Are you running on a 60MHz bus? You'd probably be safer using the Am486DX2-66 16BGC as that one is .35micron and is basically the same as the am5x86 except for having a 2x/3x selectable multiplier instead of 3x/4x. I recommend using 3x to put less stress on the motherboard unless you are sure it can handle 60/66MHz.

Feipao has a 16BGC chip running at 2x66 at 133MHz without trouble on a Biostar PCI motherboard.

mmm, its very hard to explain because i just did everything using trial and error, its weird because i set the mobo to 33mhz, but i tweaked two jumpers that work together that allow to switch the mobo's speed between from 25-33mhz to 40-50mhz, and opened a jumper labeled am2/3x... weird

I used to have one AMI 486 PCI board and one of the AM486DX2-66. The board with this particular cpu combination would allow the change of multi from 2x to 3x (jumper was actually for enabling the WB cache on WB cache supported cpus - weird right?). With FSB at 40MHz it would run at 120MHz without a problem. This was verified with benchmarks. Only this CPU could change multi. I tried with other DX2's and they stayed at 2x.

thats interesting, also i tried something like that , but when i set directly to 40mhz fsb 3x multi weird results happen, its supposed to be 120mhz but falls down to 60mhz (?!), thats because i cannot access from 33 to 40mhz directly, i have to do a weird combination betwen two jumpers (jp19-17) to get the fsb to 40mhz then 123mhz, and sometimes just wont boot when the voltage is at 3v, i hope that wont kill the mobo because it seems stable but not fast as 100mhz with 33mhz with correct jumpers..weird

if someone is interested heres some info of the mobo : http://www.uktsupport.co.uk/acer/mb/aca1g4.htm

There are many different kinds of DX2s. Some of them (I'm looking at you AMD) may have slightly weird pinouts which may explain the odd results when playing with the multiplier and cache jumpers. My impression about amd DX2 chips is that all of the 3.3V models should have the 2/3X multiplier. I don't believe the 5V models do, but I have been proven wrong in the past by weird things like 5V 486s with writeback cache.