might be power supply on last legs, might be crushed pin in one of the slots where you were reinstalling cards
measure voltages and carefully look into ISA slots with good light

Thanks for your answer, I had a careful look on the isa slots, they looks in perfect condition, and as I tested several of them I don't think this could be the issue. I have a post card and each voltage indicator leds are turned on, but I'll try with a multimeter to be 100% sure.

Does sound ike it could be PS - be aware that "voltage" isn't the whole story.
There's also noise, caps going bad could lead to excessive noise while average voltage looks OK.

There's also the possibility of drop-outs when "high power" cards try to start, when debugging issues like this I always start with on-board video if avaiable, or a a super-simple/low-power card (like ATI VGA wonder) if not.

If you have access to a scope, watch the power rails as the system tries to start. Everythuing should be rock solid.

Also, since you've been swapping things in/out ... have you tried "factory reset" - harder with no video, but disconnect BIOS battery for a good long time - also depending on mainboard there might be a jumper.

mobo and cpu models would also be helpful

vbug wrote on 2026-03-22, 13:05:

Thanks for your answer, I had a careful look on the isa slots, they looks in perfect condition, and as I tested several of them I don't think this could be the issue. I have a post card and each voltage indicator leds are turned on, but I'll try with a multimeter to be 100% sure.

It would be enough for any two opposite pins in any ISA slot to touch to mess all other slots. If the card isn't recognized in one ISA slot, it is very expected it won't be recognized by others, unless the issue is a severed trace.. but that is unexpected in your case, as traces don't suddenly degrade and get severed by normal use.

My bet is on one of the caps giving up finally. Cap failiures have this symptom of gradual degradation until final death. A quick continuity test across the caps might be useful.

vbug wrote on 2026-03-22, 11:31:

(AMI bios, 2 short (high), 8 long (low))

2 short, 8 long? that doesn't seem to be a standard ami beep code .
I guess you don't have a post card you can stick in the machine to see which post code it get stuck on ?

Either way, I'm gonna advice you to take everything out of the case, all cards & mainboard, memory, and hook it up outside of the case one by one.
Verify the behaviour of each step, start with the mainboard without memory hooked up to the psu,
add the memory,
add the keyboard,
add the video card with monitor attached.
verify beeps & led behaviour at each step.

So I tested the voltages when turning the computer on, it seems okay, stable voltage for both 5V and 12V. I have a postcard, when the motherboard starts to beep the post code is 1-2F for 1-2sec, then it moves to 1-3B and stick to that code. I already tested to turns it on with only CPU, RAM and basic ISA video card, same result. About the caps that could be dying, do you mean on the PSU, or other caps on the motherboard ?

I tested the DALLAS battery and as it was a bit low I replaced it (so I removed it and wait for some minutes so I assume the bios has been reset). Now it stops on code 3B meaning it waits for the user to press del key (probably because the bios has been reset). But still no display

I join a PDF presenting the motherboard model, and some photos of the ISA/VLB ports

here is a video of the boot sequence (as zip file because video files are not supported as attachment !

Have you tried removing the ram and reseating it, or if possible using different memory modules?

I tried to remove all the ram sticks and put them back, I'll try with other modules on next week-end !

vbug wrote on 2026-03-22, 20:33:

I tried to remove all the ram sticks and put them back, I'll try with other modules on next week-end !

Sorry, maybe nevermind the ram suggestion. I had another look at your video and during the beeps the post code is 31 and previous one is 2F

12F	Video memory tested
230	Video adapter tested
331	Alternate video memory tested

I also see 37 39 38 98? 3B and then it settles on 41

https://theretroweb.com/motherboards/s/bcm-lx400a-p-vlb
ok so the CPU is 5V

weedeewee wrote on 2026-03-22, 21:37:

3B and then it settles on 41

yep, 41 seems to be

AMI Old BIOS (AMI Plus BIOS); 08/15/88�04/08/90:
40 Ready to disable gate A20
41 A20 line disabled successfully
42 Ready to start DMA controller test

AMI BIOS Post Codes (After April 1990):
40 Cache memory now on...Ready to disable gate A 20
41 A20 line disabled successfully
42 i486 internal cache turned on

AMI New BIOS; 02/02/91�12/12/91 & AMI New BIOS; 06/06/92-08/08/93:
3B Hit message displayed. Virtual mode memory test about to start.
40 Preparation for virtual mode test started. Going to verify from video memory.
41 Returned after verifying from display memory. Going to prepare the descriptor tables.

got excited seeing that A20 error, that would mean maybe keyboard controller is seated badly and its IO pin responsible for A20 stopped connecting reliably. No idea how AMI bios determines successful A20 disable, if it just means command to keyboard controller was received or if it actually tests ram mapping.
Another option is failed DMA controller test, that would be bad as DMA controller is inside chipset
Another option is failed CPU L1 cache test - CPU crashing
Last option is after video test, starting CPU setup and crashing

In any case Video is initialized way earlier than the post code. No video output combined with most likely 5V CPU crashing does strongly point at failing 5V from PSU (high noise/ripple) caused by dried caps. Cant really test that with a multimeter as it shows average voltage while in reality it might be jumping up and down. Requires oscilloscope to see it properly.
EEVblog #594 - How To Measure Power Supply Ripple & Noise#594 - How To Measure Power Supply Ripple & Noise https://www.youtube.com/watch?v=Edel3eduRj4

Open PSU, desolder big 5V capacitors and either
-test them using ESR meter if you have one
-just replace with similar ones. Dont have to be exact parameters as long as they are 6.3-16V rated and around couple thousand uF. Either order new or borrow from power supply of some other broken junk 😀 - just for a test. If that works you can order proper replacement caps and perform real repair, if not we will have to go deeper.

The filtering caps are on the mobo closest to the AT power connector. Those tend to fail aside from the PSU itself. But ISA has several filtering caps as well that can short out when dead..

shorted tantalums on motherboard would force PSU to shutdown or burn out some traces

So maybe the best move would be to test with another PSU ? I don't know how to test the tantalum ones on the motherboard, I already had another motherboard with a tantalum that burned, I replaced it but I was never able to get the mobo working back.

For the Bios Post codes I'm not sure but I think mine is 06/06/92-08/08/93

additional pictures focused on the tantalums

So I opened the original PSU and capacitors seems visually to be in good condition. I tried with another PSU, and with another set of RAM sticks, I got the exact same result, so I think the issue is on the motherboard itself 😒

Tantalums visual fail is quite spectacular, unlike electrolytic which might just slightly bulge out. Might be a good idea to test a few of those caps and see if there are any shorts or very low resistance. Just get your multimeter on the two legs of any of those caps in ohms mode and read the value. Try it on several caps. If one of them is shorted, chances the short will appear on others as well.. so you don't need to probe each and every cap.

wbahnassi wrote on Today, 14:42:

Tantalums visual fail is quite spectacular

not always https://youtu.be/1GgYDlVjrc8?si=3IILRirmRfFPVPCr&t=612

but shorted tantalum would make the PSU shutdown .. unless it already burned a track
vbug set multimeter to resistance mode, with turned OFF computer: stick one probe into molex connector with red wire, with other probe measure https://en.wikipedia.org/wiki/Industry_Standa … SA_Bus_pins.svg
pins B3 B29 D16 in every slot
for good measure you can also stick one probe into hole with yellow cable and measure resistaonce to pin B9

Its slightly possible POST card would still power up using parasitic power from just the bus.
Next remove all cache chips and jumper board for no L2 cache.
Next debugging steps would involve checking all ISA termination resistors - those long black 10 pin noodles between last ISA slots on 20260328_115617.jpg RP9 11 13 14, those are resistor packs in parallel (with one common leg) and all should have ~similar resistance.
Next scope on pin B30 should read 14MHz.
Next poking all pins on big sis chip with a toothpick while using loupe too see if any move.
then it becomes too involved for fixing over the internet 🙁