I recently installed Ubuntu 13.10 64 bit (dual boot with Win 7) on my desktop PC, but every once in a while the screen turns off suddenly, the fans slow down, and the system appears to crash. I can't see what actually happens, because the screen turns off (no input), but the sound crashes too. All I can do is to press the reset button on the computer itself, after which it boots up without any issue or mention of a problem.

At first this happened randomly and unexpectedly simply from normal use, just by opening Firefox or moving the cursor around the desktop right after a clean install. Now it seems to happen only when I do something CPU intensive, like watching videos or importing my music library. I'm not sure what changed, because all I did was turn it off for a day and then come back.

I've since cleaned the inside of the computer for dust, and all fans appear to be spinning.

The system specs are:

AMD Phenon II X4 965 3.4 GHz

ATI Radeon HD 4870, and Ubuntu says the driver is Gallium 0.4 on AMD RV770

  • Is the system overclocked or using reduced cooling? If not such instabilities are rare. Run a memory test (available from the Ubuntu bootloader): memory can go bad over time. Nov 8 '13 at 21:48
  • @Gilles Not that I know of. The memory test found nothing wrong either
    – user14991
    Nov 9 '13 at 9:19

I posted something very similar on Ask Ubuntu a long while ago, when I had a laptop that I just purchased produced similar symptoms to yours, from overheating¹.

When the symptoms occur, is the system powering down? (Is just the screen off, or does the whole system shut down?)


Of the machines I've had that were actively overheating, in my experience:

  • Windows will just shut down. No blue screen, no warning, just "oh, my machine is off." It doesn't log the reason for this, but does do a "hmm, I appear to have not been shut down properly."
  • Linux will do a graceful, but forced shutdown. The disks will unmount fine and everything, but applications will probably be forcefully closed. (Unsaved data will likely be lost.) If heat was the problem, Linux will, in my experience, log this. (This probably depends on exact setup & distro, though.)

Determining if heat is your problem

"Over heating" can be a difficult problem to isolate.

  1. Find it in the logs. Check files in /var/log, particularly, dmesg logs, kernel logs, etc. In /var/log/messages I had:

    [ 9855.198401] thinkpad_acpi: temperatures (Celsius): 85 43 35 62 27 N/A 28 N/A 35 44 60 N/A N/A N/A N/A N/A

    I later found another message that was even more explicit, along the lines of "hey, hit a temperature trip point, something is going critical". Unfortunately, I didn't record on Ask Ubuntu where I found this.

  2. Watch the temperature of the machine. You can do this in a terminal by repeatedly running sensors, which will show you something like:

    temp2:        +96.0°C  (crit = +100.0°C)

    Here, my second temperature sensor was at 96°C. (Note that the temp here is in Celsius: in this example, which was from a laptop I (briefly) owned, I'm a few degrees away from being able to boil water with this CPU. This is not a good thing.) The crit = +100.0°C bit indicates a "trip-point", a point at which the OS will get concerned about the state of the hardware. (And shut down.) I was hitting at least one of these.

Being 4°C away from a trippoint is too hot. Being at 96°C is too hot. For comparison, my current desktop runs at 66°C when all the CPUs are at 100% — as I installed the heatsink/fan myself, a professionally built system should probably run cooler. When idle, it's about 33–36°C.

Note that while 96°C is hot enough to burn you, the laptop I had was touchable, and only "very warm". By the time you got to the case, the heat had dissipated enough.

Forcing the problem to reproduce

WARNING: what follows will cause a CPU to heat up, and if your CPU isn't being properly cooled, runs the risk of damaging the CPU. Use at your own risk. (I think some modern CPUs might have internal fail-safes against heat, but I'm pretty sure you're voiding your warranty here.) You can "provoke" an overheating quite simply: launch a number of infinite loops to heat up the CPUs. python -c "while True: pass" will do this. (Ctrl+C to kill it.) (You might need to spawn multiple processes, one for each CPU. Take it gradually, and watch the heat output with sensors) (Also, I've found that Minecraft is quite good at this as well.)

In my case, I used this to attempt to replicate the symptom, and further establish that it was indeed heat causing the problem.

¹Graciously, the purchase of my laptop came with a 30-day return-it-for-any-reason. The retailer's "techs" were "confident" that the machine wasn't overheating. (I'm confident they were wrong.)

  • The ATI GPU is around 90C during normal use, but sensors doesn't list its critical level. I got it up to 121C by watching a video fullscreen. It didn't crash until about a minute after I closed the video and stopped measuring the temp, however. The CPU seems to stay well below its critical level, so it seems to be the graphics card overheating
    – user14991
    Nov 8 '13 at 20:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy