CPU Overheating Problem When Running 4 CPU Intensive Threads with Latest Untained Kernel

[Efraim Flashner]

Also, nice -19 is higher priority than nice 19.

my netbook generally is in the mid 70s for temperature.  When doing
something intensive, such as starting chrome or installing a new kernel
(update-initramfs or rebuilding modules) it'll go up to 99C, and then
go back down over time.  I have found that with the catalyst drivers my
netbook runs cooler and uses less battery, but I haven't had it
overheat and shutdown.  If it regularly is near the top of the thermal
limit then you should be worried, but only occasionally shouldn't be a
big deal.

I did have a laptop that when using 3/8 threads at 100% would pass its
thermal limit and shutdown.  That one was clearly poorly designed.

-Efraim



On Mon, 19 May 2014 10:12:23 +0300
"E.S. Rosenberg" <esr+linux-il at g.jct.ac.il> wrote:

> Afaik all modern CPUs are fitted with mechanism that will slow them
> down or turn them off when they get too got, those mechanisms are on
> die and I'm pretty sure you can't disable them from the OS.
> 
> 
> 2014-05-19 8:07 GMT+03:00 Oleg Goldshmidt <pub at goldshmidt.org>:
> 
> > Shlomi Fish <shlomif at gmail.com> writes:
> >
> > > I'm getting many high temperatures in the log:
> > >
> > > coretemp-isa-0000
> > > Adapter: ISA adapter
> > > Physical id 0:  +98.0°C  (high = +82.0°C, crit = +102.0°C)
> > > Core 0:         +98.0°C  (high = +82.0°C, crit = +102.0°C)
> > > Core 1:         +97.0°C  (high = +82.0°C, crit = +102.0°C)
> > >
> > > They eventually are dropped but it's still alarming.
> > >
> > > My questions are:
> > >
> > > 1. Is this a problem with the kernel? (Muli? Anyone? Can you
> > > comment?)
> >
> > You can either wait for Muli's authoritative response about the
> > kernel or run a test with a signficantly different kernel version
> > yourself. It is reasonable to assume that if the kernel were to
> > blame for CPU overheating for *any* reason at all it would have
> > been noticed by people in any version that is more than a month
> > old. So if your cutting edge 3.15.0-rc5 runs hot and some
> > reasonably stable kernel (preferably distro-compiled and not
> > self-compiled to avoid Kconfig weirdness on your end) doesn't you
> > may start to suspect something.
> >
> > If you do get suspicious run your test (with both kernels) on a
> > different computer - convince someone to boot off your USB or
> > something. Then you may decide to report that 3.15.0-rc5 with a
> > given configuration seems to act weirdly on more than one system.
> > If it does, of course.
> >
> > In general, I wouldn't blame the kernel too soon. You do not
> > provide any information on your computer's cooling facilities, form
> > factor, enclosure type or brand, or, indeed, provenance (brand name?
> > self-specced and self-assembled? overclocked by an unscrupulous
> > neighbourhood computer shop without your knowledge?  etc.). Could
> > it be that its cooling is insufficient at high load? Have you
> > verified that all your fans work fine?  Is air flow good enough
> > (you won't know, but [some] big brands are more likely to design
> > for that)? Does anything (from dust to a wall that is too close)
> > clog or block ventilation holes? Are the sensors correcty located,
> > well calibrated, and working properly?
> >
> > As an additional wild hand-waving, does anything change if you
> > switch hyperthreading off in the BIOS?
> >
> > --
> > Oleg Goldshmidt | pub at goldshmidt.org
> >
> > _______________________________________________
> > Linux-il mailing list
> > Linux-il at cs.huji.ac.il
> > http://mailman.cs.huji.ac.il/mailman/listinfo/linux-il
> >



-- 
Efraim Flashner    
efraim.flashner at gmail.com 4096R/CA3D8351 created: 2013-10-08
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
URL: <http://mailman.cs.huji.ac.il/pipermail/linux-il/attachments/20140519/7cc2bdf6/signature.asc>