CPU Error related to powercuts?

[deadcpuguy]

I have run across a very strange phenomena with my PC that seems related with mains power cuts.

I have a PC with an Acer P8Z77-V MOBO with a Intel i3-3225 CPU.

After a power cut the PC was completely dead when I tried to re-start, the CPU error LED on the MB was lit.

I removed and re-installed CPU to no avail. Also reset CMOS RAM with no effect. But after switching the system off for a longer period of time (1hr without mains connected) it started again.

I had these strange CPU fault two times earlier, but thought it had to do with a bad contact. Now it seems to be more clear: the problem starts with a power cut and the system seems to "recover" after some rest.

I can not make the connection between this 'rest-period' and a CPU fault or bad contact. It seems more likely a dodgy capacitor somewhere...


Any ideas what might cause these strange effects?

I'm about to order a new CPU, but not sure if that will solve the problem.

Thanks

Peter

 

[Eximo]

A somewhat common symptom. The computer doesn't know if the incoming power is good or not, so it refuses to POST. Shutting it down for a while will clear this fault as the capacitors in the system drain out and no information about the previous state can be stored. Nothing to really worry about. Though if you don't have a surge suppressor that the computer is plugged into, invest in one.

 

[deadcpuguy]

A somewhat common symptom. The computer doesn't know if the incoming power is good or not, so it refuses to POST. Shutting it down for a while will clear this fault as the capacitors in the system drain out and no information about the previous state can be stored. Nothing to really worry about. Though if you don't have a surge suppressor that the computer is plugged into, invest in one.

Hmm, you might be very well right!

Just while eating my dinner and pondering about it, I thought: I never switch off and on within 2 minutes or so. This problem might only come to my attention because I *do* switch ON/OFF within 2 minutes when a power cut occurs!

Some questions remain though:
Is this a thing of the power supply then?
I know they often have a "POWER GOOD" signal for the MOBO.

And why would the MOBO show this as a CPU fault?
The MOBO makes it's own low voltage supply and has its own power-good signal?

Thank you for your help and I hope you can answer my remaining questions concerning this as well!

Peter

 

[Eximo]

Computers do store a last known state. In the BIOS this can be configured so that if the power goes out the computer will automatically try to remember its last powered state. If it was on it will boot again, if it was off it will stay off, being the defaults. You can also set it to stay off after power failure.

What may be happening in this case is that the power sags or spikes briefly. The PSU will interpret that as bad power and maybe prevent its on internal POST before the motherboard even has a chance to try starting. If it isn't getting 12V from the power supply, nothing to run the CPU. The BIOS usually runs off of 5V, which via Standby power might be getting to the board.

I still say don't worry too much about it.

 

[smorizio]

with newer mb make sure the cmos battery is good. if it weak a lot of these new mb wont post. on the cpu red led. that led can come on if the cpu is bad or if the eps power to the cpu/mb is not there. intel has a free cpu test program if your cpu passes try a test power supply next time. another issue can be fake power surge. on some pc with nvidia gpu the nvidia gpu were tripping asus power surge till asus dropped a bios update.

 

[deadcpuguy]

with newer mb make sure the cmos battery is good. if it weak a lot of these new mb wont post. on the cpu red led. that led can come on if the cpu is bad or if the eps power to the cpu/mb is not there. intel has a free cpu test program if your cpu passes try a test power supply next time. another issue can be fake power surge. on some pc with nvidia gpu the nvidia gpu were tripping asus power surge till asus dropped a bios update.

I will look for the test program. I don't expect really that the CPU is faulty, because once it starts it can be on for hours without any problems, but it is lways nice to see what these test programs come up with.
Thanks for your tips, the battery has been replaced less than a year ago, so that should still be ok, I'll check to make sure.

 

[deadcpuguy]

Hi, although the first proposed explanation (Eximo) seemed to be very likely true, the issue has not been solved. Indeed, when I wait more than 5 minutes before switching ON, the system works,

but...

When I re-start from my OS (Linux) or from the BIOS setup, the system resets and either switches off (so I can not switch on for 5 minutes) OR it tries to restart, but hangs with the CPU error LED lit.

So there is definitely something fishy with the POST or reset circuitry.

Any further suggestions or advise would be highly appreciated.

I have checked CMOS battery : 3V
I downloaded the Intel CPU test program, but I can only run that from within a Virtual Windows PC, which gives me a error (Memory management unit?) which seems kinda logic since VirtualBox will manage memory and probably won't allow low-level access to mem. management.

For Linux I found a CPU stress-test program (stress-ng), this ran without problems for at least an hour whereby CPU core temp. rose to 73C.


Thanks,

P.

 

[Eximo]

I have had ACPI functions fail on a motherboard before. Just dealt with it until I could replace it. Was rather interesting, had to run two copies of Windows, one with it off and one with it on. It would intermittently flip back and forth. So I installed my most commonly used programs in the second OS for when it was misbehaving.

Not sure you can still turn those features off to be honest. Also not sure how Linux behaves with it on or off. I know with Windows it meant an OS re-install.

This would be back to "It is now safe to turn off your computer" style shutdown.

Replacing the power supply is an option. Might be causing a spike or sag as it shuts down which trips up the POST. But if that doesn't do it, probably the motherboard.