Heatsink+fan the northbridge? Put a peltier cooler on it?

[this might be about to right size. https://thepihut.com/products/peltier-thermo- … er-module-5v-1a
How to properly attach it to the northbridge, and a heatsink on top, could be a problem]

I tried a Socket 478 northbidge cooler with limited success. Strangely the stability issues will not occur when the chip gets really "hot" but when the temperature rises from 22° to 28°. So the temp needs to be kept very low.
Without a Peltier or similar this is not possible when the board is in a case and it´s a hot summer day 😀

Have you tried just pushing on this chip instead of cooling it?

I first suspected some bad BGA-contacts due to flexing or insufficient soldering. So I put the board on a firm base and pushed down the whole chip first and also at several different areas. This had no effect on the issue. Only temperature does.

Then I had the idea that - if there´s some true hardware defect here - the board would get unstable at 68.5 MHz when it gets heated up.
So I used my hot air station to make all the critical parts / chips boiling hot while the board was running. No chance - it´s rock stable at 68.5 MHz.

So my conclusion for now is this northbridge just can´t do 75 MHz or above.
There´s no hint at 75 MHz FSB on the original box nor in the manual also (I have both).

After I had given up on making this mainboard run at 75 MHz, I did some occasional tests until one day a piece of soldering wire slipped under the 2nd ISA slot and shorted the 12V trace that goes to the chassis fan header..
The trace overheated, dissolved from the pcb and touched some of the ISA pins (and maybe others). You can see the remainings at the upper edge above the ISA slot.

This trace is for the fan header only so it isn´t vital for the system.

After that the board wouldn´t start anymore.
When I tried to reflash the BIOS I found the "A0" pin was dead, but replacing and reflashing the chip didn´t help.
There´s a 74ABT16245C "16-bit transceiver with 3-state output" chip that is getting boiling hot. At the moment I have no idea what it´s function is, but I suspect it´s broken.

In the meantime I replaced the transceiver chip, now it´s not getting hot any longer. The board still does not POST, but now shows "C0--" insted of "----". The "IRDY" LED is contantly lit except when you push the reset button.

At the moment I´m a bit short on ideas where to continue the troubleshooting.

This loosened (and about 10cm long) 12V trace has done more damage than I expected.
It damaged the BIOS chip, a transceiver chip and it killed the VIA "VT82C586B" southbridge.
The trace had contacted several pins at the nearby ISA slot so it´s no surprise components connected to this bus died.
After desoldering the old chip I inserted a new one:

When I had finisheding the soldering work I was happy to finally see the mainboard POST again!
But the happiness didn´l ast long after I found floppy-drives are no longer detected. It doesn´t matter if I connect a (working) floppy with a (working) floppy cable to the onboard chip or if I disable all the super I/O interfaces and try a super I/O ISA card with a floppy connected to it.

The error message is always the same: "Floppy disk(s) fail (C0)"

The ISA slots seem to be working, a ISA graphics card works perfectly.
What the hell could cause an error like this???

isa is all VT82C586B, its microscope time

Wow, that poor board. You have the patience of a saint. 👍

I once had an issue where I couldn’t clock above about 350-450mhz stable.

Believe it or not it was a bios update that fixed it. 🤷‍♂️ Food for thought.

I had doublechecked all solderings and contacts yesterday and I carefully checked every pin this morning. There are no shorts between any pins and all of the solderings are without fault and stable.
What´s irritating me is that an ISA graphics adapter works and ISA PnP cards like 3COM 509B network and a Soundblaster AWE64 are detected by BIOS correctly.

The WINBOND super I/O "W83877F" is hooked to the ISA bus like any other device I suppose but doesn´t work. Is it possible it blocks other floppy controllers from being initialized, even if it´s own interfaces are disabled in BIOS?

BIOS shows the 2 COM ports and LPT port as present even when they are all disabled in the integrated peripherals section and it shows floppy that´s disabled also. As long as this is the case, I think all those additional IO-cards will not work.

Making progress...
Because I suspected the Winbond super I/O to prevent BIOS from treating the super I/O devices from being disabled and by doing so preventing BIOS to register alternative ports, I removed the onboard super I/O - after all it´s connected to the ISA bus and potentially damaged.After that the devices showed up as "not present" in the BIOS summary screen and every ISA super I/O card works perfectly now - including floppy support!

Seems like I need a new Winbond W83877F...

I do have some W83977TF-A, but I don´t know if there´s any compatibility. But this would probably be of no use anyway, because BIOS expects the super-I/O it was compiled for.

It´s a pity since when all this is finally repaired I´m right back where it all began: A 75 MHz board that can do 66 MHz only 🙁

I was just about to say after pondering that that io chip probably got fried. But you are waaaay ahead of me haha.

That chip is probably sitting on a parts board you have somewhere I assume? 😀

Is the bios fully updated? I assume the hoard is recapped? These are most prominent in my mind for the 75mhz issue

This shows what 12V can do in no time to chips that are designed for 5, 3.3 or less volts.

I have ordered some W83877F today after I couldn´t find any donor-multi-I/O card.

Latest BIOS - ✓
Recap - ✓

There´s either a non-optimal oscillator waveform for the northbridge or the northbridge itlelf is some kind of an underperformer. I´ll give that a try when everything else has been sorted.
Problem with the northbridge is it´s A) BGA and B) virtually unavailable.

The new Winbond W83877F super-IO chips arrived last week and now all the integrated peripherals are working again.

After a lot of work the mainboard is now in the same state as before the accident again.
I also added the pull-up resistor at the clock generator to set the FSB to 68.5 MHz.
With a K6-III @ 411 MHz the performance is quite good.
To accomplish 75 MHz FSB, the northbridge would have to be replaced. I still can´t find any seller offering these VT82C597 chips.

Have you tried spot cooling it with a air can to confirm it is the weak link?

Yes I did. When you freeze the northbridge from start with coolant spray (applied at the center of the chip), the system will run stable @ 75 MHz FSB. As soon as you stop cooling the system halts.
I still cannot rule out any malformed / too weak clock signals - at least as long as my oszilloscope is limited to 20 MHz. I´d like to buy a new digital scope with 150 - 200 MHz so I could watch 100, 133, or 166 MHz signals.

Crazy. Hope you can find one! 😀