cpu overheating

Ed Greshko Ed.Greshko at greshko.com
Tue Dec 12 08:43:23 UTC 2006


Mike McCarty wrote:

> No, just cheaply designed. There are MANY processors which cannot
> run GIMPS, for example, because they have "good enough for normal
> case" cooling designs.

Change the above to read "MANY systems" instead of "MANY processors".  As
you have noted, it is the bad cooling design that causes the problem...not
the processor.

>> (possibly in the BIOS) *may* be able to slow the CPU down if it
>> exceeds some defined temperature threshold, but that's a
>> cross-your-fingers type of safeguard against a failed cooling system.
> 
> You completely missed my point. Dismissing this as "well, your
> hardware shouldn't do that" is
> 
> S-T-U-P-I-D
> 
> because, even though it means the hardware is marginal at best,
> it is a hint that
> 
> THERE MAY BE A DEFECT IN THE SOFTWARE.

So, you are saying that if I run the GIMPS torture test and my system
overheats then there "may" be a defect in the GIMPS software?

> And that should not be ignored.
> 
> My point wasn't that his or any CPU *should* be overheated
> by bad software. My point was that hints that there might
> be a defect in the software should not be ignored.

I think (hope) you mean *could*.

FWIW, I have a Dual Xeon 2.80GHz system with 2GB of RAM.  It runs my DNS
server, web server, etc.  I also run VMware and normally have at least 2
virtual machines running.  One an XP VM the other some variant of Linux.  At
times I run a RHELv4 team with 4 systems plus the XP.  During the day time I
limit GIMPS to running on 2 CPUs (hyper-threading enabled) and nice value at
19.  At night, a cron job runs and gives GIMPS all 4 CPUs and a negative
nice value.  On some summer days, since my wife is rather slim, I don't turn
on the A/C until the room temp is above 30.  My system is on 24/7.  Never
had a heat issue.  But, I do check my fans and dust out the system on a
regular basis.  Need to do that when you live in Taipei and you have 3 house
cats.

Sorry, I don't buy into your theory that overheating may be hinting at
"software with defects".  IMHO, if that were the case the virus/hackers of
the world would be putting out "defective" code with the goal of burning up
everyones systems.  :-)




More information about the fedora-list mailing list