Why would my computer hang and reboot in Windows but not Linux?
January 5, 2016 4:04 PM Subscribe
About a year ago I made the mistake of building a computer myself. One motherboard and 12 months later, I have a crash that happens consistently in Windows but never in Linux.
The PC in question is pretty simple—an AMD CPU with integrated graphics (A10-6790k), 8 GB RAM, 1 TB hard drive, very low CPU temperatures. But ever since I built it, with short and unpredictable respites, it's hung and rebooted whenever I try to run Windows on it.
I started with a retail copy of Windows 8, which I reinstalled twice; since then I've updated it to Windows 10. I'll be playing a game, or else my wife will be watching a video—typically it happens when the video or game is full screen, but not always—and then the system hangs, the sound repeats like a broken record, and the computer reboots to the BIOS. Sometimes it takes a few hours to happen, sometimes it takes a couple of minutes.
It sounded like heat to me—so much so that I bought a big aftermarket CPU fan and installed it—but that didn't do it. (The "thermal margin" on AMD's Overdrive software is always very high, and it's gotten higher since I installed the new fan, so I don't think it's a heat problem.) Then I thought it might be the motherboard shorting out, so I reinstalled it in the case and then just installed a new motherboard entirely, in case it was bad. No change. I checked the power supply with one of those Rosewill testers, and everything was stable. I checked the RAM with Memtest—no issues there. (One piece of additional heat information: two temperature readings in the Overdrive software are always impossibly high, but also grayed out—something like "-299 degrees." The number doesn't move. I assume that means my motherboard doesn't have those sensors, but I've misplaced the screenshot I took of the window.)
All that's left to replace is the RAM and the CPU. I'd keep playing hardware roulette, but I've noticed that whenever we boot the computer into Linux—first from a live USB drive several months ago, and now on a hard drive partition—I can't crash the computer. It's been playing one of those 10-hour YouTube videos in full-screen all day without issue.
I'd just keep it in Linux, but we bought the PC in part to play Windows-only games. Plus I'm curious: What could be causing this? Are the AMD drivers in Linux sufficiently different that it could be a driver problem, despite the drivers being updated pretty consistently across the life of the device (and this problem)? Is it still a hardware issue, and I'm not running into it because Linux uses fewer resources? Thanks for any light you can shed.
The PC in question is pretty simple—an AMD CPU with integrated graphics (A10-6790k), 8 GB RAM, 1 TB hard drive, very low CPU temperatures. But ever since I built it, with short and unpredictable respites, it's hung and rebooted whenever I try to run Windows on it.
I started with a retail copy of Windows 8, which I reinstalled twice; since then I've updated it to Windows 10. I'll be playing a game, or else my wife will be watching a video—typically it happens when the video or game is full screen, but not always—and then the system hangs, the sound repeats like a broken record, and the computer reboots to the BIOS. Sometimes it takes a few hours to happen, sometimes it takes a couple of minutes.
It sounded like heat to me—so much so that I bought a big aftermarket CPU fan and installed it—but that didn't do it. (The "thermal margin" on AMD's Overdrive software is always very high, and it's gotten higher since I installed the new fan, so I don't think it's a heat problem.) Then I thought it might be the motherboard shorting out, so I reinstalled it in the case and then just installed a new motherboard entirely, in case it was bad. No change. I checked the power supply with one of those Rosewill testers, and everything was stable. I checked the RAM with Memtest—no issues there. (One piece of additional heat information: two temperature readings in the Overdrive software are always impossibly high, but also grayed out—something like "-299 degrees." The number doesn't move. I assume that means my motherboard doesn't have those sensors, but I've misplaced the screenshot I took of the window.)
All that's left to replace is the RAM and the CPU. I'd keep playing hardware roulette, but I've noticed that whenever we boot the computer into Linux—first from a live USB drive several months ago, and now on a hard drive partition—I can't crash the computer. It's been playing one of those 10-hour YouTube videos in full-screen all day without issue.
I'd just keep it in Linux, but we bought the PC in part to play Windows-only games. Plus I'm curious: What could be causing this? Are the AMD drivers in Linux sufficiently different that it could be a driver problem, despite the drivers being updated pretty consistently across the life of the device (and this problem)? Is it still a hardware issue, and I'm not running into it because Linux uses fewer resources? Thanks for any light you can shed.
Just a thought - have you updated to the latest BIOS?
posted by transitional procedures at 4:27 PM on January 5, 2016
posted by transitional procedures at 4:27 PM on January 5, 2016
It's presumably a driver issue if it only happens on one OS and not another. Sadly this is pretty common.
posted by deathpanels at 4:27 PM on January 5, 2016 [2 favorites]
posted by deathpanels at 4:27 PM on January 5, 2016 [2 favorites]
Another thought - are you using stock Windows 8/10 drivers from Microsoft, or are you using drivers from the board vendor's site?
posted by transitional procedures at 4:31 PM on January 5, 2016
posted by transitional procedures at 4:31 PM on January 5, 2016
Have you checked the Windows Event Viewer for things corresponding to the crashes? (Assuming that's still around in Windows 10, but the logs must be accessible somewhere...)
Could it be a virus or some other malware that's trying to use one of those exploits that involves stochastically attempting to cause a specific glitch, a buffer overrun into the correct memory region or something like that? Maybe run through deezil's guide under Windows?
posted by XMLicious at 4:47 PM on January 5, 2016
Could it be a virus or some other malware that's trying to use one of those exploits that involves stochastically attempting to cause a specific glitch, a buffer overrun into the correct memory region or something like that? Maybe run through deezil's guide under Windows?
posted by XMLicious at 4:47 PM on January 5, 2016
"the sound repeats like a broken record" - which sound, the sound from the game or beeps?
If it's beeps you should be able to look up the code - usually beeps means memory, but depending on the number (beep - pause (1), beep - beep - pause (2), etc) it could be something different.
If it's the sound of the game repeating, that is pretty strange, and would lead me to think it's a driver issue - take the advice above and make sure you are using the drivers supplied by your motherboard manufacturer...
posted by NoDef at 5:00 PM on January 5, 2016
If it's beeps you should be able to look up the code - usually beeps means memory, but depending on the number (beep - pause (1), beep - beep - pause (2), etc) it could be something different.
If it's the sound of the game repeating, that is pretty strange, and would lead me to think it's a driver issue - take the advice above and make sure you are using the drivers supplied by your motherboard manufacturer...
posted by NoDef at 5:00 PM on January 5, 2016
AMD makes notoriously shitty video drivers. Before playing any more hardware roulette, I'd be switching to AMD video driver version roulette.
posted by flabdablet at 12:29 AM on January 6, 2016
posted by flabdablet at 12:29 AM on January 6, 2016
Couple more things occur to me, since this is a self-built box that you describe as a mistake.
The broken-record thing is indicative of a hard CPU lockup (the sound card hardware keeps on feeding the current content of some buffer or other through the DAC via DMA, but the CPU stops responding to interrupts so it never alters the content of the buffer).
When you installed the CPU heatsink, did you supply your own thermal grease? If you used one of the high-performance silver-based ones, and you were heavy-handed with it, it's possible that a blob of somewhat conductive thermal grease has ended up where it shouldn't be and is now interfering with the contact between a CPU pin and its socket contact, or perhaps even occasionally shorting a couple of pins.
If you have multiple RAM sticks, try running the machine with them installed just one at a time. I've seen bad RAM lock up a CPU as you describe, even after passing many hours of Memtest86+.
It might also be worth installing BlueScreenView to see if Windows has been recording crash dumps. If it has, you might get a clue from the list of loaded drivers that BSV shows you about which driver, if any, is the culprit.
You have all my sympathy. Intermittent faults are shits of things to diagnose, and given the complexity of modern computing machinery, just switching bits out until the fault refuses to manifest is still the least-slow way to track them down.
posted by flabdablet at 3:55 AM on January 6, 2016 [2 favorites]
The broken-record thing is indicative of a hard CPU lockup (the sound card hardware keeps on feeding the current content of some buffer or other through the DAC via DMA, but the CPU stops responding to interrupts so it never alters the content of the buffer).
When you installed the CPU heatsink, did you supply your own thermal grease? If you used one of the high-performance silver-based ones, and you were heavy-handed with it, it's possible that a blob of somewhat conductive thermal grease has ended up where it shouldn't be and is now interfering with the contact between a CPU pin and its socket contact, or perhaps even occasionally shorting a couple of pins.
If you have multiple RAM sticks, try running the machine with them installed just one at a time. I've seen bad RAM lock up a CPU as you describe, even after passing many hours of Memtest86+.
It might also be worth installing BlueScreenView to see if Windows has been recording crash dumps. If it has, you might get a clue from the list of loaded drivers that BSV shows you about which driver, if any, is the culprit.
You have all my sympathy. Intermittent faults are shits of things to diagnose, and given the complexity of modern computing machinery, just switching bits out until the fault refuses to manifest is still the least-slow way to track them down.
posted by flabdablet at 3:55 AM on January 6, 2016 [2 favorites]
WhoCrashed might be useful here. It analyses crash logs for you, and returns the likely culprit/s.
Nthing "sounds like a bad video driver."
posted by quinndexter at 7:12 PM on January 6, 2016
Nthing "sounds like a bad video driver."
posted by quinndexter at 7:12 PM on January 6, 2016
I put together a computer a few years ago and had issues that sound so similar to yours.
I would crash in certain situations quite predictably, although working just fine in MOST situations.
I went so far as to order another motherboard & cpu, on the theory the that first was defective or something. The second one worked EXACTLY like the first.
The problem turned out to be the power supply--more precisely, the power supply was TOO SMALL. What that meant is that the system worked just fine and dandy almost all the time.
Then when load got too high (most generally when the video card was under load, playing a video or a game) it would just mysteriously zonk out with these strange blue screens of death and the like.
Low voltage is a terrible thing to a computer . . .
I was particularly mad in this case, because I had purchased a pre-fabricated "set" with selected motherboard, case, power supply, video card etc etc etc put together by a respected online vendor. So I naturally expected things would be thought out and meet minimum requirements.
At any rate, it might be worthwhile replacing the power supply with a higher wattage model. That fixed my particularly problem permanently. It is pretty easy and inexpensive.
I'm not sure how to test for this exactly or what particular symptoms gave it away (aside from random crashes that happened when I replaced the other main culprit I suspected), but you could try plugging your system in this online power supply calculator and see if your current power supply is amply large or, perhaps, just barely big enough.
posted by flug at 7:31 AM on January 7, 2016 [2 favorites]
I would crash in certain situations quite predictably, although working just fine in MOST situations.
I went so far as to order another motherboard & cpu, on the theory the that first was defective or something. The second one worked EXACTLY like the first.
The problem turned out to be the power supply--more precisely, the power supply was TOO SMALL. What that meant is that the system worked just fine and dandy almost all the time.
Then when load got too high (most generally when the video card was under load, playing a video or a game) it would just mysteriously zonk out with these strange blue screens of death and the like.
Low voltage is a terrible thing to a computer . . .
I was particularly mad in this case, because I had purchased a pre-fabricated "set" with selected motherboard, case, power supply, video card etc etc etc put together by a respected online vendor. So I naturally expected things would be thought out and meet minimum requirements.
At any rate, it might be worthwhile replacing the power supply with a higher wattage model. That fixed my particularly problem permanently. It is pretty easy and inexpensive.
I'm not sure how to test for this exactly or what particular symptoms gave it away (aside from random crashes that happened when I replaced the other main culprit I suspected), but you could try plugging your system in this online power supply calculator and see if your current power supply is amply large or, perhaps, just barely big enough.
posted by flug at 7:31 AM on January 7, 2016 [2 favorites]
That's a good point, especially if you cheaped out on the case and got one of the no-name ones that comes with a power supply built in. Many of those are good for lighting up power supply testers and not much else.
posted by flabdablet at 9:02 AM on January 7, 2016
posted by flabdablet at 9:02 AM on January 7, 2016
This thread is closed to new comments.
posted by AugustWest at 4:21 PM on January 5, 2016