[mythtv-users] OT: CPU temperatures

Jack Trout witmore1 at gmail.com
Mon Nov 1 01:59:45 UTC 2004


Well printed on the chip is the max operating temperature symbol as
part of the model number, on all my athlon  chips (in my lifetime I
have had 5 4 2000+'s and and 1 2200+) and each were coded for 90
degrees, now I think thats the borderline of where it will work, but I
would never run a system at 90 degrees,
It is a proven fact that onboard systems for reading temperature are
not as accurate as reading the temperature on the die itself, I havent
checked into it lately but last check AMD did not offer and processors
with ondie temperature checks, so the best place to check die temp is
between the processor and the heatsink, I dont measure it from there,
but tests have shown the average between the die temp and a
motherboard sensor for temp is about 10-15% difference (onboard
sensors are often underneath the processor in the middle of the
socket) the sensors that are more accurate have a probe that actually
makes contact with the bottom of the processor
Knowing this I always give a 20% safe buffer which means 70 degrees is
the max temp I will allow my processor to run at. but in practice I
have anti-burn shield enabled on my motherboard and the shutdown temp
is 60 degrees, and warning temp of 55, my office normally maintains a
decent 70-75 degrees farinhiet which with 2 120mm case fans moving air
through my case, and a 80mm slow rotation processor fan (3000max rpm)
it stays around 38-45 degrees, on a xp2000 overclocked to equiv of a
2100 running at 80% load, if you case cant move the warmed air out of
the case you might have other issues than the size of the fan


On Mon, 1 Nov 2004 00:22:35 +0100, Maarten <mythtv at ultratux.org> wrote:
> On Sunday 31 October 2004 20:48, Adam Felson wrote:
> > You're overcautious.  Anything under 70 DEG C is fine and frankly 70-75
> > DEG C is nothing to worry about.
> 
> Yeah well, I beg to differ... because of a.o. the following points:
> 
> * Reported temperature may not be quite accurate to true temp. It can even be
> FAR off, depending on how and where it is measured (IN-die, external, ...).
> An ambient temperature is entirely different than die-temperature. And can you
> be certain the temp you're reading is measured where you think it is ?
> * Silicon ages way faster with elevated temperatures. It could easily halve,
> or much worse, the life expectancy of components. Thus, cooler==better.
> But you may say "oh well, in 18 months I'll have a new system anyway", sure.
> * Hot components may show unexpected results, not failures but
> miscalculations, instabilities and things of the sort.  I'm not saying they
> do per se, but they might. Very very difficult to track that sort of bugs.
> * What happens if you run your unattended-24x7 mythtv box @70 C and your airco
> dies / there is a sudden heatwave ? Bye bye mythtv box... fried for sure.
> You need to allow for special external circumstances, thus the need for some
> buffer zone or tolerance or whatever you may call it. 10C seems reasonable.
> 
> Case in point, remember the IBM drives that were flaky ? And in the end IBM
> specified them as "desktop drives, not to be used longer than 8 hours a day"?
> That was a clear-cut case of being especially sensitive to high(er) temp
> environments.  I have had several of those drives and not a single one of
> them have died, just because they were in a *really* well cooled server. (In
> fact they remained *cold* to the touch, that's how well cooled they were...)
> 
> Higher-than-normal temperatures are not good for humans, and aren't great for
> computers either.  Do you work just as hard and efficient as usual when the
> airco is off ?  Well, computers show similar behaviour. They *will* operate
> @70C, but they do not like it. (or maybe you were talking about fahrenheit?)
> Same goes for combustion engines.  They fact that they can go up to a real
> high RPM does not mean they encourage you to drive on the interstate in third
> gear. But sure, they can and will if you just want (push) them hard enough.
> 
> So no, I really don't feel I'm being overcautious.  But YMMV.
> 
> Maarten
> 
> -- 
> 
> 
> 
> 
> _______________________________________________
> mythtv-users mailing list
> mythtv-users at mythtv.org
> http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
>


More information about the mythtv-users mailing list