Possibly a weak component in the CPU VRM, either one or more caps are not holding charge or another VRM component is weak. I would try and test each part with a multimeter to see if its within tolerance for that part, starting with the larger caps first and working down to the smaller ones.

Also check the PSU voltages to make sure its delivering the right voltages and is a stable delivery.

While lots of 440 boards have 133Mhz fsb setting, the 440BX chipset does not officially support 133MHz fsb - anything over 100MHz is considered an overclock. It's entirely possible your board is not stable at 133Mhz.

That aside, all 440BX boards I've tried that have a 133Mhz fsb option ran fine at 133 (apart from the lack of a proper AGP divider) - so there might actually something physically wrong with your board. Bad caps maybe?

Not running at 133, running at 100 or 112MHz for exactly that reason

Socket3 wrote on 2023-06-07, 17:34:

While lots of 440 boards have 133Mhz fsb setting, the 440BX chipset does not officially support 133MHz fsb - anything over 100MHz is considered an overclock. It's entirely possible your board is not stable at 133Mhz.

That aside, all 440BX boards I've tried that have a 133Mhz fsb option ran fine at 133 (apart from the lack of a proper AGP divider) - so there might actually something physically wrong with your board. Bad caps maybe?

From my experience only boards that have FSB options above 133Mhz are stable at 133 FSB. My ECS P6BXT-A+ is not stable at 112 either. There is some instability with heavy hard drive usage in Windows at that speed. Cheap boards were by design unstable at anything higher than 100Mhz FSB. It isn't a major problem though as there is PIII 900E in sufficient numbers.

It isn't a good idea to put too much stress on these old boards. You will likely need to replace caps and vrm as stated previously.

As in my first post, it is stable at 100MHz AND 112MHz with the pentium 3 450 and the pentium 2 233.

At 100MHz is unstable with the 933 and the 1Ghz (but used to be stable at 100MHz and 112MHz).

Sounds like everyone else thinks caps/vrms too.
Will have to do some disassembly at the weekend!

not all BX boards run at 133Mhz FSB ... haha just kidding. I like how you had to repeat 3 times what clocks you are running :]

- Coppermines are really sensitive to Vtt (1.5V). Some boards didnt have good Vtt supply. Fixing ASUS P2B freezes on memtest86+ modulo 20 test
- doesnt matter how capacitors look, you need to measure voltage ripple/sag or individual capacitor ESR. Sadly you cant do this with a multimeter as Trashbytes suggested

I dotn think its explicitly a power draw issue as Coppermines usually use less power. On the other hand Voltage might have something to do with it. Older chips start at 2V and can tolerate more drop. I would also try 66MHz. You could try setting 1.85-1.9V Vcore on the slotket for the celeron.
all in all best bet would be putting oscilloscope on Vcore, Vtt and 3.3V rails.

rasz_pl wrote on 2023-06-08, 09:48:

not all BX boards run at 133Mhz FSB ... haha just kidding. I like how you had to repeat 3 times what clocks you are running :] […]
Show full quote

not all BX boards run at 133Mhz FSB ... haha just kidding. I like how you had to repeat 3 times what clocks you are running :]

- Coppermines are really sensitive to Vtt (1.5V). Some boards didnt have good Vtt supply. Fixing ASUS P2B freezes on memtest86+ modulo 20 test
- doesnt matter how capacitors look, you need to measure voltage ripple/sag or individual capacitor ESR. Sadly you cant do this with a multimeter as Trashbytes suggested

I dotn think its explicitly a power draw issue as Coppermines usually use less power. On the other hand Voltage might have something to do with it. Older chips start at 2V and can tolerate more drop. I would also try 66MHz. You could try setting 1.85-1.9V Vcore on the slotket for the celeron.
all in all best bet would be putting oscilloscope on Vcore, Vtt and 3.3V rails.

I only suspect a VRM issues as the PC did initially work fine with both the CU CPUs @ 112 Mhz, I suspect the age of the components and the high draw of the CU CPUs may have nuked one of the components putting it out of spec enough to cause issues with the CU processors.

Personally I would be removing the Caps and checking each one with a component tester or just recapping the VRM, along with doing the testing you suggested to rule out deeper issues.

It sound like a lot of work but its a nice BX board and would be worth doing recapping if only to keep the old girl working for many years to come.

Took the PC apart, inspected the motherboard, looked at the caps etc.
No signs of bulging or any signs trace corrosion etc.
Checked caps with a MESR100 ESR meter and all read fine (though the power caps around the ATX power connector are all very close and all in parellel, so not sure I really trust the readings there -- that said similar ones more distant around the ISA and PCI card slots all read fine too)

The voltage on a group of the power caps is the same as vcore (and changes with different CPUs matching their vcore voltage) using a hanktek6022BE which doesnt do AC so not great for ripple measurement (?) but showed peak-to-peak variance of around 40mV on the 933 and about 20-30mV on the P450 - all measured with just the CPU, RAM and a POST card connected.

Reconnected video card and the system seemed to boot fine with the 933. All symptoms before disappeared, Replaced all the other cards one by one, added back drives one by one - everything seems fine and stable.

All very odd!

Anyway, will see how long it lasts!

Do those voltage measurements sound like they should?

Connect the oscilloscope to a battery powered laptop with the AC cable physically disconnected. It's never clear if it's power noise or common mode noise. Use only Ground Spring for probing.

I would change all the small electrolytic capacitors near the socket, memory and chipset - like 10-100uF.
Check Vtt voltage.
There was a forum thread about an MSI-made slotket with a voltage clamp chip that did not support 133FSB.
Presumably because of it`s.
PS. It is hard to believe that large electrolytics have an ESR of 15mΩ after 25 years.