Results of Overclocking the Memory Bus
By now, it should have been noticed from the PCI overclocking charts that increasing the bus speed is most effective with current Socket 7 processors, but is that also true of Slot 1? The following charts illustrate that this is actually not the case! Due to the speed of the L2 cache being tied to the processor speed, it is the difference in processor speed alone that determines the performance improvement. Note that when the AMD K6-3 (or whatever they end up calling it) is released, the performance at equivalent processor speeds should be even better than the Pentium II due to the full speed L2 cache.
|AMD K6-2||Winstone98 Business (Win95)||Winstone98 Business (WinNT)||High-End Winstone (WinNT)||Quake 1.06 (DOS 7)|
|2.0×100 (200MHz)||18.8||18.9||18.0||13.1 fps|
|2.5×100 (200MHz)||20.4||20.3||19.7||14.0 fps|
|3.0×100 (300MHz)||21.8||21.6||20.6||14.7 fps|
|3.5×100 (350MHz)||22.8||22.5||21.3||15.3 fps|
|Intel PII (Klamath)||Winstone98 Business (Win95)||Winstone98 Business (WinNT)||High-End Winstone (WinNT)||Quake 1.06 (DOS 7)|
|2.0×100 (200MHz)||15.5||15.5||16.2||13.3 fps|
|2.5×100 (250MHz)||20.6||19.6||19.0||14.7 fps|
|3.0×100 (300MHz)||22.5||21.9||21.2||16.6 fps|
Once again, tests performed with the Classic Pentium were not published, but this processor gained the most from the increase in bus speed vs. just the CPU speed. This is again a result of where the data is coming from. With the smaller L1 cache, and with the L2 cache being tied to the memory bus speed, increasing the speed of data transfer across the memory bus provides a very noticeable improvement. At the other end of the scale is the P-II processor. With the L2 cache running at ½ processor speed, there is almost no difference between 3×100 and 4.5×66 settings. The small difference that exists is due to the faster data transfer from system RAM after a cache miss when the system bus is running at 100MHz. Of particular interest is the poor performance of the PII at 2.0x100MHz, showing that the K6-2 core is much faster than the Pentium II core This can also be seen on the CPU overclocking page at 2.0x66MHz, where both are running the L2 cache at the same speed.
One significant issue to note is that once the processors were pushed beyond their capabilities, Winstone experienced numerous crashes and timeouts. In some cases, after several retries the system would hang and have to be reset, rather than shut down ‘gracefully’. For example, at 5.5×66, the AMD processor would run most of the Winstone tests, but could never complete the entire Win95 business suite. It is very probable that even pushing the K6-2 processor as much as 15% will cause system and data reliability problems, so be aware of the limitations of each processor you are overclocking.
Be the first to discuss this article!