PC freezes when booting, but only if it's been off.
August 27, 2018 1:33 AM   Subscribe

We have a desktop PC that's behaving weirdly on cold boot, and cold boot only.

It will completely freeze up (no response to anything, mouse doesn't move cursor, cannot even toggle numlock, nothing to do but hard reset or power cycle) or spontaneously restart after a couple of minutes (around when Windows 7 has mostly finished loading up) the first few times it boots after having been off. Three to five boots in, it 'takes' and does not lock or restart, and then works completely reliably fine for however long it's on - hours, even days. Doing a soft reset (shutdown-restart) works just fine and does not induce the weird behaviour.

The computer is effectively a self-build (put together by a local computer shop) and is only a few years old. It's running an AMD six-core (well, with the three FPUs - an FX6300?), an ASUS GTX750, an Antec VP700P 700W PSU, ASRock 970 Pro3 motherboard, and a single DVDRW and hard disk (oldschool, not solid state) and a single 8GB Kingston Hyper Fury DIMM.

The computer is completely responsive and acts utterly normally right up until it freezes or restarts itself. This is well after the Windows 7 desktop has displayed, but before all the startup cruft (onedrive, antivirus, etc.) has finished settling down. I can't find anything on what could be causing this. I suspect it's the PSU, or maybe faulty caps on the motherboard? Any ideas for how to go about diagnosing this?
posted by Dysk to Technology (12 answers total)
I assume you'd mention, but just in case, you haven't over or under clocked it have you? I've had weird behaviour like that with under-clocking particularly.
posted by deadwax at 2:38 AM on August 27, 2018

Doing a soft reset (shutdown-restart) works just fine and does not induce the weird behaviour.

So it could be literally temperature related; you should try physically heating it up, by placing it on top of a warm appliance or in a hot attic or something, before cold booting it. If being physically warm prevents the issue, I would try going through the whole thing and re-seating connectors just in case they've loosened themselves via thermal expansion and contract just enough when cold to cause a faulty signal somewhere.

As you may know in Performance Monitor/Resource Monitor you can find temperature readouts on some systems if you fish around a bit, though the value is often unhelpfully measured in Kelvin.

(Conversely, I've also seen the initial behavior you describe occur due to overheating, thanks to a faulty fan, but that would seem inconsistent with your case because the problem stops happening. In my case, the freeze-up happened sooner and sooner as the system physically warmed up and part of the way I diagnosed it was to position it with its vents directly in front of an air conditioner—noticing that it survived longer when I did so, and for example processor-intensive activities like watching video killed it sooner because they generate heat.)
posted by XMLicious at 4:20 AM on August 27, 2018

Definitely not temperature related, I've been through that and there is no correlation at all that I can see - it's warm by the third boot (that's ten minutes at least) but still persists. It's been going on for a few weeks now, steadily getting worse. This morning it took eight or nine boots to get it to take, including two actual blue screens (which is new!) saying something about page file writing to non paged area, and attempting to write to read only memory. Maybe the RAM is bad? But then why would it be fine - for days at a time - once past the initial hurdle?
posted by Dysk at 4:42 AM on August 27, 2018

(And no, everything is running exactly to spec, both in terms of clock speed/RAM timings and voltages)
posted by Dysk at 4:43 AM on August 27, 2018

I agree that the power supply is a strong candidate. If you can borrow and swap in a known good one of sufficient capacity, that would help diagnose.
posted by exogenous at 4:49 AM on August 27, 2018 [3 favorites]

Yeah, that would've been my first step if I had access to one. I really don't have the money to speculatively throw at one though, and that's the only way I can think to get my hands on one.

I suspect I'll end up having to start setting money aside for it, given that I can't think of anything else to try.
posted by Dysk at 5:54 AM on August 27, 2018

You can get a sufficient power supply for any PC without a particularly power hungry video card for $30-$40 on Amazon. Yes, a few bucks more might get you 80+ silver or 80+ gold, but in the real world the difference in efficiency is tiny unless you're using most of its capacity. In fact, some 80+ gold units will use more power at 10-20% load than an 80+ bronze of the same wattage rating.

I say this because I had a similar problem recently where my desktop would work perfectly if I could get it to boot, but making it get even as far as initializing the onboard video was a fraught process at best. There were also some unrelated minor annoyances that made me question whether it was a dying motherboard or the power supply, so I dithered for the longest time about it.

Eventually I decided ordering a power supply from Amazon was the best option since I could always return it if it didn't actually fix the problem. Happily, it turned out that the PSU was indeed going bad, so no return was necessary. The reduction in stress was very much worth forgoing a couple of lunches and and Uber trip to make up for the $35 the replacement cost.
posted by wierdo at 6:45 AM on August 27, 2018

There's nearly 300W of power draw just between the CPU and GPU, so I'm looking at 400-500W PSUs, and they look like they start at about forty-fifty quid (plus shipping) on amazon for anything I'd trust (I don't generally trust things with several reviews claiming failures that took other components with it). No idea what the whole 80+ or precious metals stuff means. Last I knew my shit on computers was when I built a socket 754 Athlon64 well over a decade ago, and I'm pretty sure I don't remember those terms being thrown about back then.

Reseating the RAM and power connectors did nothing. None of the caps on the motherboard are visibly bulging or otherwise visually suspicious.
posted by Dysk at 7:25 AM on August 27, 2018

The 80+ thing is just a power-efficiency certification. Not strictly necessary, but generally PSUs with the certification are better-made. Enthusiast-grade units are typically 80+ Bronze or better.
posted by neckro23 at 10:05 AM on August 27, 2018 [1 favorite]

Run a SMART test on the hard drive.
My PC was doing a similar won't-start-from-cold trick, running fine after getting a few dud boots. Its problem was a short in the hard disk circuit board.
posted by anadem at 11:26 AM on August 27, 2018 [1 favorite]

SeaTools gives the drive a pass on every scan it can do (including SMART).
posted by Dysk at 6:29 PM on August 27, 2018

Pretty sure it's not the hard drive - had a total freeze in the BIOS utility earlier. Definitely setting money aside for a new PSU, as that's looking increasingly likely to my mind.
posted by Dysk at 2:19 AM on August 29, 2018

« Older Back me up buttercup!   |   Confusion Concerning Radical 96 – a Hanja Question Newer »
This thread is closed to new comments.