The Strangest Fault I've ever found!

[Arbitration]

Don't laugh. My question is about a Pentium 4.
I'm not a gamer and I'm a believer of "If it ain't broke don't fix it."

My XPC Media Centre was a P4 3.4Hz (although it started life as a Celeron D 2.8). It worked happily for several years until the PSU took early retirement.

I installed another PSU and the computer was happy again - for about 3 months. It began crashing within 10 seconds of booting. On the rare occasion you could get it to boot it would run fine until it was powered off. Then came the day when it would boot no more.

Not a problem - a new media centre was overdue.

Here comes the weird part . . .

Bored, about to throw the old PC away. I installed the old Celeron D - for something to do. Hey presto - it boots and runs fine.

The obvious solution here is that the P4 was fried.

Tried another P4 (3.4) - the old problem returned. Back to the Celeron - it works fine. (I can even overclock the Celeron to 3.4).

I only want to solve this because it's a puzzle. What could the problem be?

 

[okcnaline]

Then that's something with the motherboard...

 

[JobCreator]

See if there's a BIOS update.

 

[PEJUman]

your system is very old, therefore the could be multitudes of failure point. Based on your symptoms, I would guess your power regulator on the motherboard is on it's way out. it maybe only able to handle the ~75W or so the Celeron D requires (if you overclock without voltage increase, the power draw would only scale linearly with clock speed). When you swap to the p4 3.4 (I am assuming Prescott core here), you overwhelmed the system with ~85W power draw.

 

[Arbitration]

Maybe you are right. It's a Cedar Mill (86w). I can't see it drawing that much more power than the overclocked Celeron.
The current CPU voltage is displayed as 1.31 volts.

The most confusing part of this problem is that when the problem first developed . . . If you got the P4 booted it would stay booted and run fine for days.

I've tried reducing overall power consumption by removing the graphics card - makes no difference.

The only area in which the two processors are different is fsb 533 v 800 . . . I don' know if it will affect anything.

 

[PEJUman]

Maybe you are right. It's a Cedar Mill (86w). I can't see it drawing that much more power than the overclocked Celeron.
The current CPU voltage is displayed as 1.31 volts.

The most confusing part of this problem is that when the problem first developed . . . If you got the P4 booted it would stay booted and run fine for days.

I've tried reducing overall power consumption by removing the graphics card - makes no difference.

The only area in which the two processors are different is fsb 533 v 800 . . . I don' know if it will affect anything.

The graphic card should draw from separate power lines, it should bypass the CPU Voltage regulators, so removing it would not help weak CPU VR.

but your staying booted part is quite odd. That makes me think the FSB is the issue. Typically there is an adaptive scheme at boot for the CPU-CHIPSET communication, aimed at correcting timing the FSB interface. Maybe the weak part here is the motherboard chipset itself; it no longer able to operate/train itself for proper timing at the faster P4 FSB.