Arctic Silver vs. Silicone Compound

[J. Clarke]

Who said anything about creating a layer? All I am talking about is
making sure you have got enough on there to make sure there is good
contact between
the CPU and the heatsink. Which you can very easily fail to do if you
don't put enough gloop on.

Sure, you don't want to go mad. I think this all started because people
were putting *loads* of the stuff on and it was ending up all over the
bridges. But if you put just enough to thinly cover the chip itself, then
thats fine. As I said before, any extra gets squeezed out. You don't have
any "layer" in between.

The AS instructions that you only need a "haze" of the stuff and highly
dangerous, unless you have a lapped heatsink and a lapped CPU. Otherwise
the lack of flatness of the two surfaces can easily mean you don't have
enough.

What AS instructions did _you_ read? The ones on the AS site say to use a
layer the thickness of a sheet of paper _unless_ you have a lapped HS and
CPU in which case they say you should use a "haze".

 

[Debug]

Someone makes one. Can't remember who, but they do exist.

Chip.

Would be interested to see it but I would'nt really buy one.

 

[Debug]

The pad is supposed to do that. First time the CPU comes to working
temperature the pad melts and if all went as designed the only place where
there's any of the material left is in irregularities in the surface.

Ok, then that flies in the face of Chips contention there should be a
thin layer between the cpu and HSF. My last cpu install I tried the
extremely thin layer method and it runs pretty cool.

 

[Chip]

What AS instructions did _you_ read? The ones on the AS site say to use a
layer the thickness of a sheet of paper _unless_ you have a lapped HS and
CPU in which case they say you should use a "haze".

Exactly. I have read their instructions and I disagree with them. I think
its bollocks.

They are interested in getting the lowest possible temps. 1C lower than
something else is a result for them. They don't give a stuff if you fry
your CPU. Personally, I think an extra 1C (even if I accept that it makes a
difference - which I am not sure I do) is worth it for the peace of mind in
knowing that your CPU is making good contact with the HS.

And like I have said already in this thread: put too much AS3/5 on and the
extra gets squished out of the way anyway. Even AS admit that, hence their
instructions for how to apply it to Intel chips.

Chip.

Chip

 

[Chip]

Ok, then that flies in the face of Chips contention there should be a
thin layer between the cpu and HSF. My last cpu install I tried the
extremely thin layer method and it runs pretty cool.

Sorry debug, but I have to correct this. I am *not* saying that there
should be any "layer" of compound between the HS and the CPU.

What I am saying is: make sure you put enough on. Any excess will get
squished out of the way. Don't go mad and drench the CPU in the stuff of
course. But make sure there is enough. The AS "apply a thin haze"
instructions are dangerous, imho.

Chip.

 

[J. Clarke]

Exactly. I have read their instructions and I disagree with them. I
think its bollocks.

They are interested in getting the lowest possible temps. 1C lower than
something else is a result for them. They don't give a stuff if you fry
your CPU. Personally, I think an extra 1C (even if I accept that it makes
a difference - which I am not sure I do) is worth it for the peace of mind
in knowing that your CPU is making good contact with the HS.

And like I have said already in this thread: put too much AS3/5 on and the
extra gets squished out of the way anyway. Even AS admit that, hence
their instructions for how to apply it to Intel chips.

???? They don't have different instructions for different chips. You said
that their instructions say to use a haze. Their instructions do not say
to use a "haze" except in the specific circumstance in which you yourself
say that that won't cause problems. If you are getting 1C less than the
guy next to you then how are you going to "fry your CPU"?

The extra doesn't get squashed out of the way quickly--takes time, possibly
long enough to fry your CPU if you get it _too_ thick.

 

[J. Clarke]

Sorry debug, but I have to correct this. I am *not* saying that there
should be any "layer" of compound between the HS and the CPU.

What I am saying is: make sure you put enough on. Any excess will get
squished out of the way. Don't go mad and drench the CPU in the stuff of
course. But make sure there is enough. The AS "apply a thin haze"
instructions are dangerous, imho.

Which might be true if the AS instructions said to do that. They do not say
to do that unless you have lapped surfaces that probably don't need it
anyway.

And before you go off about "1 degree cooler" and "fry your CPU" again,
anybody who goes to the trouble of lapping the heat sink and the surface of
the processor is going to _want_ that 1 degree cooler and with the lapped
surfaces the chance of "frying the CPU" due to insufficient heat sink
compound is approximately zero.

I don't know why you insist on misrepresenting the content of the
instructions. Or perhaps you are not a native speaker of English and have
trouble understanding that they give you two options, one for lapped
surfaces and the other for unlapped surfaces.

 

[Debug]

The AS "apply a thin haze"
instructions are dangerous, imho.

Chip.

Works for me.

 

[Chip]

???? They don't have different instructions for different chips.

Yes, they do actually! Which even further strengthens my argument.

You said
that their instructions say to use a haze. Their instructions do not say
to use a "haze" except in the specific circumstance in which you yourself
say that that won't cause problems.

But the problem is, it encourages people to use too little stuff. Which is
duff advice.

If you are getting 1C less than the
guy next to you then how are you going to "fry your CPU"?

Because if you use too little, you can end up with an inadequate contact
with the HS. So in your quest for this mythical 1C difference, you end up
with worse temps, perhaps even some system instability and ultimately even a
toasted CPU. What on earth is the point of all that, when you can avoid
*any* risk by simply applying a little more compound. And the only downside
is a *possible* (although not certain) tiny increase in temps. Its a
no-brainer. I could even argue that is better, because it forms a good
interface between the *sides* of the raised chip and the HS. Which you
won't get if you only put the barest minimum of compound on.

The extra doesn't get squashed out of the way quickly--takes time, possibly
long enough to fry your CPU if you get it _too_ thick.

Honestly this is nonsense. AS3 - even AS5, which is much thicker - is
instantly squished out of the way, by the installation of the HS. Try it:
take a CPU, put a fair amount of stuff on the top, apply the heatsink and
then remove it straight away. What you will see is a *very* thin layer of
compound on the CPU and heatsink at the point of contact and stuff sqeezed
out all around. The very thin layer is like a haze. A perfect interface
between chip and HS, formed without any risk.

*Critically*, you will probably also see that its thicker in some places
than others, because of the lack of flatness of both the CPU and the
heatsink. This alone quite clearly demonstrates what I am talking about:
if you try to start out with that amount, you might well not even make good
contact on certain parts of the chip.

Chip.

 

[Chip]

Which might be true if the AS instructions said to do that. They do not say
to do that unless you have lapped surfaces that probably don't need it
anyway.

OK, I can see you are getting all worked up about this, so lets start again.

All I am saying is that it is -in my humble opinion - better practice to put
a reasonable amout of compound onto the cpu, than to try to eek out the
barest minimum, which is what AS tell you to do. Their suggesting that you
manually try to apply a layer 0.003" thick is quite ludicrous.

And before you go off about "1 degree cooler" and "fry your CPU" again,
anybody who goes to the trouble of lapping the heat sink and the surface of
the processor is going to _want_ that 1 degree cooler and with the lapped
surfaces the chance of "frying the CPU" due to insufficient heat sink
compound is approximately zero.

True, its unlikely. But what is much *more* likely is that you won't
actually have enough stuff applied to even fill the gaps and the cooling
will be worse than if you applied more in the first place.

I don't know why you insist on misrepresenting the content of the
instructions. Or perhaps you are not a native speaker of English and have
trouble understanding that they give you two options, one for lapped
surfaces and the other for unlapped surfaces.

I am not trying to misrepresent anything. Yes, I agree the "haze" BS is
only supposed to be for lapped heatsinks. Its still BS. The problem with
it is, it encourages people to think that the thinner the layer they apply,
the better. And I strongly challenge that. In ideal conditions, seeking
out the very very finest performance, then AS may be right. But for your
average Joe, slapping the average HS on, I don't agree.

Incidentally, I have lapped my SLK900. It was only after I had done this
that I realized just how unlevel the top of the CPU can be. With normal HS
milling marks, you don't tend to see it, but the polished underside of a
lapped heatsink demonstrates is quite clearly: often the CPU core doesn't
sit completely level. One side might be a few microns higher or lower than
the other. If I apply a small amout of compound and then apply and straight
away remove the heatsink, I can see one side has a thicker layer of AS than
the other. Its still very very thin, but one side is almost completely
transparent and on the other side, less so. You need to make sure you have
enough AS to fill any gaps, even with lapped heatsinks!

Chip.

 

[J. Clarke]

OK, I can see you are getting all worked up about this, so lets start
again.

All I am saying is that it is -in my humble opinion - better practice to
put a reasonable amout of compound onto the cpu, than to try to eek out
the
barest minimum, which is what AS tell you to do. Their suggesting that
you manually try to apply a layer 0.003" thick is quite ludicrous.

That's actually a fairly thick layer. Take a vernier micrometer or a set of
calipers and set .003" and you'll find that it's actually quite thick. A
thousandth of an inch isn't as small a quantity as one who has never seen
it measured would expect.

True, its unlikely. But what is much *more* likely is that you won't
actually have enough stuff applied to even fill the gaps and the cooling
will be worse than if you applied more in the first place.

If it's lapped there aren't going to be that many gaps.

I am not trying to misrepresent anything. Yes, I agree the "haze" BS is
only supposed to be for lapped heatsinks. Its still BS. The problem with
it is, it encourages people to think that the thinner the layer they
apply,
the better. And I strongly challenge that. In ideal conditions, seeking
out the very very finest performance, then AS may be right. But for your
average Joe, slapping the average HS on, I don't agree.

The "average Joe", "slapping the average HS on", uses whatever comes in the
box or whatever was cheapest. Arctic Silver is not used by "the average
Joe".

Incidentally, I have lapped my SLK900. It was only after I had done this
that I realized just how unlevel the top of the CPU can be. With normal
HS milling marks, you don't tend to see it, but the polished underside of
a
lapped heatsink demonstrates is quite clearly: often the CPU core doesn't
sit completely level. One side might be a few microns higher or lower
than
the other. If I apply a small amout of compound and then apply and
straight away remove the heatsink, I can see one side has a thicker layer
of AS than
the other. Its still very very thin, but one side is almost completely
transparent and on the other side, less so. You need to make sure you
have enough AS to fill any gaps, even with lapped heatsinks!

If one side is thicker than the other, you didn't get the heat sink on flat.
Give it a week to settle and see what you get.

You're the one getting worked up over nothing. The simple fact is that you
aren't going to fry the chip if you use no heat sink compound whatsoever.
It's going to run hotter than you want and you may get instability, but it
takes more than the mere lack of some grease to actually damage the
processor.

 

[Chip]

That's actually a fairly thick layer. Take a vernier micrometer or a set of
calipers and set .003" and you'll find that it's actually quite thick.

But your not applying it with a vernier micrometer, are you. With a credit
card is what AS suggest. And your hand. In that context, 0.003" is quite
ludicrous.

A
thousandth of an inch isn't as small a quantity as one who has never seen
it measured would expect.

Maybe. But I do know how thick 0.003" is.

If it's lapped there aren't going to be that many gaps.

Nonsense. There are still gaps are at the microscopic level. The AS5
particles are as small as .00002", just to fill such gaps. Lapping only
really takes the "huge" milling marks off - which helps, but it doesn't fill
in the microscopic gaps. Even the best lapping kits only get you down to 10
microns. At a rough (no pun intented) guess, the surface contact area is
probably halved without filling in these microscopic gaps with some
compound.

And apart from anything, this completely ignores the fact that the CPU core
probably isn't flat to anything better than this tolerance either.
Undertstandable when you consider that the chip manufacturers recommend a
thermal pad anyway.

The "average Joe", "slapping the average HS on", uses whatever comes in the
box or whatever was cheapest. Arctic Silver is not used by "the average
Joe".

OK, the above average Joe. But there's still a lot of people messing around
with PC's as a hobby, who don't really know what on earth they are doing.
You only have to look at some of the posts.

If one side is thicker than the other, you didn't get the heat sink on flat.
Give it a week to settle and see what you get.

And how, pray, did I manage this feat; not getting the heatsink on flat? I
sat it squarely on the four rubber feet and pressed down the clip. Oh, I
must have missed the toenail clipping jammed under one end. Or perhaps not.
;-) LOL.

You're the one getting worked up over nothing.

I am not even remotely worked up as you can see.

The simple fact is that you
aren't going to fry the chip if you use no heat sink compound whatsoever.

Are you sure? I wouldn't like to try it.

It's going to run hotter than you want and you may get instability, but it
takes more than the mere lack of some grease to actually damage the
processor.

Is this pure speculation on your part? Personally I would think you are
cruising for a damaged CPU if you were to try this. I could be wrong
though.

Chip.

 

[Debug]

All I am saying is that it is -in my humble opinion - better practice to put
a reasonable amout of compound onto the cpu, than to try to eek out the
barest minimum, which is what AS tell you to do. Their suggesting that you
manually try to apply a layer 0.003" thick is quite ludicrous.

http://www.subzeropc.com/article/thermalinterface.htm
When applying thermal paste you don't want to apply too much.
Remember, the idea is to fill in the tiny air gaps between the
heatsink and CPU. A thick layer of paste will actually decrease
performance since it prevents direct contact. Apply the compound to
the contact side of the CPU and make sure that 100% of the surface is
covered. Make sure that a very thin layer is applied so that the
contact surface is still visible. Too much paste will cause a gap
between the heatsink and the CPU, this means the heatsink is not even
in direct contact with the CPU.

 

[Chip]

http://www.subzeropc.com/article/thermalinterface.htm
When applying thermal paste you don't want to apply too much.
Remember, the idea is to fill in the tiny air gaps between the
heatsink and CPU. A thick layer of paste will actually decrease
performance since it prevents direct contact. Apply the compound to
the contact side of the CPU and make sure that 100% of the surface is
covered. Make sure that a very thin layer is applied so that the
contact surface is still visible. Too much paste will cause a gap
between the heatsink and the CPU, this means the heatsink is not even
in direct contact with the CPU.

So what's this supposed to prove? Its still bollocks.

If you can be bothered, why no do yourself a little test:

1. Put a *tiny* amount of AS onto a CPU, as per their instructions.
Install heatsink, remove heatsink, inspect CPU.

2. Put a little more AS on the CPU. Not loads, just a bit more, to make
sure there will be good contact with the HS. Install heatsink, remove
heatsink, inspect CPU.

Do you see *any* difference between 1 and 2, when you look at the CPU? No.
Because in (2), all the excess get squished out of the way anyway.

Chip.

Chip.

 

[J. Clarke]

So what's this supposed to prove? Its still bollocks.

If you can be bothered, why no do yourself a little test:

1. Put a *tiny* amount of AS onto a CPU, as per their instructions.
Install heatsink, remove heatsink, inspect CPU.

2. Put a little more AS on the CPU. Not loads, just a bit more, to make
sure there will be good contact with the HS. Install heatsink, remove
heatsink, inspect CPU.

Do you see *any* difference between 1 and 2, when you look at the CPU?
No. Because in (2), all the excess get squished out of the way anyway.

Define "tiny amount" and "little more" quantitatively. One man's "tiny
amount" is another man's "vast quantity".

 

[Debug]

So what's this supposed to prove? Its still bollocks.

If you can be bothered, why no do yourself a little test:

1. Put a *tiny* amount of AS onto a CPU, as per their instructions.
Install heatsink, remove heatsink, inspect CPU.

2. Put a little more AS on the CPU. Not loads, just a bit more, to make
sure there will be good contact with the HS. Install heatsink, remove
heatsink, inspect CPU.

Do you see *any* difference between 1 and 2, when you look at the CPU? No.
Because in (2), all the excess get squished out of the way anyway.

Chip.

Chip.

My own testing has shown me that less is better than more.

 

[Debug]

So what's this supposed to prove?

I posted it because it backs up what I say. You have implied in this
thread that I know noting and you know it all. I say bollocks to that.

 

[Chip]

I posted it because it backs up what I say. You have implied in this
thread that I know noting and you know it all. I say bollocks to that.

Hey Debug, I am sorry if I have offended you. Really I haven't meant to
imply anything of the sort. Clearly I don't know everything, and clearly
you do know a lot.

The only reason I harp on about this is because the IT industry gets on
these "bandwaggons". Someone somewhere decides something is a good idea,
and suddenly everyone is saying it like its some devine god-given truth.
You see it on this very newsgroup all the time.

I wonder how many of these review sites that are on the
use-a-really-thin-layer-of-as bandwaggon, have ever actually bothered to
test whether its true or not? That's the reason I questioned your quoted
article: it *proves* nothing; its just another 1 on the bandwaggon.

Well I have tested it. Personally, I could detect no difference in temps
depending on how much AS I applied. I am not saying there is no difference:
merely that if there is any, it is small and I couldn't detect it.

So based on that, I think its safer to put a little more AS on - rather than
a little less - just to be on the safe side.

I realize this flies in the face of all "evidence" (i.e. lots of other
poster all saying the same god-given "truth"). Thats why I harp on about
it. But just because I am the only 1 saying this, and everyone else
disagrees with me, it doesn't mean I am necssarily wrong.

Anyway, sorry once again if you have felt offended: that was not my
intention.

Chip.

 

[GMAN]

Is one or two degrees even worth the price of the tube? I think a 6gram tube
sells for like $6.00.

That tube can be used for multiple processors. Plus if one spends hundreds of
dollars on a speedy processor, whats $6 to make sure its cooled properly.

 

[Wblane]

Even more importantly is the fact the cheap rat-shack silicone TIM breaks down
after 1 year of use. Try using that crap and a year later your system will be
overheating big time.
WHen I had this problem and took a
look at the cheap silicone shit it was all dried out and flaky -- which is
great for a pie crust but not much good for thermal conductivity.

That tube can be used for multiple processors. Plus if one spends hundreds of

dollars on a speedy processor, whats $6 to make sure its cooled properly.

-Bill (remove "botizer" to reply via email)