Ok, I found a different AT keyboard.
System crashes the same way, keyboard ruled out.

Tried with different memory modules too. Same.

This one below is still strange.
Additional beeps when keyboard is disconnected.

-One long beep during POST. - Same when keyboard is connected.
-One "minor"/short beep when memory count is done. - When keyboard is connected there is no such thing.
-One "minor"/short beep when booting OS - When keyboard is connected there is no such thing.

This is a 486, so...

Bad A20 gate control through keyboard controller chip?

wierd_w wrote on 2026-09-03, 12:36:

This is a 486, so...

Bad A20 gate control through keyboard controller chip?

Forgot to mention; I did a KBC swapin'.
The swap is working; system still crashin'.

When doing a visual inspection it was strange that every chip manufacture date was middle of 1995 except the KBC (made in 1993).

Also all the previous observations hold up, I think.
When a benchmark or a game is running in a loop or Windows in idle state there is no issue and no freeze.
Once I start using keyboard the system freezes at some point.
(I have not tried a mouse and use it exclusively in Windows and check if it crashes.) <- This could be a next step to check using a serial mouse.

Can something else besides the KBC have an effect on the keyboard and crash?

JAKra85 wrote on 2026-09-03, 12:52:

Forgot to mention; I did a KBC swapin'. The swap is working; system still crashin'. […]
Show full quote

wierd_w wrote on 2026-09-03, 12:36:

This is a 486, so...

Bad A20 gate control through keyboard controller chip?

Forgot to mention; I did a KBC swapin'.
The swap is working; system still crashin'.

When doing a visual inspection it was strange that every chip manufacture date was middle of 1995 except the KBC (made in 1993).

Also all the previous observations hold up, I think.
When a benchmark or a game is running in a loop or Windows in idle state there is no issue and no freeze.
Once I start using keyboard the system freezes at some point.
(I have not tried a mouse and use it exclusively in Windows and check if it crashes.) <- This could be a next step to check using a serial mouse.

Can something else besides the KBC have an effect on the keyboard and crash?

Yes, the keyboard chip can go bad.
In your case I hope you have a spare for a test or it is at least socketed.

Check with the documentation on trw

Nexxen wrote on 2026-09-03, 13:06:

JAKra85 wrote on 2026-09-03, 12:52:

Forgot to mention; I did a KBC swapin'. The swap is working; system still crashin'. […]
Show full quote

wierd_w wrote on 2026-09-03, 12:36:

This is a 486, so...

Bad A20 gate control through keyboard controller chip?

Forgot to mention; I did a KBC swapin'.
The swap is working; system still crashin'.

When doing a visual inspection it was strange that every chip manufacture date was middle of 1995 except the KBC (made in 1993).

Also all the previous observations hold up, I think.
When a benchmark or a game is running in a loop or Windows in idle state there is no issue and no freeze.
Once I start using keyboard the system freezes at some point.
(I have not tried a mouse and use it exclusively in Windows and check if it crashes.) <- This could be a next step to check using a serial mouse.

Can something else besides the KBC have an effect on the keyboard and crash?

Yes, the keyboard chip can go bad.
In your case I hope you have a spare for a test or it is at least socketed.

It's socketed and I have 3 KBC chips. Tried with two and same issue.

I think I narrowed down the issue to something keyboard related.
- Boot with keyboard connected.
- NumLock LED lights up.
- Boots and Memtest+ started.
- Press "C" for configuration, works fine. Also, I can navigate in this menu system.
- Pressing "Space" activates the Memtest+ scroll lock.

Now:
- Pressing NumLock button does not turn off the NumLock LED. It has no effect.
- Pressing "Esc" freezes the PC.
- Restart. Boots and Memtest+ starts again.
- Pressing "Ctrl+Alt+Del" ... well, I loose the keyboard but the Memtest+ continues to run fine, no freeze.
- Re-connect the keyboard, acts as nothing happened. I can press "C" I can navigate in the menu system, I can press "Space", so on.
- Pressing "Ctrl+Alt+Del" disconnects the keyboard, again. 😁 Pressing "Esc" freezes the PC.

Turn off typematic rate setting in CMOS Setup if enabled, and if there is a KBC clock option set it low, like 7.16Mhz was an option on one of my boards vs 11 or 12mhz. See if that's any better.

BitWrangler wrote on 2026-09-03, 13:26:

Turn off typematic rate setting in CMOS Setup if enabled, and if there is a KBC clock option set it low, like 7.16Mhz was an option on one of my boards vs 11 or 12mhz. See if that's any better.

Tried it and could not make a difference.

Gate A20 option is: Normal or Fast
Typematic Rate setting is: Enabled or Disabled.

There was once an interesting outcome when playing with these Gate A20 and Typematic settings.

In previous runs when I hit "Esc" it crashes Memtest+ at this step:
Testing: "132K - 16MB Halting..."

Once it crashed at the same step with a bit different message:
Testing: "132K - 16MB Halting... alted."
Or something very similar, I could not reproduce this.
Set the Typematic delay to: 1000

Also, that "+" sign keeps beeping when system crashes. (every time)
Memory tests are stopped.
This is why I refer to this as: soft-crash.

Hmmm that's odd. Usually when you start getting display stutters and bad characters it's something in the DMA going wonky from cache or RAM problems. Then if those check out you're wondering if anything using DMA is glitching it, leading usually to the floppy controller, which I've only seen go a bit funky when the 14.318 reference crystal has gone way off. Though your 14Mhz crystal is marked 14.365 ... curious.

Trace the keyboard to the 8042 (keyb controller).
Are there any other chips in between? Some LSxxxx? A resistance? Could be one of those not working.

I am not at home right now. Will check these.

I just had another idea. Does this ALi chipset have an internal KBC?
If yes, could that conflict with the external KBC.
There is a 2x5 pin layout near the ketboard AT connector with 2 jumpers set.

whats your ram time stats, and wait stats ?

Nexxen wrote on 2026-09-03, 14:13:

Trace the keyboard to the 8042 (keyb controller).
Are there any other chips in between? Some LSxxxx? A resistance? Could be one of those not working.

Result:

KB Connector Pin1 - NC to KBC (directy).
KB Connector Pin2 - KBC Pin: 40, 26, 25.
KB Connector Pin3 - KBC Pin: 39.
KB Connector Pin4 - KBC Pin: 2, 7, 20 and (mainboard GND).
KB Connector Pin5 - KBC Pin: 1.

JAKra85 wrote on 2026-09-03, 18:01:

Result: […]
Show full quote

Nexxen wrote on 2026-09-03, 14:13:

Trace the keyboard to the 8042 (keyb controller).
Are there any other chips in between? Some LSxxxx? A resistance? Could be one of those not working.

Result:

KB Connector Pin1 - NC to KBC (directy).
KB Connector Pin2 - KBC Pin: 40, 26, 25.
KB Connector Pin3 - KBC Pin: 39.
KB Connector Pin4 - KBC Pin: 2, 7, 20, (mainboard GND).
KB Connector Pin5 - KBC Pin: 1.

The attachment KBC_Con.png is no longer available

The attachment KBC_Chip.png is no longer available

Tested a different (working 486) mainboard, different result!
KB Connector Pin1 - NC to KBC (directy). - Same.
KB Connector Pin2 - KBC Pin: 40, 26, 25. - Same.
KB Connector Pin3 - KBC Pin: 39. - Different! (on pin 39 I have 10KΩ resistance)
KB Connector Pin4 - KBC Pin: 2, 7, 20, (mainboard GND). - Different! (Also pin: 33)
KB Connector Pin5 - KBC Pin: 1. - Same

Tested a different (working 386) mainboard, exactly the same results as the board in question.
KB Connector Pin1 - NC to KBC (directy). - Same.
KB Connector Pin2 - KBC Pin: 40, 26, 25. - Same.
KB Connector Pin3 - KBC Pin: 39. - Same.
KB Connector Pin4 - KBC Pin: 2, 7, 20, (mainboard GND). - Same.
KB Connector Pin5 - KBC Pin: 1. - Same.

BitWrangler wrote on 2026-09-03, 14:11:

hough your 14Mhz crystal is marked 14.365 ... curious.

That is actually a 14.3G5P marking and 24.0E5P the other one.

nuno14272 wrote on 2026-09-03, 14:49:

whats your ram time stats, and wait stats ?

The SIMM modules are currently 2x4MB 60ns. The other ones are 4x4MB 70ns.

In BIOS, it does not matter to what is it set. Currently: slow.
I have Slow, Normal, Fast and Fastest options.

One more thing I noticed during memtest+.
I press NumLock button and LED not turning OFF (stays ON).
However there is a delay here that I can "play with."
Maybe someone experienced similar and can figure out what is happening here.

I press NumLock button and then press "c" for configuration menu -> the menu pops up in an instant.
I press NumLock button quickly 5 times and press "c" for configuration menu -> the menu pops up with a small delay. (~7-10 sec)
I press NumLock button quickly 10 times and press "c" for configuration menu -> the menu pops up with a BIG delay. (20+ sec)
I press NumLock button quickly ~20 times and press "c" for configuration menu -> the menu pops up with a delay almost reaching 1 minute! Huge delay.

When pressing NumLock button in a quick succession there could be a small (NumLock) LED dimming once or twice. Barely noticeable, in the millisecond range. - not sure how relevant this could be.

-=Possible root cause of all this instability=-
Measured voltage on the keyboard connector.
During boot it is near and around 0V a little above 0V (~0.2V).
By the time it posts, quite late... it rises to 4,96V.
When I press ANY button on the keyboard the voltage drops to 0.xV
It drops to 0.xV for around 3 seconds before rising back 4,96V.

There is a fuse marking, but missing built without. Factory default.
Then... the two tantalum caps!
More exactly: TC2 that was dismissed previously.
Could there be a 3rd state? Working, exploded, "leaky(?)".
Leaky=I hit anything on the keyboard drops voltage then recovers.

Sorry, my fault: measured Clock(5V) and Ground instead of 5V and Ground.

Note that SMD components also fail, so maybe C1 could be the problem. However, measuring caps in circuit isn't very reliable. In any case, if something is off about that section of the circuit you should take them out and measure again to see what you get. Usually when measuring resistance of caps standalone the value should go up the longer you measure.

I don't think you'll notice the charge time on small value caps though, probably need at least 10uF or so.