Question PSU voltage outside 4% killed CPU?

NathanShirley

Junior Member
Sep 24, 2021
10
0
6
My computer started cutting off randomly, sometimes with a BSOD, but mostly straight to black. Usually I could get into Windows before it would die, but other times it wouldn't even make it into the BIOS. Running Linux off a USB also ended up dying.

I tested the PSU voltages and the 12v was a good bit out from the 4% tolerance I've read it should be within. The PSU was a bit old so I replaced it with a used one. Same problem. I checked this PSU and it was also a bit out from the 4%, but not quite as far (only the -12v I believe). The first PSU was a Corsair CW600, the other, some EVGA 450w. I then tested a really cheap generic PSU I had which was further out than the other two (both 12v and 5v I believe). I took a 400w PSU out of my kids' computer and tested it--all voltages within 4%. I installed that in my computer...and it died as before.

Apparently the problem was the CPU had gone bad (i7 3770). Replacing it with an older CPU (Pentium 4) fixed the problem.

My question is, could the 12v being a bit out have killed my CPU? And would it be a terrible idea to keep running this with a PSU that's a little out from the recommended 4% tolerance? It's an older machine so I don't want to throw too much money into it.
 

NathanShirley

Junior Member
Sep 24, 2021
10
0
6
Ha, no not trolling...I just got mixed up. I double checked and it's a Pentium G2030 at 3 GHz. I still had it laying around from when I upgraded the computer my kids use, years back. The motherboard is a P8H67-M LE.

So if the CPU does run on 12v there is a chance my PSU killed it, right?

The reason why I'm wondering all this is because I now have two PSUs that don't output 12v within that 4% tolerance range (plus a really cheap PSU that I am pretty certain was killing my HDDs a long time ago, which I'll definitely get rid of). I checked voltage with each PSU disconnected from the motherboard, with a paper clip to power them on. I ordered a new PSU for a new system I'll be putting together, but am trying to figure out if running this old system with one of these two PSUs is a bad idea. At this point I'm guessing it is, but would it be worth checking them under a heavier load? I'd have to look up a good way to do that.

Thanks for the help.
 
Last edited:

NathanShirley

Junior Member
Sep 24, 2021
10
0
6
By the way, my -12v measured at -11.37 VDC. I read it should be between -11.52 and -12.48.

EDIT: It seems there's a lot of conflicting info out there...I've also read a +/-5% tolerance should be okay, with a +/-10% tolerance for -12V. Link here:


The main guide I was following was the video here:


So based on those lifewire tolerences, all my PSUs might actually be fine. I'll have to retest to be sure.
 
Last edited:

mindless1

Diamond Member
Aug 11, 2001
8,028
1,440
126
So if the CPU does run on 12v there is a chance my PSU killed it, right?

It is very unlikely. The VRM supplying the CPU can generally tolerate 13V or more, and if you have a PSU problem the more likely issue is it doesn't have a fast enough rise time to compensate for a sudden increase in CPU load so instead of damage, undervoltage just causes a crash. It is more likely that a poorly installed heatsink, or bad motherboard killed the CPU, but there is some tiny chance the CPU itself had some manufacturing flaw instead.

The reason why I'm wondering all this is because I now have two PSUs that don't output 12v within that 4% tolerance range (plus a really cheap PSU that I am pretty certain was killing my HDDs a long time ago, which I'll definitely get rid of). I checked voltage with each PSU disconnected from the motherboard, with a paper clip to power them on.

That is an invalid test. PSU manufacturers providing full specs, have often spec'd what the minimum load on each rail needed to be to maintain regulation, and even then, certain PSU designs are subject to cross loading issues but with no load, the PSU is only depending on the internal design to keep it from becoming completely instable.

It is not abnormal at all for some PSU without a load to fall outside of specs, and unsurprising that many PSU now have minimal internal resistance to load the rails because once hooked up to a system that is essentially wasted power. The most relevant measurement is using a multimeter at the power connector, powering the system at various load states or of course an oscilloscope which also catches momentary highs and lows in the ripple.

I ordered a new PSU for a new system I'll be putting together, but am trying to figure out if running this old system with one of these two PSUs is a bad idea. At this point I'm guessing it is, but would it be worth checking them under a heavier load? I'd have to look up a good way to do that.

Nothing mentioned so far indicates to me that either PSU is bad, but of course the older a PSU gets, the closer it is to end of life. I would pop the case off the PSU, examine for burst capacitors, clean dust out, and lube the fan while you're at it if it is a sleeve bearing fan (any bearing "name" besides dual ball bearing).

Your negative (-12V) 12 volt rail being off has no consequence at all. Nothing these days uses it, back a long time ago, integrated audio did, and RS232 serial ports before their drivers became able to do themselves, but even an imbalance on that for audio, had no effect because the resultant DC offset between it and the positive 12V rail would be eliminated by coupling capacitors on the output.
 

NathanShirley

Junior Member
Sep 24, 2021
10
0
6
Thanks so much for taking the time to reply, that's all really great information, very helpful.

I'll see if I can test these PSUs under various loads. I don't have an oscilliscope, but do have both a digital and analog multimeter (I've read analog can be good for looking for sudden changes in voltage). Thanks for the tips on PSU maintenence, I'll be sure to do that too. I did open one to see if the capacitors looked alright.

I bought the CPU used about 5 or more years ago, and was using the original Intel heatsink...maybe that was it. I never noticed it running hot, but hadn't checked it in a really long time.

Thanks again, I greatly appreciate it.

It is very unlikely. The VRM supplying the CPU can generally tolerate 13V or more, and if you have a PSU problem the more likely issue is it doesn't have a fast enough rise time to compensate for a sudden increase in CPU load so instead of damage, undervoltage just causes a crash. It is more likely that a poorly installed heatsink, or bad motherboard killed the CPU, but there is some tiny chance the CPU itself had some manufacturing flaw instead.

That is an invalid test. PSU manufacturers providing full specs, have often spec'd what the minimum load on each rail needed to be to maintain regulation, and even then, certain PSU designs are subject to cross loading issues but with no load, the PSU is only depending on the internal design to keep it from becoming completely instable.

It is not abnormal at all for some PSU without a load to fall outside of specs, and unsurprising that many PSU now have minimal internal resistance to load the rails because once hooked up to a system that is essentially wasted power. The most relevant measurement is using a multimeter at the power connector, powering the system at various load states or of course an oscilloscope which also catches momentary highs and lows in the ripple.

Nothing mentioned so far indicates to me that either PSU is bad, but of course the older a PSU gets, the closer it is to end of life. I would pop the case off the PSU, examine for burst capacitors, clean dust out, and lube the fan while you're at it if it is a sleeve bearing fan (any bearing "name" besides dual ball bearing).

Your negative (-12V) 12 volt rail being off has no consequence at all. Nothing these days uses it, back a long time ago, integrated audio did, and RS232 serial ports before their drivers became able to do themselves, but even an imbalance on that for audio, had no effect because the resultant DC offset between it and the positive 12V rail would be eliminated by coupling capacitors on the output.
 
Last edited:

NathanShirley

Junior Member
Sep 24, 2021
10
0
6
Well you were right. In the end it turned out to be the motherboard that was going bad. What exactly went bad on it I may never know. I found a similar used motherboard (needed to rescue my Intel Raid 10) and the CPUs appear perfectly fine, as do both my PSUs.
 

Leeea

Diamond Member
Apr 3, 2020
3,599
5,340
106
Yes, ever since ATX 2.0 it does.
No, it does not.

Even with ATX2.0 the mainboard down converts it to the exact correct voltage, the input voltage on the 12v DC line can swing around a bit without effecting things.

Unless of course the mainboard is going bad, at which point still not a PSU problem.