tauro wrote on 2023-03-05, 07:57:

If you have any ideas about what to measure/replace/do, please share them.

CuMine is less power hungry, then mendocino.
Electrolytic capacitors need replacing, I think.

shevalier wrote on 2023-03-05, 08:12:

tauro wrote on 2023-03-05, 07:57:

If you have any ideas about what to measure/replace/do, please share them.

CuMine is less power hungry, then mendocino.
Electrolytic capacitors need replacing, I think.

You're right, they draw more current and run hotter.

I already checked all the capacitors a week ago and I don't think they can deteriorate that fast. But a different component could have been damaged because of heat, or maybe all the AGP cards I tested put some strain on a transistor. I measured 2v. on a MOSFET with a Mendocino inserted, so it probably gets the right voltage. I'll check that and confirm if it's the case.

The Celeron 566 requires only 1.5v to work, so it makes sense it could be power-related. I wouldn't know how to proceed if in fact it is, other than replacing the transistors. The bad part is that everything looks fine, so it's nothing obvious.

for power issues oscilloscope is indispensable, even slow low bandwidth one will be enough to see bad ripple

Did you measure only capacitance or ESR too?
20 year old capacitors can be with normal capacity and esr in a around one or two OHm.
New cheap polymer capacitor - 10-30mΩ
Samwha FB, Capxon PS, Lelon OCR - cheap, but good enough for light load.
Ideal for replacement on older hardware.

The CPU seems to be getting the right voltage. And yes, this indeed seems to be a power issue.

1.7v stops at error D9
1.65v P3 boots (according to the POST card) but I get no video.
1.5v Cel same values from the POST card, but this time I get video.

For reference:
2A - Initialize video
51 - Posts
85 - "CMOS settings wrong/enter SETUP"

shevalier wrote on 2023-03-05, 09:06:

Did you measure only capacitance or ESR too?

Both. They were within 10% of the capacitance value and ≤0.1Ω ESR.

rasz_pl wrote on 2023-03-05, 09:03:

for power issues oscilloscope is indispensable, even slow low bandwidth one will be enough to see bad ripple

I'll have to look into it. Where could it be coming from and why it appeared so suddenly and not gradually?

tauro wrote on 2023-03-05, 10:08:

and ≤0.1Ω ESR.

Too much,
UNITED CHEMI-CON KZE 1500u*6.3 (as on my Aopen Mx3S s370) has 0.023 Ohm (23 mOhm) according to the datasheet.
https://www.newark.com/united-chemi-con/ekze6 … itor/dp/23K4005

tauro wrote on 2023-03-05, 10:08:

Where could it be coming from and why it appeared so suddenly and not gradually?

20+ year old stuff...

tauro wrote on 2023-03-05, 10:08:

The CPU seems to be getting the right voltage. And yes, this indeed seems to be a power issue.
I'll have to look into it. Where could it be coming from and why it appeared so suddenly and not gradually?

Most likely some other smd cap got cracked while you wrestled with mounting radiator, small caps are critical for filtering away high frequency interferences. Doesnt have to be on main cpu power line, can be Vtt.
or just some tired electrolytic caps finally dying