Friday, October 22. 2010
overheatd - is your CPU too hot?
Trackbacks
Trackback specific URI for this entry
No Trackbacks
Comments
Display comments as
(Linear | Threaded)
Hm, braucht man das wirklich? Gefühlt macht mein Ubuntu das eh schon. Oder vielleicht ist das hardwaremässig realisiert bei mir, k.a. ich merk es nur da dran, dass wenn ich länger 3d-Sachen mach, es anfängt zu ruckeln, weil der Prozessor runter taktet. Ich müsste wohl mal die Lüftung entstauben, dass setzt sich da bei mir immer fest und dann kriegt der Prozessor nicht genügend Luft.
Mate, try linux-phc. My core2 love it, in load, before I got without any problems 93'C, after undervolting cpu by linux-phc I am getting MAX 73'C. With this undervolt I workalways with performance governor. (2x 2.2GHz, core2 t7500).
From what I understand, the Nehalem processors actually do this automatically. Installing the Indigo Xtreme "thermal interface" involves running the thing as hard as you can with all the fans switched off in order to melt the thing onto the surface of the CPU. This never would have worked in the past but this new safety feature means you can get away with it now.
If you're ever seeing 100C on your CPU, something is very wrong. Make sure you have no dead fans and that the heatsink is seated properly with thermal paste.
yeah man it unfortunately seems to be normal, ask my dell studio 15 (core i7, type 1557). this piece of crap overheats and BANG!!! got it stable now with some zalman notebook cooler, stays under 100° under full load and doesnt crash anymore. Putting down the frequency (933MHz is AFAIK the lowest) didn't help at all. It's a hardware design flaw.
cpufreqd practically does the same job. If your CPU get´s at 100°C you should invest in better cooling, instead of writing software that actually exists.
You're looking for a solution to a non-problem. Get a good CPU cooler. No software can help a broken hardware.
Hello, I would suggest that you can select two different thresholds for off and on, like on an iron (in "ironing clothes"). If you put the same threshold for up and down and have a high load on the computer, the system will switch between low and high frequency everytime the script is executed. Ultimately CPU throttling does this for you automatically, and the few processors I tried have more throttling levels than frequency levels. So your regulation could gain in accuracy by reducing one step of throttling everytime the frequency is lower than the target.
cat /proc/acpi/processor/CPUxx/throttling
The only problem is cpufreq tools are userland while I don't know to control throttling not being root.
cat /proc/acpi/processor/CPUxx/throttling
The only problem is cpufreq tools are userland while I don't know to control throttling not being root.
> everytime the frequency is lower than the target.
well actually I meant "everytime the temperature is higher than the target."
well actually I meant "everytime the temperature is higher than the target."
Please please don't automatically insert an apostrophe every time you use the word "its". Sometimes "its", requires an apostrophe; sometimes it doesn't. The way to determine whether "its" needs an apostrophe is to replace it with "it is". If the sentence still makes sense, use "it's". Otherwise, use "its". For example, "decided that it is better" makes sense, so you were correct to use an apostrophe there. However, "reducing it is speed" and "it does it is job" make no sense, and so you should have used "its" in both of those places.
If it's ok under MS windows (try), then it's a kernel bug. IIRC, if "fan" is disabled in the kernel config, the fan is always at its maximal RPM on my PC. (It must be better than roast PC.)
For your information, many reports that fan does not work after s2ram/s2disk [1].
[1] https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.22/+bug/113081
Good luck.
For your information, many reports that fan does not work after s2ram/s2disk [1].
[1] https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.22/+bug/113081
Good luck.
Well, I am on a passively cooled AMD 4850e (2 cores @ 2500 MHz max) and using the ondemand or conservative governor plus CPU driver and other thing I could find in the Kernel cfg (without any undervolting, I know there is a framework in Kernel but I don't know if it works with every mainboard so I would do this in the BIOS) I never come to critical temperatures, not even while compiling that what is now Libreoffice ;) or gaming. My other boxes run VIA C7 eden and with these you can freeze icecream ;) (ok they don't crunch numbers but hey).
I wonder what is causing your CPU to go so high. Pulling the powersave (statically lowest frq) sounds like a real emergency trick to me. Normally all these ondemand (or even userspace with a corresp. userspace part) thing should clock down rather quickly as soon as the CPU is not in much use. E.g. my 4850e now constantly runs a 2x1000Mhz and rarely ever comes up (emerge something and it will). Did you check that the governor you used before does clock down also? I had the counterpart once that conservative kept my VIA C3-2 laptop on 400Mhz even though it was compiling and that was a bug.
Undevolting sounds very nice but then you have to be careful not to stretch the limits too far or you'll encounter random errors when some transistors won't play the game.
Furthermore: Are you sure your sensor readings are correct? Normally it works today but some years ago I got really messy readings on some computers.
I wonder what is causing your CPU to go so high. Pulling the powersave (statically lowest frq) sounds like a real emergency trick to me. Normally all these ondemand (or even userspace with a corresp. userspace part) thing should clock down rather quickly as soon as the CPU is not in much use. E.g. my 4850e now constantly runs a 2x1000Mhz and rarely ever comes up (emerge something and it will). Did you check that the governor you used before does clock down also? I had the counterpart once that conservative kept my VIA C3-2 laptop on 400Mhz even though it was compiling and that was a bug.
Undevolting sounds very nice but then you have to be careful not to stretch the limits too far or you'll encounter random errors when some transistors won't play the game.
Furthermore: Are you sure your sensor readings are correct? Normally it works today but some years ago I got really messy readings on some computers.