On Sunday January 26 2003 01:43 am, Paul wrote:
> Yes, they both show on LM Sensors (through Gkrellm).
> The CPU (Athlon 1200) shows 78C (149F) and something else shows
> 29.2C (75.5F). I am not sure if I set the multiplier readings in
> Gkrellm correctly, but this is what I also could see (plus or minus
> a few degrees) when then machine was still upright. So no change
> there.
Type 'sensors' in a terminal if you aren't sure you have gkrellm
configured correctly. If 'sensors' is showing 78C, that's your
"weirdness" problem. First of all, AMD Athlon's (unlike Intels) don't
have an internal diode for reporting the actual core temp. An
external probe is used. AMD says to add 10 to 20C to the probe temp
to better approximate the actual core temp.
IME, about 12C (10C from a pin probe, to 15C for a contact probe)
is the typical amount to add. To illustrate, silicon is an insulator.
Measuring the internal temp of a cpu with an external probe is sort'a
like tryin to measure the temp of electrical wires inside a wall, by
mashing a thermometer against the plaster. OK, maybe that's a little
exageration, but you get the idea ;)
AMD specs Athlons to fail at 90C core temp. This upper limit will
be lowered if you overheat the core even once, but it usually takes
several times. When you see 78C from the probe, your core temp is
probly 90C, maybe even as high as 98C. Poof! You're lucky the system
quits, rather than completely fryin that Athlon and the motherboard.
Re-mount your heatsink on that athlon squarely and firmly using
thermal grease, not a thermal pad. The $2 grease from Radio Shack is
all you need. Apply a thin layer. Keep your case temp as close to
room temp as possible. You should be able to keep that cpu under 55C
from the probe, even under sustained 100% load. 'Course the 55 probe
is really upper 60's core temp, but far enough under 90C that even
occasional spikes in the core temp shouldn't cause failure.
After fixin your heatsink, see if the system won't run properly
(upright). If it's still runnin too hot, try pointing a table fan
into the case with the cover off. If you can then run at more normal
temps, you need to improve your heatsink, fans and case ventilation.
If you still have "weirdness" with improved cooling, it's likely you
overheated the core one too many times and the cpu is internally
damaged. You may have also caused some damage to the capacitors on
the motherboard, particularly those surrounding the cpu socket.
If you suspect you've permanently damaged the cpu/mobo, all is
still not lost. Underclock it, if the mobo supports it. The
multiplier on that 1200 is not locked. Keep the FSB at spec, but try
dropping the multiplier by .5 to as much as 2. IOW's, if it's 9x133,
try droppin the multiplier to 8 or 8.5. If it's a 100mhz FSB cpu
(12x100), try 10 or 11x100. Might keep you goin till you can replace
the system. If it comes to that, I strongly suggest you only use AMD
recommended power supplies, motherboards, and heatsinks, if you
aren't now doin so.
--
Tom Brinkman Corpus Christi, Texas
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
