I tested them all (tantalums), no shortage. I tried with another CPU as well, no change. I'm wondering if a track on the board could have been broken when I removed/put back some cards due to the mechanical constraints, but visually I can't see any cut. I found that page https://dosdays.co.uk/topics/bios_ami.php with BIOS code, my bios looks like the AMI Color one.

@rasz_pl I did the tests you suggested with red wire to ISA B3 B29 D16, same for yellow wire and B9, on each isa slots, got [0-1] ohm, that seems correct. I tested all ISA termination resistors as well, seems okay, all around 4.7kohms. Tomorrow I'll try with removing all the cache chips, and check for chipset pins. Thanks for your advices

So I tested all pins around the chipset with a loupe and a toothpick, it is very hard because very very small, but I checked them all 2 times and did not found any moving/broken/unsoldered.

I'm not sure I can try to remove the cache chips, according to the motherboard doc I linked on a previous post here I only have options to set between 64kB and 256 kB, with the 64kB configuration requiring different chips than the 256kB configuration, I can't find a jumper option to disable it completely. About the 14Mhz on B30 I don't know how to test that, I only have a multimeter, no osciloscop.

I already had those beeps boot sequence sometime, mostly when testing some video cards, each time I had to remove and replug the video card and it was working back.

I can provide more photos if you think it is useful, in case I missed something

looking at photos your board matches closer to https://theretroweb.com/motherboards/s/bcm-lx400a-g minus the PS2 port.
a single photo showing whole motherboard in high resolution (good enough to read chip markings) front and back would help.

So far we eliminated:
psu ok
slot pins look ok
no shorts on power rails

I cant tell how the FSB jumpers (most likely jp10 11) are supposed to correspond to fsb. clokc chip is U38 ics av915-5 https://theretroweb.com/chip/documentation/av … 29924961515.pdf
pin 16=FS0 15=FS1 1=FS2 10=FS3
0: 0000 2 MHz
1: 0001 8 MHz
2: 0010 16 MHz
3: 0011 20 MHz
4: 0100 25 MHz
5: 0101 33 MHz
6: 0110 40 MHz
7: 0111 50 MHz
8: 1000 4 MHz
9: 1001 16 MHz
A: 1010 32 MHz
B: 1011 40 MHz
C: 1100 50 MHz
D: 1101 66 MHz
E: 1110 80 MHz
F: 1111 100 MHz

did computer always run at correct 66MHz speed (33 fsb)? you can try all 4 combinations of jp10 jp11 just for a laugh, or use continuity setting in multimeter and trace U38 which jumpers pins pin 16=FS0 15=FS1 1=FS2 10=FS3 connect to. Its highly likely 1 and 10 are hard connected to ground and power with only pins 16 15 being on jumpers.

re seat U38

re seat cpu

pull 9 sram cache chips off, board doesnt seem to have "disable" jumper and auto detects presence

re seat RTC chip, check if all pins are straight and go into socket holes correctly

re seat keyboard controller (next to rtc) for good measure

Are POST codes and beeps the same with and without graphic card? To me it looks like BIOS executes VGA bios fine and crashes somewhere later. Lack of video signal can happen due to
- not initializing Video chip by not executing VGA bios at all
- executing VGA bios, but CPU is somehow failing and does it wrong before crashing
- no power to DAC chip, you tested +5 +12 so not that
- no 14.318Hz OSC signal on ISA bus and a VGA card relying on that clock. Not many cards need that clock and "I tried several video cards in several slots (ISA and VLB), same result. " would eliminate this possibility unless you are very unlucky with VGA selection. Which cards did you try? do they have clock crystals on them?

have you tried another monitor?

try slightly pressing on chipset while resetting

try slightly pressing on CPU while resetting

Thanks for your advices and your time. Actually both cards looks closed to each other, but mine seems to be exactly none of them. But it is a BCM one. The computer did always run at 66Mhz (fsb 33). I don't understand the test you're asking for tracing U38 to jumpers, but I re-seated U38, and the CPU as well, same for RTC and keyboard controller. I'll test removing the cache chips but there is 9 chips and I don't know if I should remove the 9 or only the 8 that are aligned in 2x4 configuration. Without video card : 41-3B, with video card : 1-3B, beeps are the same. I tried with 2 different ISA video cards (one Trident TVGA9000B, and one OAK OTIO77), and 2 VLB (the usial one : Diamond Trio64, and a Tseng ET4000-w32). I didn't tried another monitor, it is a "modern" fullHD one with a VGA port I'm currently using with my modern computer, and it has always well worked with my 486 computer. I just tried to press on the chipset and on the CPU then turn the computer on, no changes.

Are those photos good enough ? I had to resize them because of the 5MB file size limit on the forum, but I can provide others of specific location on the MB if you need

Actually I noticed I had a loose connection with one segment of my post card, I fixed it, the stop value is always 41-3B, here is a video of the whole boot sequence in case the first one was wrong due to that loose connection

vbug wrote on 2026-03-29, 15:39:

I don't understand the test you're asking for tracing U38 to jumpers

- multimeter set to continuity mode, the beepy test
- one probe on chip U38 pin 16 (last one, next to blue capacitors)
- remove jumpers from JP10 and JP11
- second probe to JP10 and test if one of its 3 pins beep
- if no beeps repeat with second probe to JP11 and test if one of its 3 pins beep
- repeat with chip U38 pin 15
- put jumpers back

This will confirm JP10 JP11 FSB selection combinations.

vbug wrote on 2026-03-29, 15:39:

I'll test removing the cache chips but there is 9 chips and I don't know if I should remove the 9 or only the 8 that are aligned in 2x4 configuration.

all 9 to be on a safe side

vbug wrote on 2026-03-29, 15:39:

Without video card : 41-3B, with video card : 1-3B, beeps are the same.

>2 short (high), 8 long (low)

I think I watched first video without headphones, listening again its 1 long 8 short standard no video code https://blog.theretroweb.com/2024/01/20/amibi … ist/#Beep_codes

so its complaining about no VGA card every time, meaning its failing to execute VGA Bios, or maybe even detect its presence

vbug wrote on 2026-03-29, 15:39:

vbug wrote on 2026-03-29, 15:39:

I didn't tried another monitor

no need to, Im sorry, I somehow thought due to POST codes so high that VGA Bios is being initialized. I checked again and can confirm AMI will stop somewhere around 30-40 when it fails to detect Video card.

vbug wrote on 2026-03-29, 15:39:

Are those photos good enough ?

pictures are almost great, hard to follow tracks but will do

Ok so the board is unable to
- detect Video bios extension
or
- execute VGA bios
or
- VGA bios fails to detect card

POST card uses port IO writes so at least we know
- 8 bit data bus works
- 16 address lines work
- /IO_WR signal works

To execute VGA bios successfully we also need at least
- /SMEMR isa pin B11 because bios is usually 8bit
- /IOR pin B14
- LA17-23

we also might need
- /MEMR isa pin C9
- /MEMW isa pin C10
- SD8-15

re-reading symptoms it does sounds like motherboard developing an intermittent connection that turned into permanent break after shuffling cards. It can be a cracked joint, broken pin, broken connection in one of the chips, a chip that simply died due to mechanical/thermal stress and no longer works.

isa pinout https://en.wikipedia.org/wiki/Industry_Standa … SA_Bus_pins.svg
chipset pinout https://theretroweb.com/chip/documentation/85 … 5d517668401.pdf
74ls245 pinout https://www.circuits-diy.com/74ls245-octal-bu … tput-datasheet/

It would involve patiently and slowly tracking connections between chipset pins and ISA slot:
- /SMEMR
- /IOR
- /MEMR
- /MEMW
- SA0-1 -> SA0-1
- A2-23 -> SA2-19 and LA17-23
- data bus gets complicated quick, D0-15 -> SD0-15 SD8-15 -> SD8-15?

The trick is those arent direct connections, there are buffers between everything and with just multimeter you will only be able to check if there arent any breaks between, but not fully test if everything functions as desired.

Example: lets say we want to check CPU D0 -> ISA DATA 0 (pin A9)
- lay mobo on the back. remember pin numbering will be flipped
- find ISA slot data 0, its pin A9
- first scan all pin A9 pins of all slots to check if there are any obvious tracks going to chips on top
- for A9 there is one track on second ISA slot, somehow called SLOT4 on this board 😮 😀
- this A9 track on second slot then goes up (if you orient board so the CPU is in lower left corner)
- it lands on pin 18 of a 20 pin Chip hiding under the BIOS chip 😀 this chip is most likely 74ls245
- pin 18 of 74ls245 passes signals to pin 2
- pin 2 of that chip most likely connects to BIOS chip pin 11 (data 0)
- this wasnt what we were looking for (CPU D0) so we walk back to ISA pin A9
- sine we dont see any more obvious tracks now begins tedious part where you probe every pin of every 20 pin chip hoping to find one that beeps
- when you find beep you check what chip it is, if is 74ls245 you check the pin corresponding to the beeping pin in 74ls245 datasheet
- you continue from that new pin buzzing CPU socket looking for beeping

and repeat for every one of ~30 pins that need checking.

With oscilloscope procedure is different, you can probe ISA pins directly hoping to find missing or "funny" signal.

as you can see this is a TON of work 🙁

Thanks again for your help, btw I'm impressed by your knowledge !

So I did the tests from U38 to jumpers, it looks okay. I tested the card after removing all cache chips, no changes (so I put them back)

I started the example test, until "pin 2 of that chip most likely connects to BIOS chip pin 11 (data 0)" it's okay, but after that "since we dont see any more obvious tracks now begins tedious part where you probe every pin of every 20 pin chip hoping to find one that beeps" I'm lost !

A2-A9 (Data0-Data7) goes all to 74ls245, but when I'm on that 74ls245 I don't understand what I should check

"A2-23 -> SA2-19 and LA17-23" -> What are SAxx and LAxx ?

"D0-15 -> SD0-15 SD8-15 -> SD8-15" -> What are SDxx ?

And, following tracks on the back side of the cart is doable, but on the front side it seems very very complicated as there are a lot of elements hiding them 😒