test of thermal pad on AMD

amd's web site recommends pads, not grease.

If grease is better, why are they recommending pads?

~Aart wrote:
amd's web site recommends pads, not grease.

If grease is better, why are they recommending pads?

Less likely to bugger it up with a simple pad...


--
remove the underscores to unmung the email address...

On Mon, 17 May 2004 12:36:13 -0700, wrote:


I'm glad this topic came up. I wanted to ask all of you if you've
ever seen a degadation in performance of AMD's thermal pads. I built
an Athlon XP 2200 in April 2003 using the stock HSF and pad along with
standard case fans and no overclocking, and it's been 24/7 ever since.

A year ago the idle temp ranged between 30-40 deg C depending on the
ambient room temp, and full sustained load temp of 47 at most. Now in
the past month or two the idle temp tends to hang aroung 40-45 and the
full load temp has gone up to 52.

I cleaned out the case and HSF fins of dust multiple times and it
didn't make a difference. The fan RPMs are still reporting 4000 rpm
in the BIOS so I don't think it's the fan.

Is there a limitation in the life of AMD's thermal pads? I have fears
that it's thermal transfer properties are going to break down
completely. Should I remove the HSF and pad, sand it smooth, and use
some Artic Silver?

Yes, if you suspect that the TIM isn't working effectively you "COULD"
take the 'sink off, thoroughly clean off original pad material (petroleum
solvet works well for this, but do not scrape with a metal object). If
you then find the 'sink bottom to be rough you might mildly lap it with
fine-grit sandpaper (at least 320 grit or finer for the final pass).

If you lack any polishing compound then any generic thermal compound or
especially arctic alumina will polish it up some... not as good as many
polishing compounds, but "good enough", and the residue is even
beneficial, doesn't need thoroughly cleaned off but if it remains in the
tiny crevaces of the metal it's a GOOD thing.

However, it is not necessary to use arctic silver. With a heatsink having
a good surface the difference in using arctic silver isn't significant.

On the other hand, a full load temp of 52C is not too hot, it should still
be completely stable. There are also other reasons why the temp might
read differently, for example a bios update can change the temp offset
used by the bios and immediately show a temp difference similar to what
you've seen. If you were overclocking it would be more of a concern but
it may not be worth the bother to do anything at 52C.

Stacey wrote:
~misfit~ wrote:

Stacey wrote:


Not a fair test. The pad needs a few days of hard work to heat up and
squeeze out the excess material. If you'd run SETI for a week with
the pad, checked temps, then tried goo it would have been valuable
data, as it is, it's junk science.
--


It ran 24/7 for 2 days, how long does it need to run to -start-
working right?

48 hour should be getting close, *if* it was under 100% load for that time,
say, running Prime95. It's phase-change material, meaning it needs to get
hot for a while to do it's thing.

BTW have you ever looked at the surface on the bottom of the newer
HS's they give you in the retail box? Looks like it was surfaced with
a 40 grit grinder.

Yep, they're pretty rough alright.
--
~misfit~

On Mon, 17 May 2004 15:11:21 -0700, wrote:


Thanks for the response. Maybe I should leave it alone. I know that
the 52 is not considered hot since AMD's spec sheets say 90 as a max,
and the machine has always been completely stable. I was just
concerned that the change is an indicator of pad breakdown and I'm
scared that it will breakdown completely. I never updated the BIOS
since there were no updates from the MB manufactuer in the past year.
This is strictly some kind of breakdown occurring in the pad and/or
HSF. I never touched the CPU/HSF after I built the machine, other
than spraying air at it to blow out the dust from the fan blades and
fins.

I was at Fry's yesterday looking at new HSF assemblies and considering
some of the simpler ones, but they are all so large that I'm not sure
if they'll fit. The only way to test them is to actually install them
and if it turns out they don't fit then I'll be forced to go back to
the stock HSF and follow your procedure below.


If it's going to worry you or if you see a continual decline in thermal
transfer, that is, ever-increasing temp, then it may be worthwhile to
remove 'sink, lap and use compound just for that piece of mind. Although
previously I wrote that Arctic Silver isn't necessary, there is a benefit
to synthetic compounds in general, that in long term use on hot small CPU
cores they tend to stay mixed, not separating so readily as silicone based
compound can. Generally with a heatsink using silicone compound and a
very hot CPU with such small core area (Athlon or Duron) I'd plan to
reapply compound every 18 months or so, but certainly it depends how hot
it gets... some of my boxes run full load for hours on end and certainly
generate more heat than others.

As for the Frys heatsinks, there are no Frys here and i"m not familiar
with their selection but in general it's good to buy biggest 'sink that
will fit, one that accepts an 80x25mm fan. Larger metal surface area
(will an all copper bottom) allows lower CFM airflow, in conjunction with
the large fan, can result in quieter operation, longer fan lifespan, and
often a longer interval between cleaning.

AMD designates what's called a "keep out zone", which is the area around
the CPU socket in which manufacturer of motherboard should avoid placement
of any components that interfere with heatsinks meeting same keep out zone
specification. Generally most 'sinks do meet this, with notable exception
of Thermalright SLK-900 and possibly others with wing-like fins, possibly
the Zalmans also. A careful measurement of the board should determine if
a 'sink will fit, you can get specs on the 'sink measurement at it's
manufacturer's website or many online vendors. In some situations I'll
get a 'sink knowing it's barely too big and simply take a hacksaw,
grinder, or dremel tool to the conflicting area on the 'sink, as I really
like to use biggest 'sink possible within reason, due to aforementioned
benefit of allowing selection of optimal fan.

Still, with your 'sink having degraded performance yet still keeping CPU
at 52C, you should be content with that heatsink, there's no benefit to
lower CPU temp unless you want to push the theoretical limits of CPU
lifespan and try to use it for a couple decades or longer.

"Ben Pope" writes:
Don Taylor wrote:
Does anyone know of a relatively cheap smoke detector like device
BUT it will switch off maybe 1000 watts of power when it thinks
that something has caught fire?
....
Cheap smoke alarm with a relay? Maybe you'd want some help with keeping the
relay contacts connected - I suspect it would eat batteries. The other
concern is of course fail safety... no point in having a battery open
circuit the relay - when the battery runs out, it'll close again.

Besides, if the thing is on fire, there seems little point in switching it
off. Kind of a bit late?

I was thinking that cutting off the power might help stop feeding the
fire as the fan was going up in flames.

I actually observed this happen with a house fan a few decades ago.

On Mon, 17 May 2004 19:32:46 -0500, (Don Taylor)
wrote:

"Ben Pope" writes:
Don Taylor wrote:
Does anyone know of a relatively cheap smoke detector like device
BUT it will switch off maybe 1000 watts of power when it thinks
that something has caught fire?
...
Cheap smoke alarm with a relay? Maybe you'd want some help with keeping the
relay contacts connected - I suspect it would eat batteries. The other
concern is of course fail safety... no point in having a battery open
circuit the relay - when the battery runs out, it'll close again.

Besides, if the thing is on fire, there seems little point in switching it
off. Kind of a bit late?

I was thinking that cutting off the power might help stop feeding the
fire as the fan was going up in flames.

I actually observed this happen with a house fan a few decades ago.

Why not just put a solenoid valve on a fire extinguisher and connect that
to the fire alarm?

wrote:


Is there a limitation in the life of AMD's thermal pads? I have fears
that it's thermal transfer properties are going to break down
completely. Should I remove the HSF and pad, sand it smooth, and use
some Artic Silver?



That's what it sounds like. I was shocked at how course the finish was on
the HS, very little of the surface would ever touch the die prior to
sanding it smooth. I'd guess it would work as well now with no thermal
compound as it did with the original surface finish and their pad.
--

Stacey

~misfit~ wrote:

Stacey wrote:


BTW have you ever looked at the surface on the bottom of the newer
HS's they give you in the retail box? Looks like it was surfaced with
a 40 grit grinder.

Yep, they're pretty rough alright.


Maybe that's the real reason why they are saying not to use paste anymore?
The retail HS's are so rough it wouldn't work?

Seriously, I would bet a lapped HS would cool as well as this rough one did
with a pad. But no, I'm not going to test that on this machine. :-)

--

Stacey

kony writes:
Don Taylor wrote:
"Ben Pope" writes:
Don Taylor wrote:
Does anyone know of a relatively cheap smoke detector like device
BUT it will switch off maybe 1000 watts of power when it thinks
that something has caught fire?
Besides, if the thing is on fire, there seems little point in switching it
off. Kind of a bit late?
I was thinking that cutting off the power might help stop feeding the
fire as the fan was going up in flames.
I actually observed this happen with a house fan a few decades ago.

Why not just put a solenoid valve on a fire extinguisher and connect that
to the fire alarm?

I'm trying to get somewhere in that direction. Since I think we would
agree that the probability of fire is very small but we want the
probability it will successfully work if there is a fire to be very
high then I'm a little hesitant to believe in a hobby-modification
project to make this work. I am still looking at all the possible
solutions to this though. Thanks