On 30 Jun 00, at 6:39, Mikus Grinbergs wrote:
> In my opinion
> the main thing you have to worry about is how much hotter the inside
> of the chip is than what your probe (outside the chip) is measuring.
Yeah.
> I use a wild-eyed-guess of 5-8 degrees for this difference.
I suppose this depends critically on where the temp sensor is. With
Athlons the temp sensor isn't integrated into the chip package, you
have to stuff it between the fins on the heat sink. In these
circumstances you may be dropping a lot more than 8 degrees.
>
> In my own experience, I have seen CPUs run for years at above 50 degrees,
> and have *not* seen any CPUs stop working at even a measured 62 degrees.
My dual PII systems both use Supermicro boards which have integrated
sensors. Typically the M/B sensor runs 4C below the CPU temp, I
reckon to keep below 45C for safety.
One of the systems had the PSU fan fail, the temps rose to 55C (MB)
59C (CPU) at which point errors started. (So I rewound to a previous
"safe" save file when the fan was fixed.) Of course it could be that
the errors were due to overheated M/B components, or memory, rather
than the CPU.
> The torture-test mode of 'prime' should soon detect any CPU that
> misbehaves (for instance, because of heat). If you do not get any
> errors in 48 hours of running 'prime' (with room temperature as hot
> as you will ever have it) then my suggestion is to *not* worry about
> CPU temperature.
Agreed.
Regards
Brian Beesley
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
