My PC randomly (once a week) turns off and I cannot turn it on or reset via the button until turning off and on the PSU. I think there is still some light from the case etc. at that time.
So far it was only happening at idle or even during the night.
No overclocking except enabling EXPO on the memory.
The PC runs Linux but it’s probably irrelevant. Nothing unusual in the logs, but it definitely turns off because when it happens at the middle of the night there are no log entries for several hours.
I updated the BIOS last week, the same on the latest and June.
ASUS PROART X870E
AMD 9950X3D
Nvidia RTX 5080
2 x 48 GB DDR5 6000 MHz
Seasonic Vertex 1200W
It is connected through APC Back-UPS Pro BR1600SI but the UPS does not turn off or anything, and other devices connected to it work fine.
It can also be overheating and that is why you need to “reset” it before it starts up again.
So you can try to clean it out and then try again. ALWAYS DISCONNECT THE PSU FROM POWER AND DISCHARGE IT BEFORE DOING ANYTHING!
The are current inside a psu that is DC, and DC levels required to kill is not particularly high.
Unplug it from the wall socket and touch both stifts on something that is grounded to discharge. To be even safer, wait 30 mins after that before opening it to clean.
When a psu is having issues, there are all kinds of strange things that can happen, but if it shuts off and you can’t restart it with the power button like you would be able to if it shut down due to overloading for example, most of the time it is due to overheating. Overheating can happen if something is wrong inside the psu, not only because the load is high.
Whatever you do, just be very careful if opening it, I’m not joking.
Please, no one do this as it is very dangerous (and incorrect) advice.
There is no guarantee that grounding out the input will discharge all of the caps and 30 minutes is most certainly not enough time for them all to naturally bleed off. Even pushing the power button on the PC side doesn’t guarantee that the board will drain the very separate 12V, 5V, & 3.3V rails. You would need a bleed off capacitor or LED with long wire legs and go around to each individual cap and make sure they’re completely drained. Absolutely not worth it when you can just leave the cover on blow air through it.
On that note, please don’t take the cover off either. It’s kind of on there for a good reason, just like the one on your microwave.
It’s true, and you are correct. The cable could be damaged too. But it’s 99% secure.
I have worked as an electrician, and I assumed the cable was safe since it actually powers, and if that is true, so is the grounding because if the computer is not grounded, it dies, and pressing off and on on the switch does not involve ground.
But good catch, always safer to say as it is at 100%.
Discharging with a dedicated ground stick directly to ground points on the PSU is the actual correct way to do it.
Just figured most people don’t have that kind of equipment available and this is pretty darn safe method.
And about the discharging. As long as the psu has an IEC certification (ie bought after mid 90:ies, at least in europe) it will be discharged within 30 seconds, or it would not be certified.
No. Corsair HXi does, though, and there’s a few other USB connected units which probably do as well.
What you’re describing isn’t unusual in my experience of AM4 Asus boards, can’t say about AM5 as their AM4 bugs and ongoing bloatware-security issues are bad enough I haven’t built Asus lately. Or Windows 11 in general, if Microsoft is applicable, so I wouldn’t jump to the assumption it’s the supply.
Seasonic is not particularly known for competency, however, and is known for inability to admit having a similar problem with false protection triggers. But, while that was an idiotic pre-Vertex voltage sense design fail on their part, it pretty much happened only with dGPUs under load.
How could they break the power on/reset button from the software?
But is there anything better? Asrock is killing CPUs this year, Gigabyte is also not known to be great + e.g. leaking some thermal thing from the GPUs now.
ASRock, MSI, plausibly Gigabyte. Maybe Sapphire, NZXT, Maxsun, … We tried to build a case ASRock’s Granite Ridge failure rates are actually elevated in the thread for that but were unsuccessful.
Generally the problem that comes up around the L1 forums is ASRock and MSI don’t have direct ProArt competitors. So mostly either you accept different specs, throw Gigabyte dice, or put up with Asus. X870 Taichi Creator has 5+10 GbE, though.
FWIW I just ordered B850 Edge for 9950X + 2x48.
Potentially just takes something on the mobo getting into a weird state. The latest turn of this Gigabyte thread I think isn’t 11 specific but is one such example. Similarly, I did a build with Cybercore 1000 and B550 Steel Legend a while ago where if you don’t wait for the power supply relay click on shutdown the board won’t turn back on.
We have a fair bit of hardware that ran fine on 10 and, now that it’s on 11, just sometimes isn’t up in the morning. Same kinda deal as you’re seeing, just no logs for a good part of the night. There’s a couple cases I’ve traced to Microsoft making 11 changes but failing to push required accompanying AMD chipset driver updates.
Typical Redmond fail. Nvidia’s poor drivers might be contributing in some cases but most of the GPUs are AMD, so can be excluded as a sole graphics factor. Most of the drives are DRAMed, so same for Microsoft’s HMB issues. ¯\_(ツ)_/¯
