cpu overheating

[Gordon Messmer]

Mike McCarty wrote:
> 
> I give up. Linux is a religion. Linus is the Pope. No one can
> question it. Heretics are vilified and run out of town on a rail.
> 
> Here we have a chance to take 5 minutes to run top, and see
> what the CPU utilization is. It would literally take five
> minutes.

No, it isn't.  In fact, I personally think that Linux is pretty awful. 
It's less awful than some alternatives, but awful all the same.

> It costs actually considering the possibility that Linux might
> have one line of code which is not perfect.

Linux has lots of imperfect lines of code.  We all know that.  That's 
not the reason that we're trying to convince you that the problem is 
cooling.

> It's futile to suggest actually to collect any data on Linux behavior
> if there is any threat it might reveal a wart on its nose, because
> the suggestion will be rejected out of hand.

There's no point in collecting data.  The OP already told us that the 
system overheated when he was running an application that uses a lot of 
CPU time.  Beyond that, Linux does lots of things that Windows doesn't, 
and will use more CPU resources.  Off the top of my head: all the 
default cron jobs (possibly run at another time of day by anacron), and 
beagled on the Gnome desktop.

Looking around for software culprits will only distract the OP from the 
problem: the CPU isn't sufficiently cooled.  That needs to be corrected.

We get that GIMPS has an application to test for poor cooling, already. 
  You know what?  That doesn't mean that "it's a well-known fact that 
faulty software can overheat CPUs".  At best, it means that the GIMPS 
team doesn't want to be blamed when a system that's already failing goes 
tits up.  If your system can't cool a system under sustained CPU 
utilization, then it's only a matter of time until it fails. 
*Something* is going to make that CPU overheat.

What, exactly, do you suggest that software applications do?  Should 
they be designed to sleep() enough to let the CPU get a breather? 
People don't buy a 2Ghz CPU to get 1Ghz of processing.  Most systems 
already throttle back if they detect a cooling failure, but failing 
that... the CPU could fry.  It happens.  It's never the software's fault.

Overheating is always a failure of the cooling system.

I don't mean to be rude, but have you considered that Linux is not a 
religion, and that there is no conspiracy to suppress the wisdom that 
you bring, but that you're simply wrong on this one point?