Thermal paste application

[Ratz O. Fratzo]

Properly lapped heatsinks with mirror finishes will only
require a translucent haze."

I think that is the key phrase there. Most heat sinks aren't machined
to that specification.



______________________

-=Ratz O. Fratzo =-

 

[W7TI]

I mean the following passage at that link which implies
that a properly thin layer of thermal paste (perhaps thin
enough be a mere haze) wouldn't be able to ooze out:

"The flatter the mating surfaces, the thinner the layer that is
required. Stock processors and/or heatsinks with normal
surface irregularities will require a layer 0.003" to 0.005"
thick as shown below to fill the resultant gaps. (Equal to
the thickness of about 1 sheet of standard weight paper.)
Properly lapped heatsinks with mirror finishes will only
require a translucent haze."

_________________________________________________________

I don't doubt that the above statement is true, but I believe it to be
irrelevant for nearly all of us. Who among us has ever seen a "Properly
lapped heatsinks with mirror finish"? The purpose of the thermal
compound is to compensate for real-world gaps, peaks, concavities,
convexities and other irregularities.

If you think about what's happening when you press the heatsink down in
place, what difference does it make if a little extra oozes out? On the
other hand what difference does it make if you don't have enough?
Potentially a lot of difference. I'm not talking about oozing major
gobs that drip all over. Just enough to see around the edges.

If you're sure you have no such irregularities waiting to bite you,
spread it as thin as you like.

 

[Timothy Daniels]

I don't doubt that the above statement is true, but I believe it
to be irrelevant for nearly all of us. Who among us has ever
seen a "Properly lapped heatsinks with mirror finish"?
The purpose of the thermal compound is to compensate for
real-world gaps, peaks, concavities, convexities and other
irregularities.

If you think about what's happening when you press the
heatsink down in place, what difference does it make if a
little extra oozes out? On the other hand what difference
does it make if you don't have enough? Potentially a lot
of difference. I'm not talking about oozing major gobs
that drip all over. Just enough to see around the edges.

If you're sure you have no such irregularities waiting to bite
you, spread it as thin as you like.

I agree that a layer thick enough to ooze a little doesn't hurt,
and that it gives acceptible thermal conductivity, and that what
you profess is what the CPU manufacturers expect in the real
world from average do-it-yourselfers. My question is really
about the best one can do without lapping and without
laboratory conditions. And I'd expect that the thinnest layer
that a real world person could manage (using a razor edge per
Arctic Silver's instructions) that is 0.003" thick would be so thin
that the viscosity of the paste would prevent oozing. IOW,
a really good (i.e. really thin) layer wouldn't ooze even when the
parts are pressed together, i.e. oozing implies that the layer is
thicker than it could be. Whether one could get the two paste
layers flat enough and to be able to mate the two parts slowly
enough to not trap air between them is probably a bigger question -
one which would lead me instead to put a small dab in the middle
and then to lap the two parts together to work the paste out
without allowing air in. But, as I've said, that is just my intuitive
guess and not the result of experiment.

*TimDaniels*

 

[JAD]

ever use spot putty to fill a scratch in a paint surface....that's what the AS is supposed to accomplish....the term coat doesn't
apply, as it is not mean to cover the surface, but rather to fill imperfections...really there should be no ooze.

 

[Gary W. Swearingen]

without allowing air in. But, as I've said, that is just my intuitive
guess and not the result of experiment.

An experiment occurs to me: Get a glass bezel off a flashlight and
pretend to mount that on a CPU or simulator. See how easily air
pockets form and how translucent (==thin) the goop gets.

 

[Timothy Daniels]

"Gary W. Swearingen" ignited and said:

An experiment occurs to me: Get a glass bezel off a flashlight and
pretend to mount that on a CPU or simulator. See how easily air
pockets form and how translucent (==thin) the goop gets.

That's a pretty good idea. Could, uhh... you do that for us?


*TimDaniels*

 

[jeffc]

If you follow the Arctic Silver instructions (i.e. spreading on a film just
thick enough to discolor the surface), the small gap and the viscosity
of the paste would probably prevent any "oozing". This comment taken
from the instructions is indicative of the thickness of the layer of paste:

"The flatter the mating surfaces, the thinner the layer that is required.
Stock processors and/or heatsinks with normal surface irregularities
will require a layer 0.003" to 0.005" thick as shown below to fill the
resultant gaps. (Equal to the thickness of about 1 sheet of standard
weight paper.) Properly lapped heatsinks with mirror finishes will only
require a translucent haze."

Yeah, I'm sure it's a matter of definition. If you think of oozing like
batter from a waffle iron, then no. But even with mirror finishes, under a
microscope I'm sure you'd see something. Probably not what the OP meant, I
concede.

 

[jeffc]

Apply only enough thermal compound so that it spreads thin
in middle half of CPU. If too much compound is applied - if
it oozes out - then thermal compound also obstructs contact
between CPU and heat sink. Furthermore, thermal compound
oozing out and getting on CPU contact pins means potential
electrical problems. Never apply so much thermal compound as
to ooze out.

Not necessarily true. It oozes out because there was too much, and the
force of contact brings metal to metal where possible. If you put too much
on and it DOESN'T ooze out, then you have a problem.

 

[w_tom]

Apply no thermal compound - no problem. Put too little on,
still no problem but a small increase in thermal
conductivity. Put right amount on and get better thermal
conductivity. Put too much on and get less thermal
conductivity AND get problems with CPUs electrical pins.

If so much has been put on as to ooze out, then human does
not understand the concept. Too much thermal compound is the
worst of all applications.

Remember why thermal compound is being applied. For any
acceptable heatsink, CPU temperature will drop no more than 10
degrees compared with no thermal compound. And since so many
buy inferior heatsinks, even too little thermal compound
compensates for that poorly manufactured, overpriced heatsink
or a heatsink improperly applied.

BTW properly machined heatsink does not mean mirrored
finish. It has more to do with basic machine surface AND
pressure applied between CPU and heatsink. But many heatsinks
have no surfacing let alone shaped to apply pressure in the
tiny region where almost all heat is transferred.

But too much thermal compound only means one or many bad
things can happen. Too much thermal compound is indicated by
oozing.

 

[W7TI]

Put too much on and get less thermal
conductivity

_________________________________________________________

Absolute nonsense. You can believe any damn thing you want. I'm tired
of talking about it. End of conversation.

 

[Craig]

Arctic Silver III is considered best thermal paste you can
buy because ... it is equivalent to most every other thermal
compound; but at something like 10 times the price. Arctic
Silver does noteven provide thermal conductivity numbers.
Why bother? They are not selling to people who first want
facts. They are selling to those more influenced by
propaganda and hype

Sure they do...your just not looking in the right place
Here is information on the brand new Arctic Silver 5
http://www.arcticsilver.com/as5.htm

Craig

 

[w_tom]

Too much thermal compound means a hotter CPU. Now for
formulas:
theta = p * t / A where
theta is the thermal resistance in degree C per watt.
We want to decrease theta to decrease CPU temperature -
decrease thermal resistance.

p is the thermal resistivity constant of thermal compound
expressed in 'degree C - inches per watt'.

t is the average thickness of thermal compound - in 'inches'.

A is contact area between heatsink and CPU - in 'square
inches'.

The only variable is thermal compound thickness. Increase
that thickness - put on too much thermal compound - and
thermal resistance (theta) increases. Thinnest thermal
compound means better conductivity. But then of course. The
more heatsink directly touches the CPU means better heat
transfer. Thermal resistance increases with change of
medium. Resistance is higher in 'CPU to thermal compound to
heatsink' than it is in 'CPU to heatsink'. Again, why thermal
compound must be applied minimally - no oozing.

Unfortunately there are many who just know this is wrong -
science be damned.

 

[JAD]

tom....

.....over done.....well done... burnt....stick a fork in it...
as soon as the algebraic equations come out you know this post is bordering the lunatic fringe

 

[HMSDOC]

Me again..I started the thread :>)

I read the Arctic Silver directions and assume they would be the same for the
thermal compound that comes with the CPU (no thermal tape in mine, just a
syringe with compound). For the *Intel P4* as opposed to CPUs with smaller
heatsinks (if that is the proper term for the metal plate) it clearly says not
to ooze at all. There is even a picture of removing the fan after the
fan/heatsink has been properly applied to show that the entire plate on the CPU
top should, in fact, not even be totally covered...just the segment over the
core. Believe it may be different for non P4s but I didn't read that part
carefully.

Howard

 

[Timothy Daniels]

I read the Arctic Silver directions and assume they would be
the same for the thermal compound that comes with the CPU
(no thermal tape in mine, just a syringe with compound). For
the *Intel P4* as opposed to CPUs with smaller heatsinks
(if that is the proper term for the metal plate) it clearly says not
to ooze at all. There is even a picture of removing the fan after
the fan/heatsink has been properly applied to show that the
entire plate on the CPU top should, in fact, not even be totally
covered...just the segment over the core. Believe it may be
different for non P4s but I didn't read that part carefully.

From a physics point of view,
1) the thinner the layer of thermal compound, the better;
2) air gaps are bad, even microscopic air gaps;
3) for very thin gaps between planar surfaces, viscosity
is a very strong impediment to fluid flow.

Conclusion: If it oozes without pressure, it's definitely
too thick. If it oozes with applied pressure, it's thicker
than it has to be. If it doesn't ooze at all, even with
applied pressure, it may be too thin. Therefore,
just a little ooze with applied pressure is probably
the best one can do for peace of mind.

From a manufacturer's point of view:
1) People are idiots;
2) People are fumble-fingers;
3) The product has to work even when applied by People.

Conclusion: It doesn't matter too much HOW the
friggin' thermal compound is applied, just keep
out the bubbles.


*TimDaniels*

 

[jeffc]

Apply no thermal compound - no problem. Put too little on,
still no problem but a small increase in thermal
conductivity. Put right amount on and get better thermal
conductivity. Put too much on and get less thermal
conductivity AND get problems with CPUs electrical pins.

No, not necessarily. Only if you put WAY too much on.

If so much has been put on as to ooze out, then human does
not understand the concept. Too much thermal compound is the
worst of all applications.

Again, define "ooze". How much? If you're thinking something like jelly
oozing out from a 6 year old's sandwich, then yes, that's way too much. But
too much if it is only slightly too much is not the worst - not at all.
It's better than too little. As long as it's only a tiny bit too much,
which is not difficult to get right. What's difficult - nay, impossible -
is *exactly* the right amount.

 

[jeffc]

Me again..I started the thread :>)

I read the Arctic Silver directions and assume they would be the same for the
thermal compound that comes with the CPU (no thermal tape in mine, just a
syringe with compound). For the *Intel P4* as opposed to CPUs with smaller
heatsinks (if that is the proper term for the metal plate) it clearly says not
to ooze at all. There is even a picture of removing the fan after the
fan/heatsink has been properly applied to show that the entire plate on the CPU
top should, in fact, not even be totally covered...just the segment over the
core.

Of course - we are only talking about the core "plate". And again, for the
millionth time, "oozing" is not necessarily a bad thing. It should be an
amount that you can barely see, like with a magnifying glass, if you could
actually see under the heat sink anyway, which you can't.

 

[JAD]

Again, define "ooze

exactly... its an arbitrary term with NO definition that will cover all areas. like I said before...its supposed to FILL defects in
the surface, NOT create its OWN surface.. the result of ANY coming out of the side, melting/changing the viscosity COULD result in
contamination of the solder points of the chip....I would rather caution on the side of NOT ENOUGH than too much...

 

[w_tom]

Some years back, four Intel engineers wrote a paper that
described the entire heat transfer process out of CPU. Even
included getting heat out to CPU surface. The future of CPU
cooling was described as phase change materials. Anyone
looking for a better CPU cooling alternative may be interested
in these phase change materials rather than the so often
overhyped Arctic Silver:
Ultrastick from Thermalloy-

http://www.thermalloy.com/products/options/greases.shtml#ultra
8040905C from ThermaFlo-
http://www.thermaflo.com/ifmat_phase.shtml

 

[w_tom]

tom....
....over done.....well done... burnt....stick a fork in it...
as soon as the algebraic equations come out you know this post
is bordering the lunatic fringe

DINNER'S READY!