my opinion, lessen the quality or stability of this board for the professional who would not be attempting to play with non-default settings.
Be the first to discuss this article!