TAG chip can be either 8k or 32k SRAM, some boards have switches for that as well and/or expect the TAG to be matched to the cache size. But for 256k it should be 32k anyway.

Regarding the burned trace, it goes to the 33ohm resistor near the upper UMC chip, correct? In that case the resistor is a line terminator to improve switching transients, and the other side of it should be connected directly to big chip. In this particular case it goes to a via so that connection is either on the other side or inside (inner layers) of the PCB.
As you can see pretty much all of the signals to SIMM slots are done like that, so you can:
1) find which pin on the UMC chip that one goes to, exactly
2) compare the resistance of this connection to SIMM vs several other ones (pick at random), you should get similar results of about 33 ohms
If you get considerably different values it's worth investigating why.

I like it I really like it, this is great thinking and logic.

So I found the umc chip pin, its the 6th pin down on the right side of the chip, the resistance checks out 33.8 ohm to be exact from the back of the simm where I placed the jumper wire, the 35th and 40th pin form the simm (only the back not the front) show 33.8 ohm to the 6th pin on umc chip.

I got the front pins in the simm confused but found the right ones and the resistance from inside the simm to the umc chip is ok, high 33 ohm

I used the same pins on simm2 which correspond to pin 11 on umc chip and get 33.6 ohm resistance l, probably because I have a jumper wire on simm1 so its slightly higher

I am actually thinking about giving it a shot to change the chipset, I see a few on ebay which are all UM82C491F but the next 2 bottom lines are different, would a chipset like that be compatible or all the lines need to be a exact match?

my impatience got the better off me, I had to lift off the chipset to have a look. Unfortunately nothign exciting under, just a clean board.
At least none of the pads came off, so now time to look for a new one. But hey what have I got to lose have to attack the big boss and get it over with, can't keep soldering random chips

I got my hands on a replacement umc chip, soldered it using the drag technique with a flat tip, didn't really expect it to boot but it did, however still the same problem, this is crazy.

out of ideas for now, ill find some random part to remove or resolder soon

It's possible you have a micro-cracked trace or a via, and not between cache/RAM and chipset but rather the CPU socket and chipset. And finding that would be... difficult.
What CPU clock are you using now? Does it help any to go down to 25MHz?

using 25fsb now, and actually I found at 25fsb it doesn't give me parity errors like at 33fsb but only detects 64kb conventional memory instead 640 kb.

at 33fsb it detects 64kb conventional memory but also gives parity errors and doesn't get as far to the boot screen.

Good hint, I guess next job when I get the itch will be to start reflowing all the traces to the cpu from chipset.

This is VLB mobo so inspect all the 32-bit extensions to ISA slots - like really good. Those are directly connected to CPU and there is a reason why any mobo has at most 3 of those, even though originally 6 were planned in the standard - any more and the CPU will be unable to drive all that copper and parasitic capactitances fast enough to actually work. Point here is, any sort of slightly conductive dirt in one of those slots will cause havoc. And since this is a 2nd hand mobo, you don't know what it went through.

yeah will add that to my list of things, thanks

I think ill try replacing some of the smd resistors too, look at the traces from cpu to chipset and inspect the 32bit vlb extensions

Also what I just found is keeping it at 25fsb but changing from a sx to a dx cpu and using correct pin configuration for both types when a dx cpu is ran it gives me the offboard parity error where with a sx cpu it goes to boot screen but has 64kb conventional memory

Maybye because dx clock is doubled even less stability?

I checked the vlb slots but they are perfectly clean, being on the topic of checking I decided to check all the smd resistors and capacitors in circuit, for the sake of it.

I found 1 resistor that was giving me no reading at all, so I had a closer look and it its connected to TC4069UBP cmos hex inverter from my google search.

So I got excited, before trying to remove R32 (the resistor with no reading) I reflowed and added new solder and it stared to give me a reading. Got 1 successfull boot, but then same problem again, coincidence? who knows.

disappointed while had my tools out decided to take out the cmos hex inverter and ofcourse I socketed it like the rest of the board.

I found the data sheet for this part.

So I tested with my meter on continuity positive to input negative to output. I have a cheap meter that doesn't beep just shows a readout .01 means its full continuity and higher numbers are some bigger resistance.

So all the input and output pins on the left side give me exactly the same reading 5.xx but the input and output pins on the right side gave me nothing, some moments my meter flashed numbers for 1 second then nothing.

I am hopeful this could mean its a faulty chip, being a cmos inverter it is on the relevant topic where the failures are, but another needle in a haystack. Time to look for this part and wait.

You're checking for continuity through an inverting buffer? I don't think that's meaningful at all. Much better to power the chip from 5V, connect each input to Gnd and see if you get +5 on each output. Then connect each input to +5 and see if you get Gnd out. (Or use your logic tester to do the same.)

I couldn't find the chip name in my tl866 that is the first thing and easiest test I would like to try. As you can tell I am not very advanced with electronics theory.

Is there any generic buffer profile I could test it on a tl866 with, do you know?

If you look for just "4069" isn't that the right chip ? They seem to put very basic names in there, sometimes they even forget about letters it seems.

yes you are right, and I remember you also mentioned how they don't label the chips that well in the app, the chip checks out under the 4069 profile. thank you and back to my cave

also the r32 resistor is 502 but I couldn't find those online, upside down it also looks similar to a 205. Do both variants exhist?

So I have done so much stupid unnecessary guessing random fixes on this board which got me nowhere. I noticed a tantalum capacitor nexy to the clock gen had a scotch mark on it, previously I thought it was somebody marked it with a black texta a black line, because the capacitor was not swallen like a electeolitic one I dismissed it and went on with a crusade looking for the invisible enemy. So now I went back to basics and realised this is actually a scortch burn mark, oh how stupid am I, I was thinking should I admit this or hide and don't tell anybody.
Changing the tantalum capacitor fixed the following issues:
- Parity errors (I realised its actually cpu parity error)
- Not being able to use high fsb (25/33/40 works everytime)
- Correct base memory is detected always as 640kb

This is the guy:

* I still have the no L2 cache problem which developed throughout my adventure, it shows up 5% of the time.
External cache always shows up as disabled in bios, when turn to enable it saves in bios, after reboot its disabled in bios again. I will work on this naturally but the board does work fine now.

* to cheer myself up I tell myself that this board did have multiple issues, the crashing/freezing bios was one of those which was also fixed early on so maybye I didn't waste all my time.

bad tantalum that didnt explode? lucky!
so cant be shorted, is it very low resistance?

rasz_pl wrote:

bad tantalum that didnt explode? lucky!
so cant be shorted, is it very low resistance?

I already threw it out unfortunately

rasz_pl wrote:

bad tantalum that didnt explode? lucky!
so cant be shorted, is it very low resistance?

Not necessarily.

Actually exploding the tantalum cap requires an energy source great enough to heat it up enough to reach combustion.

If the PSU can't deliver enough power, or it is in a circuit that is current limited, it simply won't consume enough energy to explode.

Unstable power supply going to the chips due to faulty caps can give you all kind of weird issues. Especially if it's a clock generator. But at the very least you found it before replacing even more chips on the mobo.

As for R32 - what resistance values does your meter go up to? The 4096 is very often used as a clock generator - feedback and amplifier for the 32kHz crystal. This is because it's CMOS and very low power at this frequency. In fact it has so high internal impedance that there's usually a resistor needed across the crystal to improve the linearity. Though again you don't want to waste power - so for 32kHz this resistor is in 10-22 Mohms range. Many meters will show it as open circuit.

Nice to finally see some progress 😁