Yes, there were <i>some</i> PCI cards running at 25 MHz and they were being phased out. Care was taken so that the PCI runs at 25 or 33 MHz which are standard PCI frequencies (the FSB was upped to 33 from 25 in 486 times). The key device on the issue is the PCI-IDE controller, which resides in the PIIX4 Southbridge 82371EB of the 440BX chipset, which natively supports 33 MHz operation, so 25 MHz PCI was just for legacy support. The BX chipset was made to run with 100 MHz FSB, <i>and</i> had to be backward compatible with the LX/EX chipsets. So the PCI speed was by default 33 Mhz and not 25 MHz. This speed could not be overclocked to beyond 38 MHz which most overclockers have succeded but faster than that is extremely rare. The culprit is the IDE controller, which would run fine when undersped at 25 MHz or its rated speed 33 MHz but wont run much beyond that. Probabely this has to be attributed to the inability of the IDE interface which most disks implement that cannot transfer data that fast. Its actually a limitation in there!
I guess those who have succesfully run their IDE controllers at 40 or more MHz had better hard drives that supported this overspeeding, or just plain lucky!
Actually the BX chipset was never intended to run at speeds higher than 100 MHz, so the AGP divider was probabely skipped. Yet, even at 89 MHz, thats overclocking by about 30% modern cards are fine. But not the older AGP2X cards which in fact were rarely even exposed to such stresses in the reign of the BX.
girish
<font color=red>No system is fool-proof. Fools are Ingenious!</font color=red>