PCIRESET pin being pulled up and connected to ground with two caps? maybe chipset pci reset steering is fried and someone was salvaging the board - might be factory mod/fix.
PCIRESET is connected in parallel on all slots so should make no differenve where you plug your pci vga, either works in all or none in case of this particular mod.
I would look into the slots if any pins arent crushed inside.

Thank you.

I re-checked this morning with another PSU. Same thing. The BH6 can't start with the GF4 MX PCI in the first slot (any other PCI slot is OK). The system can boot with a Matrox G450 PCI in the first slot.

Quick update.

During my troubleshooting i removed the mod done to the PCI bus. And saw no difference with the reboots.
After extensive testing i'm pretty sure it was something caused by my CRT monitor. Each time the system unexpectedly rebooted there was a concomitant "pop" in the monitor.
I was not sure at first if the "pop" was the cause or the consequence. But it seems to be the cause.
I tested several graphics cards, both PCI and AGP and the same thing occured.
After switching to a VGA LCD monitor, no issues. I was able to run memtest for 2h+ and left quake 2 running for an hour or so without any reboot.
As soon as i switched back to the CRT, the issue reappeared.

I suspect some kind of interference flowing back to the card when the monitor "pops" and the BH6 might be more sensitive to that because i never saw issues with other systems.

edit : the PCI Geforce mx still won't allow the MB to start when in the first slot but it seems it's a different problem.
edit2 : i still have pictures of the PCI mod in the back and can re-apply it if required

I dont know when the pci reset is performed other than bootup, its possible its only ever needed in weird circumstances like waking up from sleep - in that case you wont see a difference without the fix it something was broken in the first place, not that anyone would ever need waking from sleep on a retro computer 😀

PCI is a weird type of bus, it uses https://en.wikipedia.org/wiki/Reflected-wave_switching I dont really understand how this works 😀 but its very sensitive to badly designed receivers (wrong impedance) causing local ringing

Could be something about the first slot being a master and the rest of them aren't? Like 1st slot below AGP is master and the rest are slave.

All slots are the ~same(*), but slot closest to chipset receives that "Reflected-wave" first and supposedly can interfere with signals the most before it all propagates to rest of the bus and calms down.

* 4 PIRQ signals are in different order but shared and every slot gets dedicated DMA req/grant pins, other than that its straight up parallel

Is the PCI slot with the issue the one next to the ISA slot?

rasz_pl wrote on Today, 04:28:

All slots are the ~same(*)

IIRC on the BH6 the PCI and ISA slots that are next to each other are actually shared bus slots and NOT the same as the other PCI slots, and always had issues with some PCI cards being on that shared bus.

Also the The AGP slot shares resources with PCI slot 1 as well. So if you have a Video card in both the AGP and the PCI next to it, there maybe resource conflicts, the exception is a Voodoo 1 or 2.

Matchstick wrote on Today, 04:43:

IIRC on the BH6 the PCI and ISA slots that are next to each other are actually shared bus slots and NOT the same as the other PCI slots, and always had issues with some PCI cards being on that shared bus.

Also the The AGP slot shares resources with PCI slot 1 as well.

no such thing as PCI sharing resources with ISA.
The problem with more than 4 PCI slots was always interrupt line sharing. There are only 4 interrupt lines to go around (rotated in each slot) https://en.wikipedia.org/wiki/Peripheral_Comp … nect#Interrupts. AGP also shares two of those lines in 440BX, PIRQ#A PIRQ#B. Badly behaving cards unable to share interrupts might conflict with each other. There is no law that states PCI card needs to use its 1st interrupt line, thats what Plug&Play is for.
PCI slot 5 will repeat same PIRQ order as slot 1 thus will be as problematic as slot 1 in a case of badly implemented pci card that
-only allows use of PIRQ#A
-cant properly share it

If i remember correctly i only tested the first two PCI slots and the AGP one.
I had the issue regardless of the slot type or number.

I focused on the first two PCI slots because i initially thought that the issue had something to do with the mod on the back.
And the first slot had the mod while the second one had not.
I wanted to see if there was a difference

I have a BH6 1.1, fully recapped, and I can't get a PCI MX440 to work in any PCI slot, no signal and 4 beeps on boot. AGP cards work fine though.

curioser and curioser