PC turns off unexpectedly when under (seemingly) high cpu load, but rarely

[Nostras]

Hi folks,

Since about a year I've been having an odd issue where my PC would just randomly turn off.
Specifically, to turn it back on I need to remove the power for a bit, then replug it and it works again.
I should mention that technically replugging the 24-pin cable does the same thing, so it may be a "panic" button of the motherboard, not the PSU.

If you're asking "wow you've had this for a year? why post now?", well, that's because it's exceedingly rare. I think I've had it happen about 4 times over a full year. All in somewhat CPU demanding games except for once in rocket league with fps cap (165, uncapped is 600+).

But what changed everything is that I have a rather good way to trigger the sudden power off, I just need to install a very large skyrim modpack with Wabbajack.
I'm not joking.
In about 3 installs I will know if it is stable or not (which it never has succeeded 3 times in a row).
It tends to crap itself in the final step, Building BSAs, which is a very high load on the CPU and some on the memory/disk.

For my build:
- X670E-I
- 7950X3D
- 7900XTX*
- 64GB DDR5 (KF556C40BBK2-64)
- SF750*
- EK 280 RGB AIO
- Dan C4-SFX*
- SN850X 4TB
- SX8200 1TB
- 660p 2TB (via NVME -> USB adapter, LC-Power LC-M2-C-NVME-2X2)

Of course I've tested some things. Of note is that I have a separate AM5 NAS which I can use as a donor (8700G, B650M Lightning, KF560C40BB-32, RM550x).
I've marked a couple with a star, I've checked these with the donor PC or removed altogether.
The 7900XTX I've simply removed from the system and done the stress with just the iGPU, similar results.
SF750 has been replaced with the RM550x, similar results.
I've removed the PC from the C4-SFX and put it on an open bench, so out of scope as well.

If I remove everything I've also tested or what I think is not relevant we're left with:
- X670E-I
- 7950X3D
- 64GB DDR5 (KF556C40BBK2-64)

The last one I can check with the donor PC as well, which is what I'm doing as I write this post.

Oh, and yes, of course, I've done a bios reset.

Anyone got some ideas? I think it's gotten to the point that I need to isolate the motherboard/CPU which is frankly not something I'm looking forward to, but I can't really fault any of the other parts.

 

[brutalame]

Usually, the problem of : " I need to unplug my computer to be able to restart it" is tied to a faulty PSU

 

[xorbe]

My PC was randomly turning off under gpu load every 1-2 weeks after I switched video cards . I swapped PSUs, and it hasn't happened since, but it's not 100% proof.

 

[Nostras]

Usually, the problem of : " I need to unplug my computer to be able to restart it" is tied to a faulty PSU

My PC was randomly turning off under gpu load every 1-2 weeks after I switched video cards . I swapped PSUs, and it hasn't happened since, but it's not 100% proof.

Understandable guys, but as mentioned I've swapped my SF750 with an RM550x and this did not affect the problem whatsoever.

I tried with the NAS RAM and that borked just as hard.
Unless someone has a great idea to identify the culprit I'll have to swap motherboard/CPU.

Would be annoying to RMA the motherboard considering the VRM header and fan wiring are broken. Appears that never worked and when I tried to replug it the PWM wire came loose and a part of the female receptacle came loose. It does work now but at 9000RPM is highly unpleasant.
I'm afraid ASUS will blame me for that if I try to RMA it, how the hell do I proof it has been like that since day 1.

 

[Noci]

That are some serious hardware specs, they need a lot of juice especially when stressed.

According to the Bequiet! PSU calculator, with a non OC your system already uses 85% of a 750W PSU capacity. The Corsair RM 550W PSU is probably not even capable of delivering soo much stable power.
You can try to borrow some more powerfull PSU from a friend to see if that solves the problem.
If not possible to borrow somewhere, I would defenitely advise to upgrade the PSU.

 

[Nostras]

That are some serious hardware specs, they need a lot of juice especially when stressed.

According to the Bequiet! PSU calculator, with a non OC your system already uses 85% of a 750W PSU capacity. The Corsair RM 550W PSU is probably not even capable of delivering soo much stable power.
You can try to borrow some more powerfull PSU from a friend to see if that solves the problem.
If not to borrow somewhere, I would defenitely advise to upgrade the PSU.


View attachment 347676

I haven't used the XTX in any of the stresstesting. Checked with a power meter as well, at worst about 200W without the XTX when Wabbajacking. Is that a verb? Maybe.
And stupidly enough the PC is doing A-OK when stressing the CPU + GPU in your typical power viruses, it's just that the power off is unexpected but only occurs rarely when not idle and easier to reproduce as mentioned in the TS.

 

[Noci]

I haven't used the XTX in any of the stresstesting. Checked with a power meter as well, at worst about 200W without the XTX when Wabbajacking. Is that a verb? Maybe.
And stupidly enough the PC is doing A-OK when stressing the CPU + GPU in your typical power viruses, it's just that the power off is unexpected but only occurs rarely when not idle and easier to reproduce as mentioned in the TS.

I don't need to know if you are jacking your Wabba .

So during high CPU load combined with heavy data transfer, did you by chance monitor the temps when reproducing the problem?

 

[Nostras]

I don't need to know if you are jacking your Wabba .

So during high CPU load combined with heavy data transfer, did you by chance monitor the temps when reproducing the problem?

I do yes, as the X3D loves it it immediately slams against the 89°C and 95°C thermal limits of their respective CCD's.
Everything else is fine, but I did have to put a fan on the CPU block as the AIO fans are no longer creating airflow over everything on the motherboard.
Chipset got a nice toasty 90°C without that (to note, it didn't go much above 80 in the C4 because of the AIO).
Arctic's AIO fan on the CPU block would've been highly appreciated, truly ahead of their time.
I thought temperature was suspect, but the fan reduces most toasty ones to about 70°C which should be well outside of the danger zones.

 

[micropage7]

btw have you check the bios setting, looks like when the CPU dragging more power then it shutdown

 

[Nostras]

btw have you check the bios setting, looks like when the CPU dragging more power then it shutdown

That's the thing, it doesn't really seem to do with power consumption. It's more related to a specific workload, or at least way more likely to happen.
But I'm not sure what Bios setting you're suggesting I look at.

 

[Noci]

Despite the fact that you say the power draw is not that high, I would really focus first on improved cooling.
The Arctic Freezer fan on the pump would not make much of a difference as you say the VRM fan is already spinning a 9000 rpm.

Still looks to me like a combination of power and temps that trigger some sort of saftey feature as it happens only during intense compiling with Wabbajack.

 

[Nostras]

Despite the fact that you say the power draw is not that high, I would really focus first on improved cooling.
The Arctic Freezer fan on the pump would not make much of a difference as you say the VRM fan is already spinning a 9000 rpm.

Still looks to me like a combination of power and temps that trigger some sort of saftey feature as it happens only during intense compiling with Wabbajack.

You misunderstand, I unplugged the VRM fan as 9000 rpm is beyond annoying.

But today I'll check it with a clean Windows install and if that doesn't change anything I'll swap the motherboards.

 

[Madmackz]

The tm550 you swapped to is not nearly enough power which would have created the exact same issue . Before I went and blew any funds on it at all I would be firing up HWINFO Logging and then running a benchmark, also i would be checking through evet viewer as well as , Reliability History. You can narrow down exactly why your issue persists without much effort. I highly doubt a motherboard cpu swap is in need. My system was doing this for weeks, and it eventually led to my AIO pump getting jammed and it would spike the temps momentarily, were talking split second and my machine would reboot no blue screen nothing just gone.