My computer has lost its mind and I can't even begin to diagnose what's at fault. (It's the power supply.)
UPDATE: It was the power supply. I've never even heard of such a weirdly anal issue, but after installing a new one, everything is a-okay.
So, I've never really had issues with power supplies, and generally have always troubleshot (troubleshooted? trouble...shot?) my own issues with no real, well, issues. Until now.
The other day, I got a helluva deal on a 6800 XT on Facebook Marketplace, the guy had the printout with the receipt, it's still under warranty for two years, whole shebang. So I upgraded from a Vega 56 to it. And there were zero issues. Admittedly, my power supply is only 650 watts, so I thought I might be missing some wiggle room there, and was prepared to need to upgrade. But the other night, it was fine. I stress-tested with a nearly-maxed-out 100+ FPS Cyberpunk 2077 and had zero issues, and followed that up with moderate use 144 FPS board games and things for the next few hours with a friend.
And sometime after I went to bed (I left the computer on because I'm a bad man who doesn't take care of his things or some other vaguely acceptable excuse), Windows Update occurred. Again. It's been raising hell on me in the middle of the night any time I leave my computer on, but whatever. So in the morning (this was Sunday), I saw it wasn't working right, and just kind of... shitting itself. Had trouble getting out of BIOS, all this other stuff. Eventually, I realized it was ignoring my SSD, and after unplugging everything else and forcing it to boot from my SSD with the Windows 10 install on it, it said the install was borked and asked me to do recovery steps. None of them really worked. So at this point, I was assuming that I might have hit something with the SSD and damaged the SATA controller when moving the power for the GPU or something.
So today, I got a new NVMe drive, booted from a 16gb flash drive, installed Windows 11 on it, and everything was fine. I was able to create a functional Windows 11 install, and it was fine. Until I got to the login screen. As soon as the screen asking for my PIN (on a complete, 100% valid Windows install) would load, that first frame, it would shut down hard. No BSOD, nothing. Just immediate shutdown. So I thought, "well, this seems like an issue for the POWER SUPPLY!" and removed the GPU, plugging my main monitor directly into the motherboard. Now, it was shutting down and power cycling before it even hit the BIOS, which is... weird as hell? So I thought "well, it gets further when a video card is in, let's put ye olden Vega 56 in and see how far that gets me!" and... it just works. I'm typing this from my fresh Windows 11 install with zero perceivable issues.
So my question is: How is it that my computer was perfectly fine on Saturday night with my new video card under 100% load, but by the next day would decide seemingly at random based on some sort of schrodinger's cat theory when it would shut down and when it wouldn't.
So, in summation, the four inconsistent scenarios, in tl;dr form:
- New RX 6800 XT is installed, computer runs fine at 100% load while stress testing and then for hours afterward
- 6800 XT installed, Windows won't boot and the power supply seemingly gives up
- No video card installed, the computer starts power cycling before even reaching the BIOS
- My old Vega 56 installed, everything is perfectly fine
So, obviously there's something weird going on with my power supply, but if someone can set my sights on exactly why all of this has happened, and what the proper solution to make sure it doesn't again, or just... I don't know, typing this all out has made the last day and a half of my life feel much more worth it.
And as an aside, my theory for why Windows was broken and I assumed my SSD was dying is as such: When it did the Windows Update and started trying to install it was the first time it powered down with no warning, which just broke Windows mid-update in a bad way.
So you got a good deal, but it doesn't seem to be working. You said he had the receipt and it's under warranty. Take that opportunity and just have it serviced. Skip the headache. It'll be a steal of you get new parts.
I'm already in heavy consideration of doing this, but there's a part of me that still wants to understand why certain things kept happening, like power cycling when there was no GPU in the computer.
Warranty PC technician here. The last two years I've been working with Dells and Lenovos, but prior to that I was a components testing specialist for an e-cycler, so I'd see anything and everything electronic.
How old is your PSU? Their total wattage output will decline over time, so if it's, say, 5+ years old, you may be hitting its ceiling, even if your components don't rate the full 650W your PSU was originally rated for.
It's just barely possible that you had either a faulty SSD or the voltage regulation on your mobo is fried, but I'd be surprised. The timing is suspicious for an SSD failure, and I don't see VR failures, like, at all. I really only mention the SSD because you say you did a new OS install and it's working now, but you don't say which drive you installed it on. This sounds a lot like either a motherboard or PSU problem, and the PSU is both cheaper and easier to test than the mobo.
Hypothetically, you could undervolt your GPU using Afterburner or some such and see if it's more stable that way, but really your best test would be to find a known-good higher-wattage PSU and see if the system crashes running with that. It's not a cheap, but it's very difficult to nail down faults like this without equipment, and a new PSU is about as cheap a testing device as you're really going to find.
Got it in late 2020, and it's a great EVGA that's served me well.
The new NVMe I bought at a place that doesn't take returns, of course. That said - power issues aside - the regular SSD was able to boot into a recovered two-day-old pre-update form with no problem (until the power died, of course).
I said it in another comment, but I'm same-daying a very reputable power supply from Amazon that I'll test it with, and I'll either return it if there's still issues, or have everything fixed and ready to go.
You need more wattage. I dealt with this ~10 years ago, hard shutdowns after installing a new, beefier GPU. Eventually, I figured out from a power supply calculator like this that my existing power supply wasn't enough, so I got a 1000w one. I used the old one for when I built my Mom a desktop a bit later, and both still work great.
He probably doesn't. People over provision their wattage by a country mile because most calculators overestimate by far too much.
This one is a good conservative calculator:
https://outervision.com/power-supply-calculator
My guess is either a faulty GPU, or if it is the PSU, there's more likely to be something wrong with it rather than the total output being too low. Considering he ran the card on (probably) full powerdraw with Cyberpunk, the PSU looks like it can handle the spike. If anything it blew during the test due to being used and old, my guess it was added to the PC right around when that Vega was put in there.
This is 100% what I'm assuming (as alluded to in the title), but the part of me that needs an explanation wants to know why the setup was fine the first night at 100% power draw.
Complete guess:
I wonder if the PSU was damaged during the stress test, but kept functioning while everything was still warm. Then once parts began to cool overnight, it failed more permanently (e.g. some bits shrinking and separating/cracking).
This seems like the best idea thus far. That said, it's holding up like a champ with the old card again. Then again, I'm not stress-testing it and am over-nighting an 850 watt power supply from Amazon to test tomorrow with. If the new GPU works with it, we have our answer: The last hurrah of my last EVGA component.
I’ll add-Psu may have been marginal, such that stress test put it over the edge.
I agree with the other commenter-swap psu and see if problem persists.
Aside-the most frustrating condition in all of tech support for me is truly coincidental failure, and I’ve had a couple. More often though, a change results in the revelation of marginal components.
That, I don't have much insight on, but if I remember right, I think it took a bit before my machine started cutting off like yours.
God, if that's what it is, that's absolutely wild. I literally stress-tested for an entire night just to make sure it'd be okay moving forward, and it was happy for that long.
I'm gonna be so upset if I did everything right, only to end up doing it wrong anyway.
There was that windows update though, so technically not the same platform. Could be worth seeing what changed.
Also, once you did the 100% load testing, did you reboot during that phase?
Could be that you just had a perfect moment in time on the PC
Given that nobody else seems to have any other ideas, I'm assuming this is what happened. I didn't reboot it that night after installing the 6800 XT, though.