To me, the symptoms you describe just SCREAM "replace the caps". Even if they look okay, its been long enough that some are going out of tolerance and causing issues. When you "stress" them by adding full banks of ram, they just cant cope.

Get some nice, high quality caps and goto town. If you cant/wont do it yourself, there are members here that can do it for you.

I would add that it could also be an issue with cooling.

Those 1MB chips produce alot of heat and your cooling dosnt look very good, even that stock cooler with a fan is only just enough, the one without for sure isnt.

I would try two 256kb cache CPUs with better cooling, add a fan and make sure you use good cooling paste. The caps could still be a part of the issue but I always like to try the easy before the hard...

luckybob wrote:

To me, the symptoms you describe just SCREAM "replace the caps". Even if they look okay, its been long enough that some are going out of tolerance and causing issues. When you "stress" them by adding full banks of ram, they just cant cope.

Get some nice, high quality caps and goto town. If you cant/wont do it yourself, there are members here that can do it for you.

Quoted for truth. Brand new capacitors have a typical shelf-life of 15 years before the values begin to drift. You have used capacitors which are probably about twenty years old, now?

You could try running a single CPU, and see if the reduced stress allows it to stay at 200MHz--this could be an indication that the caps are having a hard time. But keep in mind, if you are using a power-supply of the same vintage that the capacitors could use to be replaced in that as well.

Well if he has "I also checked all the caps with a scope and they seem to be fine." I would not be totally sure its the caps. Are those Sanyo caps? they do not fail often.

I have an issue with my Intel PR440FX, Windows 2K is running slow like a mole ass if I install more than 2x 128MB memory, with 256 MB the system feels speedy. There are no memory errors in HCI memory test even when I fill all 4 slots so I think the issue is related to the different types of memory chips and general layout of the modules. In NT4 I see no performance degradation when filling all slots. Sometimes stuff just dosnt like to work together in a speedy fashion. Edit I just found a third 128MB module with similar layout and now Windows 2000 works fine with 384MB without performance degradation.

The issue with the one CPU not always getting detected do sound like a typical cap issue, but as you only have a fan on one of the CPUs I would not rule out heat issues. I had a similar issue with my PR440FX and it was because a heatspreader on one of the CPUs was extremly uneven and did hardly make any contact with the heat sink at all. From a cold boot both CPUs would be detected and the system would run for a while before freezing, after a reset only one CPU would be detected.

Yes, they are Sanyo caps, I understand the heat issue but this problem with the wrong speed happens randomly even from a cold start, keep in mind the speed selection is done with a jumper, I don't understand how that can be overridden even if the caps were bad. The caps don't even feel hot when running for a while. I could try changing them since it won't hurt to do so, I do have some high quality Rubicon, I just don't think it's going to resolve the problem, the power supply I am using works great there are no cap problems, plus I've tried two different PSU. I also don't know what would cause hieroglyphics by filling the last two banks, it only happens after it goes past the post screen not during which is odd. Also it freezing sometimes and not going past the post screen, interesting though I can go into the bios regardless of how long it sits there.

Does anyone know if what the two EZ1085CT regulators are doing is correct for this circuit? 1.5v output instead of what I assume should be 3.3v?

I believe the +1.5V that the EZ1085's is correct for the Vtt GTL termination voltage that the PPro's bus uses.. The CPU VRM's are the group of 4 SMT FETS plus the toroid laid flat down on the PCB, it has one VRM per CPU, one between the keyboard/mouse PS2 connectors and the top CPU socket, the other is the same set of parts between the front edge of the board and the second CPU socket. I've got the single CPU version of this board, i've noticed the CPU VRM get's pretty hot with a 512K cache CPU, and it has the same topology as your board's VRM's.

Thanks gg1978, that information is helpful. Do you have any ideas to what I could test or replace that might fix the problems I am having?

Could it be that the RAM SIMM's are set to the wrong voltage?? If you populate 6 x 128M SIMM's for 768M RAM, i was thinking the 128M SIMM's i've run across have been 3.3V, while most SIMM's are 5V. Turns out i was mistaken about my board, it's an Iwill P6AN, it has a weaker VRM than your's so your 1M cache chips are probably fine, though you've said you've tried a different pair of CPU's with the same results, so i doubt it's that.. I find it interesting that you have just a single jumper to do all the FSB and multiplier selections, makes it easier i guess. You might try downclocking to 180 or 166, PPro's are unlocked. Might also try some older SIMM's like 64M or smaller, those are much more likely to be 5V. I've looked at the board, i haven't noticed a set of jumpers that might set your RAM voltage down to 3.3V from the standard 5V. If you have the manual available, see if it is in there..