Computer just switched off at random, for the second time today. This time, it happened simultaneous with a power flux event (the air conditioner halting).
Ha ha heehhhh after several days of no problems my computer just shut off at random while I was sitting at it using it, and then less than five minutes later, did it again. Is there a way to get Windows to tell me, after a restart, why the hell it just restarted
Like I understand it might not know why it shuts down but surely if it restarts there's somewhere a record of why
I am terrified of a near future where I spend hundreds of dollars on a UPS, it takes up a bunch of space under my desk, and it turns out not to solve the problem
Every time I buy an Apple computer I eventually have to replace it even though it works and I don't want to replace it, because of planned obsolescence, and every time I buy a non-Apple computer it eventually just breaks
The scariest possible thing that could happen in the next few months is I buy a new mac because Apple forces you to buy a mac to develop for the Vision, then my Windows machine craps out totally and I'm having to use a Macintosh as my daily driver
@mcc I dunno… I’ll give Apple crap for a lot of things, but not planned obsolescence. They support phones “forever” in Android terms, and they support older desktop machines than MS now too…
@mcc@aerique I do not understand. If it works, why do you feel you need to replace it?
Wanting to run the new shiny is not the same thing as planned obsolence. Everybody abandons hw at some point. Not all computers can run Windows 11, to take one example.
@mapcar Apple's developer tooling and such deprecate old system support at a fairly breakneck pace compared to, say, Windows, where old software continues working much, much longer without heroic effort on the part of the developer. This is why I blame Apple for the problem of quickly disappearing software support.
@mcc I'm always worried that the power supply and the mobo won't talk to each other properly. Maybe that's just my ignorance. I'm happy upgrading every other part of the machine, but as soon as either mobo or power supply need replacing, I figure it's good money after bad and I replace the entire machine.
Like the ThermalTake has worked fine for just under two years, nothing obviously changed before the problems started except I swapped my SATA drive for a m.2 one on the motherboard, and that was like a week or two before, then boom problems
So maybe the ThermalTake isn't overloaded, just going bad?
@mcc Likely, "performance" as it is used here means a combination of efficiency, thermals, derating, and noise. Which one is subpar? who knows! Based on the specifics of the teardown in the review, it seems like a good choice.
@mcc I also had random halts which took me a while to debug.
Turned out that with the new passive cooling tower on my CPU and setting the fans as silent (low speed), the RAM did not get cooled well enough and overheated under load.
I changed the fan setting to normal and the computer hasn’t crashed since.
The hint was that I saw memory corruption messages with two crashes (after the crash the hardware warned and I saw that in the scrolling text on startup — under Linux).
@rotopenguin it could also be the heat from the m.2 driving some other component out of the save range. Those heat sinks I had to screw onto every m.2 indicate that they get quite hot.
@mcc AFAIK the heat sink (just a metal stripe with some soft heat connector taped underneath) should have already been on your motherboard (or the m.2 should be in a location where it can dissipate heat). @rotopenguin
@mcc I've had good experience with Corsair power supplies – I only replaced my previous one (RM650x) because it was too weak for new graphic cards; had it from july 2017 till january this year. Currently have a Seasonic Prime PX 1300.
@mcc Oh, and as for what I had connected to RM650x:
GTX1070, 1TB SATA SSD, 4x 8TB SATA disks inside a hotplug cage, ARC-1212 RAID controller; originally I had Asus Z9PA-U8 motherboard, Xeon E5-1620 CPU, 4x8GB DDR3 registered RAM and a PCIe USB3 controller; later I changed to Asus Pro WS X570-ACE motherboard, Ryzen 5900X CPU and 2x 32GB DDR4 ECC RAM.
BTW, check the warranty for your Thermaltake – at least Corsair has either 5 or 10-year warranty (not sure which), while Seasonic claims 12-year warranty.
@jernej__s The Thermaltake (it was a GF1 not a GF3) has a 10 year warranty. I will be exploring this. But if I have to mail them the PSU I like… I have to use my computer in the meantime.
The store near me had Corsairs but none with ATX3 which I believe I need for my video card.
@mcc AFAIK, 3xxx series still uses classic PCIe power connectors. Only 4xxx series uses the new one – it's much smaller with 12 pins and supports up to 600W per connector (which has gotten reputation for being melty; the classic 8-pin PCIe power connectors are rated for 150W, which is why high-end cards usually have two of them).
@jernej__s@mcc there's a lot forum thread which is like the goto for PSU reviews by model because so many brands are wildly variable from model to model.
I had the same issue and found an open box fractal PSU they rated highly on eBay for cheaps and it fixed my world.
@mcc So I’ve been told that modular power supply cables are not standardized or interchangeable. Dunno if that’s part of the confusion, but I figured was worth reminding.
@mcc I was terrified the first time I did this, but if you follow carefully the instructions, you should be fine.
Feel free to ask on here if you are unsure about something. That’s what I did and the Fedi helped me :).
@mcc the upside is that usually you can't really plug things in the wrong way around so you're relatively safe as long as you stick to the cables that came with the new PSU. There should be at least a 24-to-24-pins cable going from the 24-pinhole on the PSU to the 24-pins slot on the mobo. Optionally some mobos also accept an extra 8-pin thing. Modern PSUs label these cables CPU/PSU. The PCIe ones go to the PCIe cards if you have an externally powered one (e.g. GPU).
@mcc I think you may minimize the risk of that happening if you base things on the parts that need power, not starting at the power supply. When you run out of parts with unconnected power ports, you’re done!
Anything missed will just, not have power on startup, and not show up in Device Manager, so that should be an easy way to catch any oversights.
(Also I like to see the extra cables as ‘just in case’ replacement parts for down the line.)
24 pin ATX into motherboard. This is the top two connectors on the back of the power supply. On the motherboard it's usually on an edge far from the CPU
8 pin ATX12V into motherboard. This is usually near the CPU. Typically you'll only need one of them, high end motherboards might use two. Connectors at the top right of the PSU
That's everything for the motherboard
Then
*ATX12VHPWR this is the second connector from the right on the bottom row. Only GeForce 4000 series uses this
PCIe power. This is bottom right and top right (same on the PSU side as ATX12V). Used for graphics cards that aren't geforce 4000. Use as many connectors as you can, ie avoid using daisy chained cables if you have enough sockets to avoid it
SATA/MOLEX. Three left connectors on the bottom row. Basically anything that isn't covered above will use these connectors. Watercooling, random internal things that want power like fan controllers, etc etc etc
@directhex thank you, having this to compare against will legitimately help.
Here is a question. The cables came in two bags. One of them has all the cables except one, and one of them had this cable only. Do you have any idea why they might have done that?
@mcc that's the ATX12VHPWR cable. It's probably separate because power supplies with that connector are very new, whereas they have a warehouse full of the older connectors
@mcc@directhex from a power perspective, there shouldn't be too many. The big 24-pin and 6/8-pin CPU ones are typically all you need on the motherboard. The rest should be obvious - i.e. drives and graphics card.
I expect there will be plenty of spare cables afterwards
The PSU came with three seemingly identical cables, each of which has a single head at one end and two heads at the other end. I think these are interchangeable and I think one goes from the CPU/PCI-E slot at the top of my motherboard with one head left unused, and the other goes from a CPU/PCI-E slot to the 8pin slot by the CPU with one of the cable heads unused.
However, the three seemingly identical cables are NOT identical. See below. But nothing is labeled and there's no parts manifest
Wait, no! There is a parts manifest…on… the box?! That is the least convenient place to put documentation. I don't have room to store boxes. Also, it seems to be wrong. It claims it comes with two of the 8-pin "CPU" cables and four (?) of the PCI-E "split" 6+2 cables, but all I find in the box is three of the 8-pin CPU cables (Except one appears to have the "empty" pin??)
I feel defeated and incompetent. I can do literally anything (given enough time) with a computer once it is switched on, but I apparently am not capable to plug a cable into another cable.
Someone asked for a photo of the motherboard. Here's above the video card, the video card, below the video card.
The "wide" motherboard connector (first image right side, in shadow) is obvious. Similarly the cable for the video card (image 2) seems obvious. But the CPU connector (first image top right) is baffling me. EDIT: wait maybe I figured it out.
@mcc You can sometimes tell which cable is for the CPU power by checking if any of the 8pin cables split at the end. They are usually 2 4 pin cables stuck together. Then from there, you just have to orient it so that the clip latches onto the motherboard connector
Also I distinctly remember (and wrote down) I unplugged something in the bottom left (image 3 or possibly "CHA_FAN2" in image 1) and I assumed it would be obvious what but now I have literally no idea what it was. I don't think this one could possibly be essential (??) because it's not on the basic plugs lists but I am super angry at myself for not taking clearer notes. Did I hallucinate this?!
I've been stuck on this step for an hour. Just basically a disaster here
@mcc It would be a chassis (case) fan -- if all your fans are plugged into the board, you're probably fine (it's not unusual for there to be more fan headers on a motherboard than fans in a build).
@mcc The three superficially identical cables have shaped plastic parts so that each only fits where it's supposed to fit. If you can't figure it out by looking at the shapes, just try them in all combinations and orientations until you have them all plugged in.
The CHA_FAN2 is for a chassis fan. Small three-pin thingy that connects with one of the fans. If you're wondering how to plug a 3-pin connector on a 4-pin header, look at the clip on the side.
Computer might be fixed now but the only way to know for sure is to just use it for a indeterminate number of hours and after the hours-long ordeal replacing the PSU all I want to do is lay down
@mcc FYI, even though every manufacturer uses the same #%@@ connector on the PSU side, none of them agree on the pinout. Don't just toss the extra cables into a bin. Put them in labelled baggies first. (I didn't blow up a motherboard, but it was hours of frustrated debugging)
@mcc It's chassis fan connector so you might have a cable dangling off a fan somewhere - or sticking out of some cable management contraption. It's a very specific connector type newegg.com/p/N82E16812162026
It's not essential but not having the fan operational will rise temperatures.
@jplebreton It turns out that one of my 4x2 connectors is in fact the 3x2+1x2 connector labeled on the box, and they just glued the 3x2 to the 1x2, and didn't update the box . I guess there's a little bit of plastic indicating this. Maybe you can break them apart if you try.
A parts manifest on the box is fine, quickly lets you say if you need to buy extra stuff too, but it should be replicated on an info card or manual inside. and, well, accuracy would be nice.
This sounds like someone fucked up at the factory.
@mcc can you share a picture of the mobo? The 20+4 should be quite obvious in terms of visibilty. The CPU 4+4 depends on the motherboard. The others are for external cards.
@mcc since the box listings and the box contents don't seem to match here, I'm thinking maybe you got a dud or something. This doesn't smell right, imo - especially when factoring in that missing pin on one of them. What a mess.
@mcc you probably figured this out already, but CPU and PCI-E cables are keyed differently; you physically can't plug them into the wrong place.
As long as all the power inputs on the motherboard and your graphics cards have a lead plugged into them, you're good to go.
(They're just GND+12V to deal with the power consumption of a GPU, which cannot be sourced through the tiny PCI-E pins from the motherboard. Same for CPU aux connector: extra power over what an old ATX connector supplies)
@mcc That's because TT is a brand name manufacturer, and they have the budget to label the cables and/or write proper documentation.
This is usually considered expendable and of little value in computer part reviews but becomes essential when you do not install the part at least monthly, and it's not completely trivial.
Modular power supplies fall outside of the category of parts with trivial installation for sure.
@mcc I am completely ignorant of this stuff, all the hardware I do is usb C powered these days. But it looks like those cables aren’t the same, the pattern of square vs arched ‘pins’ is different. If that makes any sense. So I don’t think they all -fit- in the same places.
@mcc no idea why one is different, but it could be a quirk of that model as the PSU side connectors aren't standardized. Looking at your pictures, I suspect they packed the wrong mix of cables.
@brendan no, the gigabyte was the one that was unlabeled. The thermaltake was the one that failed after 23 months (not clear yet if due to overdraw or bad unit)
@mcc Modular power supplies are like this - there's no standard for how the cables should work exactly (pinout on the PSU end can vary from mfg to mfg!) so it tends to be a bit of a zoo.
There should be two "classes" of 8-pin "splittable" connector: one set of 2 cables that breaks into 2x4 pins (this is the one for the CPU/PCI-E) and another set of 4 cables that breaks into a 6 pin and a 2 pin "head" (which are for GPUs). All of them should have a solid 8-pin side.
@mcc There's I think three different cable types with 8 pin on one end and SOMETHING on the other - it's sort of hard to tell without looking at both ends of the cable.
That said, there should be 6+2 cables in that box. I don't know what they might look like on the PSU side, though, or exactly how they plug into the GPU (the story varies by manufacturer). Let me see if the manual says/
@mcc Is that picture all of the cables with 8-pin connectors? It might help to figure out if you're missing a cable to lay out all of them.
My suspicion is that the missing pin is on the PSU side - it's sometimes possible to elide a pin from some cables in a sort of "optional" manner, which could explain why there's only one of them.
@mcc The opposing end of the cables that go into those won't all be identical; they should either be CPU (4+4) or GPU (6+2) connectors that "oppose" that side. It doesn't matter beyond that what you plug in, though.
@mcc modular power supplies piss me off so much. It’s always impossible to tell which combination of plugs and cables are supposed to be used for anything and even if they do give a parts manifest it’s always like. really deceptive.
@mcc If there's a chance it's overloaded, go to pc part picker, plug in all your parts to a parts list, and see what it says for power draw. I take that number, round up to the next 50W and add 200 to give some room for small upgrades and transient loads.
I wouldn't go further than this unless you know for sure what your actual load will be(this gives peak load which is almost never reached) and you want to dial in maximum efficiency
@mcc I don’t know anything, but the last time I bought a power supply it was from Corsair and the actual unit came in a velvet bag(?). Like it was jewelry or a fancy adult toy.
@mcc I really like the super flower leadex III. They charge gold and test platinum. I can’t find the site that does the reviews but will link when I do.
@mcc that model I’m not sure. But super flower makes the guts for a lot of other quality PSU and you can get a good price for a very high quality if you aren’t paying brand markup
@mcc every time I see this thread I get disappointed in myself that I don’t remember enough of this to add something helpful… but most of how I used to know as much as I did was constantly having to replace parts and ending up with several computers worth of everything laying around as a result, and if you want to spend your time writing software I can’t recommend taking on that hobby. I’ve mostly switched to laptops I can’t do that with, if I needed to build I might resign myself to building 2
@mcc that is basically what I’ve been assuming. That’s what prompted my previous enquiry about whether it freezes under Linux. The graphic driver would hang at very unpredictable times and the logs didn’t reveal anything particularly insightful. After searching for many days, someone in some bug thread said that it might be a power issue. I didn’t think that was a possibility but I thought I’d eliminate it as a cause so, I swapped the power supply. Never happened again.
@mcc Gigabyte has had some high profile issues with exploding power supplies. That should be fixed but it's shaken my trust in them a bit.
The Corsair RM and RMX series are very good if you can order one. The RMX is slightly higher end but they're both good. I've got an RM850 in my system keeping a loaded 5950x system going.
Thermaltake is decent, though, if it's under warranty I'd try that.
@mcc I haven’t been keeping up on all the replies- has anyone suggested PSU yet? If it’s not thermal protection from overheating, what you’re describing sounds like the PSU being overtaxed to my ears. 🤔
@beporter I have a "ThermalTake Premium Edition 750W ToughPower GF1 Gold Modular Power Supply". Do you know how I would go about monitoring its power load, logging its power load, or determining whether it has been overloading? The only downloadables on ThermalTake's website seems to be… a program for controlling the LEDs inside, a thing I cannot imagine ever doing on purpose, ever. Maybe it also has logging.
@mausmalone In July of 2021 I replaced my video card with a fancy ass, high load 4070ti. It was too much for my psu so I got a new psu at the same time. My current psu is a ThermalTake Premium Edition 750W ToughPower GF1 Gold Modular. It's under two years old so it dying would be a surprise. I did add a new m.2 HD recently, but the problems didn't start until 1-2 weeks later.
Do you know of any way to get the ToughPower to report its load, or log its load, or report if it has been maxing out?
@mcc looks like a BSOD. the references to power are just the Kernel-Power event source, which tracks when a machine boots after a non-clean shutdown. look just before to see if there's an error event from volmgr that says "Dump file creation due to error during dump creation" - if so you had a BSOD but the filesystem wasn't in a state where the crashdump could be stored.
@mcc if you open up the Start menu and start typing "environment variables" it'll come up with an option called "Edit the system environment variables" (this isn't what you're going to do, it's just the way I remember how to get to the menu you need).
it'll take you to a System Properties window. click Start-up and Recovery Settings. make sure you've got "Automatic memory dump" enabled, with overwrite ticked. Should look like this:
@mcc if you don't see the volmgr errors in Event Viewer, take a look to see if you've got a C:\Windows\System32\MEMORY.DMP file
it might be hidden. in Explorer click View -> Options, then in View tab enable "Show hidden files" and uncheck "Hide protected operating system files". that'll make sure you can see it ok.
@gsuberland This all seems more than a little unlikely. BSODs causing instant restart rather than a BSOD, in some cases causing the computer to turn off rather than reboot, and no signs in the event log?
@gsuberland So I have been filtering for events 41,1074,6006,6005,6008, as that document suggests, in the System log. If I scroll down to the time of the last reboot event and look around it, I see nothing about a crash of any type, although also worryingly I find the events are not in chronological date order (they jump back and forth 10 seconds or so). Is there a way to add crashes to my filter?
@mcc that's not a great list of event IDs. should add 161 to that list at least, but there are lots more.
generally better not to filter on ID/source, but instead just scroll to 6/16/2023 10:44:52 and then look back in time.
if the timestamps aren't in order, you might've caught one of the column headers and made it sort. could also be due to time correction events (e.g. NTP sync) or the RTC on your motherboard being off.
@mcc if you have a MEMORY.DMP file, compress it with a password (ideally 7z, but zip is fine), upload it somewhere, and send me a DM with the link and password. (the reason not to post the dumps publicly is that it contains kernel and userspace memory that might include sensitive info)
I can load the crashdump into windbg and try to figure out where and why the crash occurred.
@mcc try disabling "Automatically restart". if it is a BSOD, it'll at least stay visible instead of instantly rebooting after the dump attempt fails.
if the crash is occurring in a storage driver, it'll be impossible to write to the event log or memory dump, so that step of the BSOD will fail and it'll just insta-reboot. (sometimes this does cause a power off instead of a reboot on some machines. it's pretty random.)
@gsuberland I do have a MEMORY.DMP file in C:\Windows, dating from 11:05 PM last night.
I had a sudden reboot at about Midnight:30 AM last night, and another reboot around an hour ago (10:45 AM). I don't… remember a reboot at 11:05, and I wasn't in the room at the time. Maybe something happened and I didn't notice. I do find this in the log from around that time:
@mcc cool. if you wanna upload it somewhere and DM me a link, I can throw it in windbg and analyse it. might point you in the right direction at least.
@mcc bugchecks (BSODs) don't always appear in the event log. it depends exactly where the crash occurs.
the automatic memory dump picks the relevant type of dump to perform depending on heuristics, so if it's doing a small memory dump it takes a fraction of a second to complete and will just look like an instant reboot.
if you untick "Automatically restart" on the system failure bit, it won't insta-reboot.
@mcc I have a very similar problem! It seemed to happen more often in sleep mode, but eventually took the form of not being able to start the computer at all without manually jumping the power supply with a paper clip.
My leading theories are: faulty power supply or (more likely) a slowly building charge on the circuit the computer is plugged into (since I live in a house with old electric)
Just happened again. Third time today. Second time in an hour? No other electrical devices seemed to have any problems. I guess I should officially be worried. :(
Add comment