try hair drier & cold spray, maybe its temp related, either margin chip between ram-cpu or partially cracked track going open once everything warms up

Does it work OK if you remove jumper JP3 ? That is the parity check enable/disable ??

rasz_pl wrote on 2023-06-13, 21:27:

try hair drier & cold spray, maybe its temp related, either margin chip between ram-cpu or partially cracked track going open once everything warms up

How to proceed ?
I'll see this as a last resort - I'm afraid to damage another "good" component by side effect, to see

Horun wrote on 2023-06-14, 00:39:

Does it work OK if you remove jumper JP3 ? That is the parity check enable/disable ??

Oh, good point and not listed but already tested -> I've still a system halt, without the full screen message and a buggy display with a thing like a ghost system prompt on the at the bottom left, with a system one beep and all is frozen

If there is anything else to try..

80386SX wrote on 2023-06-14, 05:06:

rasz_pl wrote on 2023-06-13, 21:27:

try hair drier & cold spray, maybe its temp related, either margin chip between ram-cpu or partially cracked track going open once everything warms up

How to proceed ?
I'll see this as a last resort - I'm afraid to damage another "good" component by side effect, to see

Hair drier is less than 100C, cold spray (for example compressed air for cleaning keyboards) isnt that cold either. If anything dies with those temps it was already damaged and on its way out.
You run something intensive (wolfenstein3d?) and selective heat up components on the board looking for a crash, then same deal with cooling.
But before that I would just resolder all pins on Headland chips and CPU. Sadly no datasheets for those chips so cant go do more detailed debugging without reverse engineering.

80386SX wrote on 2023-06-14, 05:10:

Horun wrote on 2023-06-14, 00:39:

Does it work OK if you remove jumper JP3 ? That is the parity check enable/disable ??

Oh, good point and not listed but already tested -> I've still a system halt, without the full screen message and a buggy display with a thing like a ghost system prompt on the at the bottom left, with a system one beep and all is frozen

If there is anything else to try..

so its not parity problem but memory/data corruption

rasz_pl wrote on 2023-06-14, 13:37:

Hair drier is less than 100C, cold spray (for example compressed air for cleaning keyboards) isnt that cold either. If anything […]
Show full quote

80386SX wrote on 2023-06-14, 05:06:

rasz_pl wrote on 2023-06-13, 21:27:

try hair drier & cold spray, maybe its temp related, either margin chip between ram-cpu or partially cracked track going open once everything warms up

How to proceed ?
I'll see this as a last resort - I'm afraid to damage another "good" component by side effect, to see

Hair drier is less than 100C, cold spray (for example compressed air for cleaning keyboards) isnt that cold either. If anything dies with those temps it was already damaged and on its way out.
You run something intensive (wolfenstein3d?) and selective heat up components on the board looking for a crash, then same deal with cooling.
But before that I would just resolder all pins on Headland chips and CPU. Sadly no datasheets for those chips so cant go do more detailed debugging without reverse engineering.

80386SX wrote on 2023-06-14, 05:10:

Horun wrote on 2023-06-14, 00:39:

Does it work OK if you remove jumper JP3 ? That is the parity check enable/disable ??

Oh, good point and not listed but already tested -> I've still a system halt, without the full screen message and a buggy display with a thing like a ghost system prompt on the at the bottom left, with a system one beep and all is frozen

If there is anything else to try..

so its not parity problem but memory/data corruption

Many many thanks for your advices, but I spent my evening to make this and resolder all Headland chips + logc probe testing, but without any result 🙁
Im afraid that just one of RAM controller (Headland HT102/B1A4000 I guess) have just an internal issue...

I've done a new check also : I made and eeprom dump of Low and High BIOS to compare with same dump findable on the web, to check if not corrupted.. but same SHA1
- L : CD5629E839B531D2DB01A297DEAE65D9E34B09A7
- H : 5CF6BE081FC23328B0C14D1F1744350CEB99AC9A

This motherboard is driving me crazy i just want to cry 🤣.. or maybe an internal copper track of the PCB that has cracked.. I don't know what to think anymore..

On the reproduction side of the problem, I just noticed the swapdos systematically crash the RAM (probably because of a reservation and loading of RAM at once: I could make you a video of this)

FYI the way parity error is usually handled on these mobos is via NMI. Could be the interrupt is somehow triggered randomly but not because of actual parity error, and that would be why the address seems to be different every time.

NMI on PCs is actually maskable, via external circuit. The way it's done is by accessing CMOS NVRAM address register, bit 7 is the mask and not part of the address (so there can be only 128 bytes stored in CMOS, less because of RTC registers taking some space, but there are RTC chips with banked NVRAM). If you can try writing a small COM program that will disable NMI after boot, see if it cures your system (NMI is not usually needed). If you don't have the tools I might have some time during next weekend.

EDIT: Here's the program, didn't have time to test it though. Run it "NMI 0" to enable NMI or "NMI 1" to disable it. Try without any memory managers first, even HIMEM, since going through CPU reset or BIOS code might change NMI state.

Deunan wrote on 2023-06-14, 23:17:

FYI the way parity error is usually handled on these mobos is via NMI. Could be the interrupt is somehow triggered randomly but not because of actual parity error, and that would be why the address seems to be different every time.

NMI on PCs is actually maskable, via external circuit. The way it's done is by accessing CMOS NVRAM address register, bit 7 is the mask and not part of the address (so there can be only 128 bytes stored in CMOS, less because of RTC registers taking some space, but there are RTC chips with banked NVRAM). If you can try writing a small COM program that will disable NMI after boot, see if it cures your system (NMI is not usually needed). If you don't have the tools I might have some time during next weekend.

EDIT: Here's the program, didn't have time to test it though. Run it "NMI 0" to enable NMI or "NMI 1" to disable it. Try without any memory managers first, even HIMEM, since going through CPU reset or BIOS code might change NMI state.

Oh guy, it's so kind.
Indeed it's about the NMI - with your COM test, the machine is much more stable!
You revive me too somewhere 😀
I added a demo of your prog in attachement (zip) to show you the results.

Finally I had to go back to 1MB of RAM in chip, because 2MB or 4MB was poorly supported : I still had some freezes.
I'm pretty sure there is a problem with one of the memory controller chips.
Indeed, on this card there are two chips dedicated to this purpose, and when I do a RAM scan under SI9, we see that the "half" of the RAM is intermittently a concern.

But hey, from there to replacing a chip - or rebluid the motherboard -, I don't feel the courage to do it for the moment, especially since I don't have the professional hardware for that at all. ;D

So for the moment - and so that I won't be fooled in the future if I forget to load your COM file when reinstalling somewhere - I made a "hard patch" 🤣 I disabled the NMI pin of the CPU (see photo below).
It's not ultra clean but it's the most elegant fix I guess at the moment, in the sense that ; that's it or I put this card in my spare scrap box on my garage.

Well here it is, so with the 1MB on board, I ran benches for several hours in a loop and zero worries, the machine is stable! Zero crashs and zero freeze! 😀
Which was not the case before, whatever the amount or the mode of memory used (512k to 4Mb, by SIPP or on board).

All that it's for a restoration project of a complete PC 286 that I had found outside on the ground at the edge of a forest...
I'll post some before/after pics 😀
Old comp restoration is a real passion let's not be ashamed 🤣

Maaaaaany and many-many thanks again ALL for your help!! <3

Cheers!

Some pictures of the restoration.
I only have two for the beginning, this pc was rusty everywhere and in a bad state.
For example, on the top of the can here it's gray and you can't see it because it dried, but there was some green forest foam. And inside, various bugs (spiders, etc).

Luckily it was rescued 😀

Cheers.

Cool machine. Glad you got it working.