PCI devices can share IRQs with each other, but not with an ISA device.

You should try memtest86 to rule out bad RAM. Also, check that your voltages aren't sagging under load-- it could be a power supply issue. It's also a good idea to take a close look at all the capacitors on the motherboard. I've seen a system look mostly stable with bad caps but then when the graphics card is exercised the real trouble begins.

The strange thing is, it was the graphics card sharing the same IRQ as some IRQ PCI controller thing which I couldn't find anywhere in Device Manager. I have no idea how it switched over from its default of 11.

I sometimes go into the BIOS to check on certain monitors, such as VCore and MB Temp and the last time I did it the AGP slot was using 3.4 to 3.5V, while everything else seemed fine.

The power supply is of 350 watts, although I'm not quite sure how stable it is. It's only 80 watts less capable than the Corsair CX430 in my modern PC.

I'll definitely take a look at my RAM modules as they provide a rather uncommon total of memory: 192MB from 2 different manufacturers in 3 PC133 modules.

One last note, literally all of the nVidia drivers identify my CPU as a Celeron or a Xeon with MMX, despite it being a Deschutes Pentium II. Windows 98 properly identifies it as such.

I forgot to mention, I've tried 2 OS's, Win 98 and 2000 and both crash at the same point. They both seem to do it regularly when menu animations are turned on and it happens all the time with dropdown list menus, although they may just survive without them.

Nevermind, I figured it was all related to the Graphics Aperture Size option in the BIOS. I had it on 64MB because I thought that was the way to set the amount of video RAM for the card. I lowered it to 4MB and everything works perfectly. I can select the double of the previous number until it maxes out at 256, any idea what I should set to achieve optimal performance?

Still seems strange to me, I don't think having the AGP aperture set to 64MB should be causing you these problems. Not sure what is best, but I think 32-64MB were typical settings.

You mentioned a Deschutes P2 processor so in the comments below I'm assuming you're running a 100MHz frontside bus.
Have you carefully gone over all the jumper settings? I'm especially wondering if the AGP and PCI clock ratios are set correctly. If AGP was running at 100MHz (set at 1:1 with a 100FSB CPU) then maybe it would explain these symptoms. If the PCI clock is at the wrong ratio it could cause all sorts of problems.

If you do any stability tests such as memtest86, then if possible I'd also try it with the FSB overclocked slightly. I don't remember the options on the P2B but it might have a setting for a 3% overclock like 103MHz. A stable system should pass testing with a 3% FSB overclock. If it fails that, then it doesn't have enough stability headroom to be considered reliable, and probably points to an issue that would cause eventual crashes in real usage.

You could also try dumbing down the configuration as much as possible to see if anything stabilizes it. In sequence I' d try setting the FSB down to 66MHz, reducing to one memory module, then removing the Sound Blaster (and any other ISA cards), then removing the ethernet (and any other PCI cards). Maybe some clues will emerge.

A quality 350W PSU contemporary to the era should be working fine, but I'm wondering if it's a good quality unit and how well it suits an older system like this. Lots of modern PSUs aren't very suitable for these old boards, and lots of old PSUs were either cheaply made or aren't in good condition at this point.
Checking voltages in the BIOS monitor is better than nothing, but sometimes those readings are way off, so using a multimeter is a lot more trustworthy. However, even a multimeter won't show high frequency fluctuations.

Yes, it is running at 100FSB. In the BIOS there's an option under PnP settings to set the PCI clock (or timing?). By default it's set to 32 and I think I might have it on 33. I checked the jumpers and indeed all of them are set to run at 100Mhz FSB and the core multiplier is 3.5x.
I should mention that the card runs at 2x AGP speed, which I read in the manual should be of 133Mhz of Bus speed.

I haven't run memtest86 yet, but tried each RAM stick individually in each slot and they all failed. They all lasted slightly longer in the middle slot, though.

PKFreeZZy wrote:

Yes, it is running at 100FSB. In the BIOS there's an option under PnP settings to set the PCI clock (or timing?). By default it's set to 32 and I think I might have it on 33. I checked the jumpers and indeed all of them are set to run at 100Mhz FSB and the core multiplier is 3.5x.
I should mention that the card runs at 2x AGP speed, which I read in the manual should be of 133Mhz of Bus speed.

The P2B doesn't have BIOS based settings for the clock speeds and ratios. The P3B does though.
The PCI setting that said 32 might be the "PCI Latency" or whatever it's called, which is something different.

I have some old pictures of a P2B, so I've attached a photo of the board markings that describe how the PCI clock is controlled. It's part of the same set of jumpers that also sets the FSB speed.

Actually, it looks like the only setting they have documented on Rev1.02 for 100FSB is the one that also puts PCI at 33MHz, so if your settings match that line then it's correct. But if it doesn't match that line, there probably is an undocumented setting that would cause the PCI to be overclocked.

I suspect this isn't your problem though, I'm more suspicious of the AGP clock.
I didn't find a picture of the jumper for the AGP setting, but as I recall there is a separate jumper somewhere which toggles the AGP clock between 1/1 and 2/3 of the FSB speed. At 100FSB you'd need to set that to 2/3 so the AGP clock will be at 66MHz. Although it does support AGP2X transfers, the AGP clock is still supposed to be at 66MHz. If the AGP is at 100MHz it would probably be very unstable.

Just checked google for images of that jumper and to my disappointment the board lacks such a feature.

It's just a plain P2B with no extra letter, revision 1.02.

Due to a user mentioning bad capacitors, I went on to check them and found only one which could be suspicious; it has a slight bump on top of it.

I'll probably make a video or post images of it soon, I'm desperate to get this thing fixed.

I must be wrong about the existence of the AGP jumper. I looked in the manual and don't see it mentioned either. Maybe it's a hidden/implied part of the FSB jumper setting.
I wish I could double check what I'm remembering. I have one of these boards somewhere but I'm not sure exactly where. (I stored away some stuff in boxes a few months ago)

Have you tried running a Prime95 torture test? Try leaving that running for a few hours. This will put stress on the CPU, chipset, and RAM but not the video card. It should also do a pretty good job of stressing the power supply. The results could help reinforce whether this is actually a graphics problem or a more general system instability problem.

Do you have a multimeter, and feel comfortable using it to measure voltages? If so, try measuring voltages not only at idle but also when under stress. The more voltages you are able to measure, the better. Maybe you'll find something that's looking off.
The easiest to measure are the +5V and +12v from the PSU. Those can be measured in one of the molex 4-pin drive connectors.

What brand and model of power supply are you using? Can you post the specs from the label?

If you remove all cards and components except the minimum required to test (only 1 RAM module, the graphics card, and the OS drive) does it run any better?
If you do this and also reduce the FSB to 66MHz, does it run any better?
Finally, while still in this minimal state, I'd also try resetting the CMOS to defaults just in case there's some unknown issue there.

First off, you're not entirely wrong about the AGP jumper, they do exist on older motherboards, I just happen to have the very first revision of it on which it is absent 😒

Since my last reply I've installed Windows 95 OSR2 which, strangely enough, works perfectly. Had to look for different Ethernet drivers, though.

I've never heard of Prime95 before. I'll definitely try it out, though.

Regarding the dumped-down configuration, I've tried running it like you suggested above (graphics card only, HDD and 66Mhz FSB) when it still had Win98SE, but with all the RAM modules, because they all passed according to Memtest86. It didn't help.

As for the PSU, this is what I'm using: http://www.electromyne.com/images/131901a.jpg

I even tried running it with the Corsair CX430 from my main PC which also didn't help.

Windows 95 also reports the card to be "PCI based" and refuses to allow AGP acceleration in DirectX as opposed to Win98SE, which did.

All in all I assume either the board or the CPU is faulty and I'm on the idea of purchasing an HP Vectra Slot 1 board which happens to be the board of my previous, long gone Windows 98 PC used.

I have an ASUS P2B Revision 1.02 board and it's working perfectly fine. FSB is set to 100MHz, the CPU Multiplier is set to 6x (for a 600MHz Pentium III "Katmai" Slot 1), PCI @33 MHz and I assume AGP @66 MHz. I'll look into my system to verify if there are jumpers or an AGP setting, but it only runs at AGP 2x anyway. I have a Diamond Viper V770 RIVA TNT2 32MB AGP card, 2x STB V2-1000 Voodoo2 12MB PCI cards in SLI (working), 10/100 PCI Ethernet, Creative Labs Sound Blaster AWE64 ISA card, a 40GB Western Digital hard drive, and a CD-RW drive running Windows 98 SE.

Sorry I've been kind of slow to reply. But if Win95 OSR2 is working, then I guess it must be a software issue somehow. Either that, or maybe Win95 isn't using AGP - I've had a couple situations before where installing an AGP driver triggered instability, but without such a driver, a system was stable. In at least one (not sure about the 2nd) of those cases I had a faulty graphics card, but I know you've tried multiple cards here.
Based on Win95 working, it tends to point away from power. But as far as the PSUs are concerned, I found a closeup picture of the label for the Linkworld LPD6-350WP and it doesn't specify a max combined wattage for the +3.3V and +5V rails. However, based on the fact that it claims higher amperage on those rails than for the +12V rail, it implies it's probably designed for this era of system. I don't know how reliable or well designed it is, I'm not familiar with it.
The modern Corsair CX430 expects most of it's load to be on the +12V rail, and is only rated for 110W total between the +3.3V and +5V rails. In principle that's certainly enough for steady state operation, but I'm not sure how reliably it would power a late 90s system while under load, but if the configuration was stripped down then it was probably enough. The other question in my mind is whether those rails are independently regulated (on some PSUs, if 3.3v and 5v are unevenly loaded then their regulation can drift in opposite directions).
Watching voltages with a multimeter helps to remove some doubts when it comes to possible power issues IMO, but still, given that Win95 OSR2 isn't having a problem, it suggests the problem lies elsewhere.
I wonder if perhaps the AGP slot isn't getting good power for some reason, but I'm not sure how the P2B powers it. I don't know if it runs the PSU's power rails straight to the slot, or has some onboard circuitry involved.
I definitely suggest trying Prime95, because it creates a scenario that will stress the system but without stressing the graphics.

It's OK, I understand. Just want to address that Win95 actually reports the card as being incapable of AGP acceleration in DirectX while it's enabled by default on Win98. None of my games have crashed on it so far either.

I'm now on the idea that it's the processor's fault. It's a Pentium II 350 with decent coolers. I'm not sure what lies beneath and I don't want to touch anything as I don't have any thermal paste to reapply. (although I'm planning on buying a tube of it)

I've confirmed that it's half software half hardware related (Win95 and 98 can work but 2000 crashes after driver installation every time) and that it's not the HDD, the RAM modules,the internal cards or the jumper settings causing it.

When I ran Windows 98 with later (Detonator 20 something and up) drivers it hung mostly after menu animations and during gameplay. It also uses twice as much RAM as Windows 95 by default, and 56MB of the page file.

I'm about to try another Win98 ISO I got, since the previous one was probably pirated as I got it from a torrent 😒