Maybe, the deterioration of the aluminum electrolytic capacitors due to aging.
Perhaps, the capacitance it still has is barely okay if the current draw is low, but not okay if it the current draw is higher.
The higher current draw is higher if the FSB is higher, so maybe this very simplistic explanation can explain the situation.

bracecomputerlab wrote on 2025-09-12, 08:00:

Maybe, the deterioration of the aluminum electrolytic capacitors due to aging.
Perhaps, the capacitance it still has is barely okay if the current draw is low, but not okay if it the current draw is higher.
The higher current draw is higher if the FSB is higher, so maybe this very simplistic explanation can explain the situation.

Capacitor failing not only affects to the capacitance negatively, but ESR also shoots up way out of spec. For VRM the capacitance can be often less than what is originally installed, but too high ESR means high ripple which definitely causes instability to circuit.

Agree that caps could be causing issues here - but back to basics first: you're getting memory issues. What memory are you using? And how is the memory clock configured (same as FSB or same as AGP?)

If the board's otherwise okay I'd suspect memory that's not handling 100mhz for some reason. Settings or just marginal parts.

I did run into boards even when SS7 was current that were flaky at 100mhz. Although that'd usually eventually show up in an errata sheet for them and the Tomato is the only one that sticks out in my memory.

I tried other memory modules: PC-100, PC-133, 128MB, 64MB, and the problem persists, multiple memory errors.
- It doesn't matter if the memory is running at AGP or FSB speeds, there's no noticeable difference.
- The capacitors are visually fine, not the slightest bulge, not scratches, cracks, or dents on the board.
- The motherboard worked fine with the K6-2 300 (66x4.5), MemTest86+ was correct.
- In all cases, the CMOS battery was absent, so it can't be a bad BIOS configuration either.

The monster must be there, somewhere, but I can't find it.
I'm going to go back to FSB 66 and install WinXP to test the system thoroughly.

In the next few days, a friend will lend me an identical motherboard for testing, so I'll keep you posted.

Elegantly said than my explanation.
I learned something from it (i.e., the ESR part).

CharlieFoxtrot wrote on 2025-09-12, 09:33:

bracecomputerlab wrote on 2025-09-12, 08:00:

Maybe, the deterioration of the aluminum electrolytic capacitors due to aging.
Perhaps, the capacitance it still has is barely okay if the current draw is low, but not okay if it the current draw is higher.
The higher current draw is higher if the FSB is higher, so maybe this very simplistic explanation can explain the situation.

Capacitor failing not only affects to the capacitance negatively, but ESR also shoots up way out of spec. For VRM the capacitance can be often less than what is originally installed, but too high ESR means high ripple which definitely causes instability to circuit.

As a possibility, even if the caps look okay from outside, they may have lost their capacitance due to dry up.
If I understand it, the moisture inside the can gradually escapes from the rubber seal below, and loses the functionality of a capacitor overtime.
Just one possible explanation.

Feanor_twh wrote on 2025-09-12, 14:44:

I tried other memory modules: PC-100, PC-133, 128MB, 64MB, and the problem persists, multiple memory errors. - It doesn't matter […]
Show full quote

I tried other memory modules: PC-100, PC-133, 128MB, 64MB, and the problem persists, multiple memory errors.
- It doesn't matter if the memory is running at AGP or FSB speeds, there's no noticeable difference.
- The capacitors are visually fine, not the slightest bulge, not scratches, cracks, or dents on the board.
- The motherboard worked fine with the K6-2 300 (66x4.5), MemTest86+ was correct.
- In all cases, the CMOS battery was absent, so it can't be a bad BIOS configuration either.

The monster must be there, somewhere, but I can't find it.
I'm going to go back to FSB 66 and install WinXP to test the system thoroughly.

In the next few days, a friend will lend me an identical motherboard for testing, so I'll keep you posted.

They've already given me the board for testing. It's the same 591P, but this time it's v3.1 while mine is v2.0.

I'm going to run a systematic round of tests on both boards and post the results.

591P 2.0:
- FSB 66: both processors, MemTest86+ RAM OK
- FSB 100: K6-III RAM failures, K6-2 restarts on MemTest86+
591P 3.1:
- FSB 66: both processors, MemTest86+ RAM OK
- FSB 100: both processors, MemTest86+ RAM OK

All tests performed with the same single-sided PC133 64MB RAM module.

I've learned that a PC100 RAM module can work fine at 66MHz and fail at 100MHz.

My motherboard still isn't working properly; perhaps it's what you're saying about the capacitors, as further RAM tests seem unnecessary...

Ah... with the PC100 RAM module I was initially using, both boards failed in the same way. That made me look for another module, even though it worked without any problems at 66MHz

If one PC100 module fails on both boards, it's probably a dud module - but if PC100 fails at 100MHz on one board but works fine on the other, I'd say your Vdimm voltages are probably out of spec and/or there's a lot of ripple on that line.