The next step is test continuity with multimeter on all the traces from the keyboard connector and keyboard chip, if you not sure how google or YouTube has alot of easy info

That's what I did ^^

Data/clock pins are indeed connected to two of the KBC's pins. Not sure if they should go anywhere else on that chip : I can't find any datasheet/pinout for it 😒

the lights flashing on the keyboard probably means the connector is ok, time to focus on the kybd chip. I would try to find info on the kybd chip and what the output on the pins should be then test the kybd chip pins with multimeter.... or swap the kybd chip, but this is all easy to say when its 20 year old parts and hard to find datasheet or replacement

I have some dead (corroded) AT mobos that might have a KBC like this, but I have no idea if they're pin compatible ...

unless they are exactly the same make and model then I really don't know,. have to wait for. somebody with more experience on that one to reply

Well I tried to replace it with an older one from AMI.

The board doesn't POST without the KBC. With any of the two KBC (original and older one), the behaviour is the same, so actually the problem lies somewhere else !? Any idea what could be causing this ?

Deksor wrote:

When I start it up, first the three LEDs of the keyboard stay on for a second, then they turn off. So far, so good. But at the end of the RAM count, they all blink rapidly, and then all turn off ... and of course the keyboard doesn't work at all.

BIOS is able to at least initialize keyboard. KBC is working and communicating with keyboard, otherwise you wouldnt see any LED activity (leds are directly controlled by the bios). But leds should not stay on for full second, and rapid blinking can be either power supply related (5V to keyboard) or bios signaling errors (blink and beep codes).

Deksor wrote:

I tried hot air on it but I don't think it fixed anything
So far I've tried to heat the chips to see if that would help

Best advice I can give you is try to put less trust in internet "Rehot CPU Bro"/"bake the gpu" memes. Heat gun is not a magic star trek healing beam, heating stuff randomly on multilayer pcb is bound to break something (thermal expansion = popcorning = internal cracks). First step should always be diagnosis. Universal PC diagnostic code card is $5 and will tell you where the problem starts. Not that it would help here, because:

Deksor wrote:

The first one is that when I first got it on its tray, I noticed they forgot to install a brass ... and screwed a screw there instead, so there was a big bend in the board.

current behavior fits cracked tracks hypothesis, this board should probably go back to the garbage bin at this point.

Well I didn't heat random parts on the board, I reflowed the pins of the chips which seems to have improved the current state of the board because it now POSTS with the right speed and memory size every time. I also fixed a non POSTing PCChips board this way, so it works if you have cracked solder joints for example (which my pcchips definitely had).

In this case, everything seems to work except keyboard. Maybe I burned some glue logic chip by mistake ? It now behaves the same whatever I do, so maybe there's some hope for it.

Also this board has a DALLAS-like RTC chip, so if I manage to turn on USB keyboard support by writing the correct value in there, this board might be usable once again x)