larry said:
One thing for sure is that you applied the correct amount of thermal
paste -- not too little (no voids seen), and definitely not too much
(100% coverage but with transparent valleys between all those peaks,
indicating near perfect thickness), so removing any paste won't help
you. And because the compound shows lots of peaks and valleys, that
indicates it's creamy and can flow freely when squeezed, therefore
preventing a thick layer from remaining in place.
The spring force between the heatsink and CPU is probably OK, and it's
common to be able to wiggle the heatsink horizontally a bit. You
can't make the CPU punch through the mobo by increasing the force, and
with your motherboard's setup for clamping the heatsink in place,
there's probably no risk in increasing the force (and probably no
benefit).
Which 2.1 GHz Athlon X2 dual core? Desktop versions seem to range
from 45W maximum (BE2350) to about 90W (4000+ X2?), and that Spire
heatsink is rated for a maximum of 95W. Spire's website says the
maximum fan speed is 2,000 RPM, contrary to the 3,000 RPM claim on the
box, and that heatsink seems to be all aluminum, but I believe the
heatsinks AMD supplied with my 45-65W Athlons had copper slugs in the
middle, but if not, at least each aluminum base was thicker than the
Spire's, up to about 1/2" in the middle. OTOH their fans are just
70mm, vs. 92mm for your Spire's, and they typically spin at 1,300 -
1,600 RPM and the CPUs run at 50-55C under moderate load.
50-55C CPU temperature is nothing to worry about, except to junk
quality electrolytic capacitors. However your capacitors don't show
any bulging or oozing, and some of them seem like Nichicon brand model
HM(M) (about 5-6 visible in your second photo, one is near the word
"GIGABYTE" printed on the motherboard and the boxy coil marked
"2R0"). Those are really good and long lasting unless they were made
from about 2001-2004. You also probably have some United/Nippon
Chemicon 3,300uF capacitors, about 7 above the CPU, near the boxy
coils marked "R60". Those capacitors should be fine if they're model
KZE, but Chemicon still has problems with their KZG and KZJ series,
which the experts at BadCaps.net say can fail even if they don't
bulge or ooze. If I had to choose between running junk capacitors at
cool temperatures and running quality capacitors at hot temperatures,
I'd definitely pick the latter every time. Bad capacitors cause
problems with reliability, not high CPU temperature.
Sorry for being long-winded, but I think your temperatures are normal,
and maybe you should simply increase the temperatures for the high
temperature alarm and shutdown. 60-70C is actually safe for
operation. Also it wouldn't surprise me if the Spire heatsink wasn't
very good.
What if it's shutting down on THERMTRIP ? I wouldn't
expect THERMTRIP to be configurable by software. Some other
shutdown method, could be configurable.
I wish I'd noticed that URL from earlier in the thread. On the
http://www.spire-corp.com/main/product_detail.asp?ProdID=1140
page, it lists theta_R as 0.388 C/W, so a person can actually
do their own thermal analysis.
http://www.gigabyte.com/support-downloads/cpu-support-popup.aspx?pid=2818
"Socket AM2 - NVIDIA nForce 520LE - GA-M52L-S3 (rev. 2.0)
Socket AM2
AMD Athlon X2 BE-2350 2100MHz 512KBx2 Brisbane 65nm G2 45W 2000 F6
AMD Athlon X2 BE-2350 2100MHz 512KBx2 Brisbane 65nm G1 45W 2000 F6
AMD Athlon 64 X2 4000+ 2100MHz 512KBx2 Brisbane 65nm G1 65W 2000 F6
AMD Sempron LE-1200 2100MHz 512KBx2 Sparta 65nm G1 45W 2000 F6
Socket AM3
Socket AM2+
AMD Phenom X3 8450 2100MHz 512KBx3 (but this is a triple core)
So the processor could be the 4000+ (to get as oldish as possible).
If the computer case air is 35C, then 35 + 65W*0.388C/W give a processor
case temp of 60C max. Nothing wrong with that. If the temperature is going
higher than that, then some assumption in that calculation is off. Either
the computer internal case temperature is higher than 35C, the processor
is actually drawing more than 65W, or the theta_R of the heatsink is higher
than 0.388C/W.
At 0.388C/W, that's the same as, or a little worse than, a retail cooler.
Some of those are in the 0.33C/W ballpark. A good enthusiast cooler,
one of those monster ones, can be 0.12C/W or so. What a monster cooler
can't help with though, is if the inside of the computer case isn't
cool enough. And then, a bigger fan on the back helps.
One other possibility, is the cooler isn't sitting flat on the processor,
but you might notice that, in the pattern in the paste when it is
disassembled. If you clean off the surfaces, apply a rice grain of
paste in the center, squish them together, then disassemble again
later, the circular squish pattern tells you something about how
the surfaces meet.
If there are old and new paste materials involved, the old stuff should
be removed first. For example, if there is a phase change compound, that
is rather hard, that can elevate the heatsink and prevent good contact.
I had one setup, where it took some scraping to get that stuff
off (I didn't use sandpaper, and used a tool with an edge to
get it off).
In the BIOS on page 45 I see
ftp://download.gigabyte.ru/manual/motherboard_manual_m52l-s3_rev.2.0_e.pdf
"System/CPU Warning Temperature
Sets the warning threshold for system/CPU temperature
BIOS will emit warning sound.
Options 60C, 70C, 80C, 90C"
If that is set up high, then maybe it'll be less annoying to
use Speedfan or the like, to watch how high the CPU temp
actually goes. THERMTRIP will shut it down, if it is too high.
On the cpu-world.com web site, they list things like this for
AMD processor feature set. What the first two are exactly, I
don't know. But the third one, is a signal that comes out
of the processor, and is used to turn off the ATX supply
by deasserting PS_ON#. THERMTRIP is the option of last
resort, and still works, even if the processor has crashed.
Hardware thermal control
Software thermal control
THERMTRIP
I can find a posting, where someone left their computer sitting
in the BIOS, watched the temps in the hardware monitor screen,
until it overheated and shut off. They list 105C as the last
temperature measured. That would be a silicon die temperature,
one would hope. To convert that to a case temperature (temp
of the metal on the outside), subtract about 20-25C or so. So
that would be about 80C case, which is 20C higher than the
60C case we expect by calculation. And your processor should
only draw the 65W, when it's running prime95. If it overheats
at idle, at that point it should not be drawing 65W.
HTH,
Paul
