I think you're putting a little bit too much thought into it. Not a bad thing. 😀

20 years of progress means our "regular" caps are effectively equal to the high end caps of yore. Low ESR is more than plenty for the DC/DC converters. Now, if it was me, i'd get solid aluminum caps. ^.^ I'd also replace every electrolytic on the board. I mean, mouser/digikey have minimum orders, so you might as well, right?

Agree. With 14 - 1000 and 4 or 5 - 1500 is actually not expensive at all. Think same Sanyo's were on the P6Np5 and they have a very long original life compared to others...

those sanyo cap is good brand and in good shape so i don't think there are any reason to replace them in the first place.
I only replace the cap if they are
- having the sign of being bulged
- the brand is known to be bad like cheapo or teapo

They're 30yo. Might as well replace them before they go spectacularly.
And on a relatively primitive early dc-dc supply, that likely measures acceptable tolerances around 10-15%. On 3.2v thats 0.32-0.48v!
I'd just replace them with any half decent modern cap.

nhattu1986 wrote on 2024-11-23, 08:45:

those sanyo cap is good brand and in good shape so i don't think there are any reason to replace them in the first place. I only […]
Show full quote

those sanyo cap is good brand and in good shape so i don't think there are any reason to replace them in the first place.
I only replace the cap if they are
- having the sign of being bulged
- the brand is known to be bad like cheapo or teapo

I was under the impression that just with age they'd have issues?

What I'm trying to track down is some instability in how the system runs....granted it could also be due to me still not having BIOS settings quite right. I install Windows 98 SE over and over again to test stability.

My observations:
1) the ISA bus speed is...slightly irregular.
My scope reads it as 7.43252 MHz...however, I can see it "twitching" occasionally. Here's a picture of a frozen frame of the slight delay that occurs occasionally.
CPU is 150 MHz. Clock speed is 60 MHz (x 2.5). Could this just be an artifact of how it generates the ISA bus CLK signal from the 60 MHz? i.e. is this normal?

2) I measured some of the tantalums on the board. It appears none of them are on the +12 or -12 lines. All are on +3.3 or +5. All are rated for 16 v, except one which is 25 V.
Seeing low resistance across some on the +5 line, IIRC around 20-30 Ohms. Measured in circuit....
One question I have is, do Tantalums migrate towards lower and lower resistance until they short? i.e. could this be a sign they are going to have issues?
Obviously the fact that it was measured in circuit means I'd have to pull one out to measure it directly.

3) The main issue I've seen is that for quite a while I could NOT get the system to get video out of ANY PCI video card I have. I thought my cards were bad (second system had the exact same issue). However, I found some more PCI cards and all of them had the same issue. After playing around with it a while I was eventually able to get ALL o the PCI video cards to output. But even after I got the card working on one system, it would still refuse to work on the second system without going through the same process of leaving it connected to the system for a while and rebooting several times...then it was happy.
This lead me to think maybe the caps needed to be replaced? I'm certainly open to other suggestions.

luckybob wrote on 2024-11-23, 03:25:

I think you're putting a little bit too much thought into it. Not a bad thing. 😀

20 years of progress means our "regular" caps are effectively equal to the high end caps of yore. Low ESR is more than plenty for the DC/DC converters. Now, if it was me, i'd get solid aluminum caps. ^.^ I'd also replace every electrolytic on the board. I mean, mouser/digikey have minimum orders, so you might as well, right?

By solid aluminum, do you mean Aluminum Organic Polymer? (still learning, pardon if it's a dump question 😀 ).

I should have stated as well, none of the caps are bulged or leaking at all that I can see.

Yup.

The Sanyo caps are of good quality, and are 95% likely to be just fine.

I don't think it's going to hurt to replace them, it sounds like you have soldering experience.

Id also just give a good physical inspection of memory slots while it's under the microscope.

Or a cold solder joint on any of the qfp chips.

Can we assume you've ran a full memory test?

Guld wrote on 2024-11-23, 15:20:

1) the ISA bus speed is...slightly irregular.
Could this just be an artifact of how it generates the ISA bus CLK signal from the 60 MHz?

Yes, ISA is not a synchronous bus, everything happens on signal edges and ignores clock for the most part.
Since you have a scope put it on VRM output, that will give you definitive cap health check answer. From my experience Asus caps are rock solid after 25 years, so maybe even 30 is fine 😀

luckybob wrote on 2024-11-23, 16:14:

Id also just give a good physical inspection of memory slots while it's under the microscope.

Or a cold solder joint on any of the qfp chips.

Can we assume you've ran a full memory test?

Am I looking for slightly dull looking solder connections? That's about the only way I know to spot a cold solder joint?

Actually, no I haven't run a memory test! What's a recommended memory test to run? I think back in the day I used to use memtest86, but not sure I have a floppy boot version laying around. And not sure I can get the system to boot from CD as I do think I have an old booting CDROM version laying around. ... just found an old floppy marked "Memtest86 3.4a"....will see if it still works!

Looks like memtest86+ now lives at, if it's the old software I remember back in the day 😁. But didn't see a floppy version, just CD.
https://www.memtest.org/archives

rasz_pl wrote on 2024-11-23, 16:46:

Guld wrote on 2024-11-23, 15:20:

1) the ISA bus speed is...slightly irregular.
Could this just be an artifact of how it generates the ISA bus CLK signal from the 60 MHz?

Yes, ISA is not a synchronous bus, everything happens on signal edges and ignores clock for the most part.
Since you have a scope put it on VRM output, that will give you definitive cap health check answer. From my experience Asus caps are rock solid after 25 years, so maybe even 30 is fine 😀

Are you talking about the VRM to generate 3.3V? I have the board set to use the ATX 3.3V so I'm not using the onboard VRM. Where would I put my scope on that VRM output? Would that be one of the CPU pins?

the Pentium pro has a split voltage setup. there is a 3.3v bus voltage and the core.

Am I looking for slightly dull looking solder connections?

A 'quick and dirty' way is to take a needle or other sharp object and gently poke each leg on the QFP packages. if one moves, that's bad. A microscope helps.

Guld wrote on 2024-11-23, 21:37:

I have the board set to use the ATX 3.3V so I'm not using the onboard VRM.

"3.3V POWER SOURCE SELECTION" oh nice, so whole VRM is skipped and those caps do nothing, no need to replace. Tantalums inside the socket are doing all the filtering. no wait, I think this is about skipping that linear reg on the right of the socket, and VRM is still powering the CPU

luckybob wrote on 2024-11-23, 22:12:

the Pentium pro has a split voltage setup. there is a 3.3v bus voltage and the core.

Am I looking for slightly dull looking solder connections?

A 'quick and dirty' way is to take a needle or other sharp object and gently poke each leg on the QFP packages. if one moves, that's bad. A microscope helps.

Sorry, dumb question, but what's a QFP? And yeah, I see what you are getting at. I don't have a microscope, but I can zoom in with other tools. Is there a particular output to focus on for this QFP chip if I can find a datasheet for it?

Quad flat pack. ^.^

https://en.m.wikipedia.org/wiki/Quad_flat_package.

You don't need datasheets, just gently see if any legs are loose.

luckybob wrote on 2024-11-24, 00:14:

Quad flat pack. ^.^

https://en.m.wikipedia.org/wiki/Quad_flat_package.

You don't need datasheets, just gently see if any legs are loose.

Ah, gotcha, QFP is not a specific chip! I missed that. Thanks!

Ran memtest 86 and got an error "Unexpected Interrupt - Halting".

That's odd....so not a memory error? But some other fault?

This smells like faulty ram or a poor simm connection...

So, I've been running some tests on the board with different combinations of memory.

I've gotten all the modules to work in the SIMM1 and SIMM2 slots, one set tested for 9 hours, the other set tested for 2 hours (hard to find runs of time to test sometimes 😀 ).

I've seen it die at 11 minutes with one combination populating all 4 slots, and die at just over 1 hr with the pairs swapped.

I used deoxit on the memory and simm slots and am testing to see if that makes any difference. I have not had a chance to try to see if any of the QFP pins are loose.

One weird thing I did notice is that in one combination, I saw that memtest 86 went past 100% (WELL past, more like 200%) for the "Pass" output. I attributed this to possibly a memtest86 bug...but maybe someone knows better?

What's a good length of time to run to establish a stable configuration? I know people like 8 hours so they can let it run overnight but I don't leave these old systems running overnight so getting longer runs in is a bit more difficult.

It can be micro fractures in tracks or parts, an even detached bonding wire inside one of ICs - all are temperature dependent, might work for hours until temp change makes it detach. You can try testing while running a hair dryer with cold/hot air at the mobo, or physically gently but firmly pressing on parts with ESD safe (wooden) stick or even your finger (ground yourself before, no fleece/wool clothes)