Thermal pad or Thermal paste?

[Peter van der Goes]

hi peter,

i can't really pinpoint any reason for the increased cpu temps since the
case mods were done after a lot of research -- my case currently has 1 rear
exhaust fan, 1 psu exhaust, 1 front/bottom intake, 3 side panel
intake/suckholes (1 bang opposite the cpu cooler with a duct, 1 blowing air
onto the agp slot, and 1 towards the hdd area on the bottom/front) and 1 top
blowhole exhausting hot air. The positions were chosen carefully and i've
also tried to balance the intake and outtake fans.

Thus i can't really understand why the temp delta (cpu to case) is much
higher than before (24C vs 18C previously). but do bear in mind, that the
actual CPU temperature is about the same now with the case closed as it was
with the case open and a huge floor fan blowing into it -- it's just the
cpu-case delta that has gone out of whack, which could suggest the heatsink
ain't doing it's job well, even though the case is cooler now.

vin

I based my comments on personal experience with adding blowholes, etc with
little positive effect.
In the end, I've found that getting a good large case (I use Antec SOHO's)
with two 80mm exhaust fan mounts top-rear (just below the PSU and directly
behind the HSF) provide the best cooling with minimum noise. I've tried all
sorts of combinations of cool air intake, but the critical ingredient (based
on my limited experience - ~ 50 systems built for myself and others) is
getting the hot air coming off the HSF out of the case. I've found that once
you set up to exhaust that hot air, intake pretty much takes care of itself.
I use a single 80mm intake, bottom-front. No case-side fans at all. Using
2500+ Bartons on Asus and Epox mobos, clocked to 200x11 and Vantec Aeroflow
HSF's, my systems run at 27 - 28C case and max 44 - 46C CPU temps in an
air-conditioned home at ~ 24C.
Admittedly, my evidence is anecdotal, but from the first time I got a case
with the two exhaust fans, all my temperature problems ceased. Guess you
could say I'm an exhaust fan fan.
In your description, you upped the intake and didn't increase the exhaust
flow.

 

[w_tom]

Your evidence is confirmed by numbers. A single fan of
airflow - one fan blowing in and the other out - is more than
sufficient to cool any standard computer system. The only
reason why this would not be sufficient is a hotspot or a
system that consumes well over 500 watts.

Too many think that more air means that much more cooling -
a direct relationship. They don't first learn the numbers.
Number say the bottleneck in CPU cooling is mostly in the
heatsink assembly. Most important number is "degree C per
watt" for that heatsink assembly - not more fans - more CFM.

Your evidence is now more than anecdotal. It works
experimentally and is in agreement with the fundamental theory
- two conditions necessary for science.

If a computer fails in a 70 degree room, then the computer
has defective hardware. A computer with the airflow of one 80
mm fan must work just fine in a 100 degree F room. Too many
think heat is the reason for failure. Semiconductors that
cannot work at that higher temperature are defective now and
will probably be failing more often in the near future at 70
degrees. Too many blame heat only because they first do not
learn the numbers. Heat is how one finds defective
semiconductors before the warranty expires.

 

[w_tom]

A bare heatsink, properly machined, is more than sufficient
to cool a CPU. If thermal compound results in more than
single digit temperature decrease, then heatsink is
defective. It probably was not sold with the all so necessary
"degree C per watt" parameter. Therefore they could not be
sued for failing to meet a specification - for selling a
defective product.

That spec is the first thing one seeks when buying a
heatsink. If the heatsink was so bad that it did not mate
properly to CPU - needed thermal compound - then heatsink was
not even machined and it would not provide a "degree C per
watt" spec. Thermal compound only reduces CPU temperature by
single digits IF heatsink is acceptable.

Thermal compound must be as thin as possible. Thermal
compound works better than pads because most of heatsink must
make contact direct to CPU - no thermal compound between.
Thermal compound only fills microscopic holes between points
of direct contact. Thermal compound and pads do not conduct
as well as a direct to CPU connection. And thermal compound
must be applied so sparingly as to only spread out in the
center half of the CPU - where all heat is transferred from
CPU to heatsink. No thermal compound should be observed in
seam between heatsink and CPU. If it is squeezing out, then
far too much thermal compound was applied - and compound is
keeping heatsink from directly contacting CPU.

 

[David Maynard]

A bare heatsink, properly machined, is more than sufficient
to cool a CPU. If thermal compound results in more than
single digit temperature decrease, then heatsink is
defective. It probably was not sold with the all so necessary
"degree C per watt" parameter. Therefore they could not be
sued for failing to meet a specification - for selling a
defective product.

That spec is the first thing one seeks when buying a
heatsink. If the heatsink was so bad that it did not mate
properly to CPU - needed thermal compound - then heatsink was
not even machined and it would not provide a "degree C per
watt" spec. Thermal compound only reduces CPU temperature by
single digits IF heatsink is acceptable.

This must explain why not a single heatsink manufacturer provides their
heatsinks without thermal compound/pad and both Intel and AMD emphatically state
thermal compound must be applied and provide instructions on how to do so.

 

[David Maynard]

Vin said that the chemists said they didn't have *isopropyl* alcohol
(isopropanol). Alcohol is at least a dozen different things, from methanol
to glycerin.

(Also note that when he said chemist, he almost certainly meant what we call
a drugstore here in the US.)




We sure don't have a bottle. Is it mandatory?




Iodine tinctures, witch hazel, hydrogen peroxide, antibacterial salves and
lotions, or a whole lot of other stuff. Including other alcohols. Using
isopropanol as rubbing alcohol might be a regional thing, because I'm used
to ethanol.

Denatured Alcohol vs. Isopropanol

Denatured Alcohol is ethyl alcohol, or ethanol, containing traces of benzene.
While ethenol, in it's natural form, is a source of combustible fuel, it
contains trace amounts of water making it less efficient for burning than in
it's denatured form. Introducing Benzene during the distillation process drives
out the water contained in ethanol creating a very pure and potent fuel source -
denatured alcohol.
NOTE: Ethanol is the alcohol consumed in alcoholic beverages. However, once
denatured, ethanol becomes toxic and can cause blindness or death if consumed.


Rubbing Alcohol is isopropanol. It has a completely different molecular
structure than ethanol and often contains up to 30% water. Though it is
flammable it is a poor choice for fuel.

Isopropanol is also used for disinfection, and is often the ingredient in the
swab your skin is given before you receive a shot at the doctor's office. Common
thinking is that you should give isopropanol ten minutes to fully kill any germs.

Mixed with ammonia, it makes a great glass cleaner. On its own, it is a great
solvent for many cleaning jobs. Most computer users will be familiar with the
substance as a way to clean mouse balls, clean screens and keyboards, and in a
high enough concentration, clean magnetic drive heads in tape and disk drives.

----

In my experience "rubbing alcohol" has always been isopropyl although I can find
an occasional references to ethanol as an alternate, but it's then called Ethyl
Rubbing Alcohol in the ones I found. I did a quick survey of friends scattered
from New York and Massachusetts to Florida and Texas and all said isopropyl, not
to mention it was in their medicine cabinet for a quick check of the label.

See http://www.med-chem.com/Products/alcohols.htm

70% Ethyl Alcohol (Ethyl Rubbing Alcohol)

Alcohol used as a disinfectant for thermometers and as a skin antiseptic. 70%
ethyl alcohol is meant for external use only


70% Isopropyl Alcohol (Rubbing Alcohol)
Alcohol most commonly used for rubdown and general massage purposes. It is
used as an antiseptic for simple wound disinfection. 70% isopropyl alcohol is
meant for external use only.


Obviously that concept is behind the other poster's question on what folks use
for "scraped knee's." I, as you mentioned, have a number of other disinfectants
as well but I suspect the reason 'rubbing alcohol', meaning isopropyl, is so
often mentioned is it's historic propensity to 'always be there', hence handy,
due to it's other varied uses and low cost.

 

[Strontium]

-
David Maynard stood up at show-n-tell, in (e-mail address removed), and
said:

Denatured Alcohol vs. Isopropanol

Denatured Alcohol is ethyl alcohol, or ethanol, containing traces of
benzene. While ethenol, in it's natural form, is a source of
combustible fuel, it contains trace amounts of water making it less
efficient for burning than in it's denatured form. Introducing
Benzene during the distillation process drives out the water
contained in ethanol creating a very pure and potent fuel source -
denatured alcohol.
NOTE: Ethanol is the alcohol consumed in alcoholic beverages.
However, once denatured, ethanol becomes toxic and can cause
blindness or death if consumed.

You are confusing this with Methanol (wood alcohol). Denatured alcohol
(benzene dirtied ethanol) will not cause blindness. Death? Perhaps, if
enough is consumed. More than likely, though, cancer as well as a multitude
of other health problems.. Having worked with Benzenes for many years, I
think I can attest to this.


<snip>

 

[David Maynard]

-
David Maynard stood up at show-n-tell, in (e-mail address removed), and
said:




You are confusing this with Methanol (wood alcohol). Denatured alcohol
(benzene dirtied ethanol) will not cause blindness. Death? Perhaps, if
enough is consumed. More than likely, though, cancer as well as a multitude
of other health problems.. Having worked with Benzenes for many years, I
think I can attest to this.

It's not my confusion; I cut and pasted that from a web site and didn't pay
attention to the 'note'. My comments were after the ------ separator.

Anyone who watches enough old prohibition period movies knows the problem with
"wood alchy"

 

[David Maynard]

-
David Maynard stood up at show-n-tell, in (e-mail address removed), and
said:




You are confusing this with Methanol (wood alcohol). Denatured alcohol
(benzene dirtied ethanol) will not cause blindness. Death? Perhaps, if
enough is consumed. More than likely, though, cancer as well as a multitude
of other health problems.. Having worked with Benzenes for many years, I
think I can attest to this.

It's not my confusion; I cut and pasted that from a web site and didn't pay
attention to the 'note'. My comments were after the ------ separator.

Anyone who watches enough old prohibition period movies knows the problem with
"wood alcy"

 

[Strontium]

-
David Maynard stood up at show-n-tell, in (e-mail address removed), and
said:

It's not my confusion; I cut and pasted that from a web site and
didn't pay attention to the 'note'. My comments were after the ------
separator.

Anyone who watches enough old prohibition period movies knows the
problem with "wood alcy"

Ahhhhhh! Didn't see your "Pasted from etc, etc..." disclaimer

 

[w_tom]

Intel and AMD do not demand that everyone use thermal
compound. They say to ill informed hobbyists to install
thermal compound because experience has demonstrated that
"degree C per watt" is too complex for some assemblers. For
those with some experience, any heatsink without the "degree C
per watt" parameter is suspect - at best.

But then, since thermal compound is so cheap (tens of times
less than what the overhyped Arctic Silver costs), then why
not provide thermal compound so that the bad CPU installation
does not cause permanent failure. Thermal compound is good
compensation for the ill informed computer system assembler.

In one Intel paper for engineers, the negative effects of
thermal compound are discussed. But that paper is beyond
scope of this discussion. A properly machine heatsink without
thermal compound will only cause single digit temperature
decrease on CPU ... IF heatsink is properly machined. If
thermal compound results in better temperature decreases, then
one must ask how improperly a heatsink was applied OR how poor
the heatsink really is.

If you dispute this, then the theoretical numbers can be
posted, obviously, OR examples provided from scientifically
controlled experiments. In the meantime, decades of
experience says that thermal compound results in only single
digit temperature reduction if the heatsink is properly
machined. This is science that was old even thirty years
ago. Why promote the hype of Arctic Silver, et al? They
already sell a product that is grossly profitable. Instead
promote the well proven science.

 

[Phrederik]

A bare heatsink, properly machined, is more than sufficient

This must explain why not a single heatsink manufacturer provides their
heatsinks without thermal compound/pad and both Intel and AMD emphatically state
thermal compound must be applied and provide instructions on how to do so.

He's still correct. The best thermal interface is the flat, bare heatsink
against the core of the CPU. Unfortunately it's hard to account for all the
different sockets and other mounting mechanisms as well as poor handling by
the users, so paste is a lot safer.

 

[kony]

He's still correct. The best thermal interface is the flat, bare heatsink
against the core of the CPU. Unfortunately it's hard to account for all the
different sockets and other mounting mechanisms as well as poor handling by
the users, so paste is a lot safer.

Untrue, or at least not true in practice because there are no CPUs
THAT flat on top. The P4 spreader is nowhere near flat, nor is the
Athlon. The closest was the Coppermine P3 or Celeron, which unlike
the Athlon didn't have any laser-etching on the core, and even then
the final coating was often less than perfect.

Taking about theoretical best interface is meaningless when it is not
possible to achieve it with ANY off-the-shelf parts, without
refinishing them first. I'm fairly sure neither AMD or Intel will
warranty a CPU after it's planed down to perfection.

It may be true that two nearly-perfect parts (CPU & 'sink) will have
single-digit difference compared to using heatsink compound, but if
one is using one of the hi-spec 'sinks expected to start out somewhat
near flat, it's going to be providing effective enough cooling that
this single-digit difference may be a significant percentage of rise
over ambient temp. It is not madness to use thermal compound, many
components in electronic gear never expected to be seen by the eyes of
a consumer also use thermal compound for the benefit it provides.


Dave

 

[David Maynard]

Intel and AMD do not demand that everyone use thermal
compound. They say to ill informed hobbyists to install

Following installation instructions is not "ill informed."

thermal compound because experience has demonstrated that
"degree C per watt" is too complex for some assemblers. For
those with some experience, any heatsink without the "degree C
per watt" parameter is suspect - at best.

But then, since thermal compound is so cheap (tens of times
less than what the overhyped Arctic Silver costs), then why
not provide thermal compound so that the bad CPU installation
does not cause permanent failure. Thermal compound is good
compensation for the ill informed computer system assembler.

And where are these 'good installers'? Not at Dell, or Compaq, or HP, or any
other manufacturer because they all use some kind of thermal interface material.

In one Intel paper for engineers, the negative effects of
thermal compound are discussed. But that paper is beyond
scope of this discussion.

Useless since you don't say which one it is and there are many papers discussing
the 'negative effects' of thermal compound, as well as thermal pads, graphite
sheet, and tape, and all the other interface materials. Nothing is 'perfect' and
they all have 'negative effects'. They also have different 'advantages'. That's
what engineering is about: selecting the best solution for the specifications
and conditions one has to work with in the real world.

A properly machine heatsink without
thermal compound will only cause single digit temperature
decrease on CPU ... IF heatsink is properly machined. If
thermal compound results in better temperature decreases, then
one must ask how improperly a heatsink was applied OR how poor
the heatsink really is.

Alpha PAL PAL8942M81. Expulsion>0 .23 deg C/W, Induction>
0.26 deg C/W. Comes with thermal grease and installation instructions spell out
to apply it. There is no 'exception' mentioned for 'if you're a good installer'.

Now you show me one of your readily available 'proper' CPU heatsinks that comes
without thermal compound and explicitly says you don't need it because it's so
wonderful.

If you dispute this, then the theoretical numbers can be
posted, obviously, OR examples provided from scientifically
controlled experiments.

Having already disputed it THIS was the time for you to post something
supporting your claims.

In the meantime, decades of
experience says that thermal compound results in only single
digit temperature reduction if the heatsink is properly
machined. This is science that was old even thirty years
ago.

What "decades of experience" shows is that virtually everyone uses some form of
thermal interface material with virtually every heatsink application involving
significant power, from CPUs to stereo power output transistors. Crack open your
PSU and you'll find thermal interface material between the power transistors and
their heatsinks. Pop the can off your P4 and you'll find thermal interface
material between it and the CPU die. The stuff is everywhere.

Why promote the hype of Arctic Silver, et al? They
already sell a product that is grossly profitable.

Where in the world you come up with the notion I 'promote' Arctic silver, or any
other product for that matter, is a complete mystery.

Instead
promote the well proven science.

There's a difference between theoretical and applied science.

 

[David Maynard]

He's still correct. The best thermal interface

Define "best."

I submit that being able to make it should be part of the criteria.

is the flat,

How flat? How smooth? (even though you didn't mention it.)

bare heatsink
against the core of the CPU.

How are you going to get it "against the core" of a P4?

Unfortunately it's hard to account for all the
different sockets and other mounting mechanisms as well as poor handling by
the users, so paste is a lot safer.

Then without it isn't "best."

 

[~misfit~]

Untrue, or at least not true in practice because there are no CPUs
THAT flat on top. The P4 spreader is nowhere near flat, nor is the
Athlon. The closest was the Coppermine P3 or Celeron, which unlike
the Athlon didn't have any laser-etching on the core, and even then
the final coating was often less than perfect.

Taking about theoretical best interface is meaningless when it is not
possible to achieve it with ANY off-the-shelf parts, without
refinishing them first. I'm fairly sure neither AMD or Intel will
warranty a CPU after it's planed down to perfection.

It may be true that two nearly-perfect parts (CPU & 'sink) will have
single-digit difference compared to using heatsink compound, but if
one is using one of the hi-spec 'sinks expected to start out somewhat
near flat, it's going to be providing effective enough cooling that
this single-digit difference may be a significant percentage of rise
over ambient temp. It is not madness to use thermal compound, many
components in electronic gear never expected to be seen by the eyes of
a consumer also use thermal compound for the benefit it provides.

Thanks for that. I've been watching this thread and also thought the same
thing about dies not being completely flat. What's the use of having a
totally flat HS base if the die is concave or convex?

 

[David Maynard]

Thanks for that. I've been watching this thread and also thought the same
thing about dies not being completely flat. What's the use of having a
totally flat HS base if the die is concave or convex?

Actually, depending on what it is, a convex surface might be desired because it
might flex under load!

I'm not suggesting that's the case here. What I'm saying is that these things
are a heck of a lot more complex than the simple 'flat' kind of analysis folks
typically look at. Like, how much will the CPU package flex under the heatsink
clip pressure, as just one example.

Here's a pdf with, although not CPU related, some interesting information about
a particular component's heatsinking: http://www.ixys.com/t112901a.pdf

Note they intentionally convex the case to allow for deforming when it's bolted
down to the heatsink. Another interesting statement is "Therefore the flatness
of the heatsink across the module mounting area should be less than 25µm with a
roughness RZ less than 10µm which is a standard requirement for heatsinks."

Contact pressure is important so "it is recommended to use M5 screws secured by
a lock washer and flat washer torqued to between 2.7- 3.3 Nm (24 – 29 lb-inch).

And, of course, the ever present "The use of thermal grease is recommended to
ensure low case-to-sink thermal resistance." (and they list recommended types)


This link, http://www.thermaflo.com/interface.shtml , has a brief discussion of
thermal interface materials. Of particular note: "Unfortunately, no matter how
well prepared, when two typical electronic component surfaces are brought
together, less than one percent of their surfaces make physical contact. As much
as 99% of the surfaces are separated by a layer of interstitial air due to a
certain roughness caused by microscopic hills and valleys (surface finish) and
by macroscopic properties such as non-planarity in the form of a concave, convex
or twisted shape (flatness, twist, etc.)."

 

[We Live For The One We Die For The One]

Its safe then, your not dead are you

or are you :0

 

[w_tom]

The assumption being that heatsink to CPU surfaces must be
flat. However a better interface is formed with maximum
pressure between those surfaces that transfer most heat.
Almost all heat is transferred from CPU to heatsink in
center. That is where two surfaces typically have most
pressure - so that maximum amount of heatsink is in direct
contact with CPU where the heat is transferring.

 

[Matt]

Okay, it seems that to w_tom a temperature improvement of 9 degrees C is
of little consequence.

 

[Strontium]

Perhaps I'm the living dead I should be dead, by all rights! Having
worked (unsafely, I might add!) with many carcinogens and toxic solvents for
the better part of 14yrs....

-
We Live For The One We Die For The One stood up at show-n-tell, in
(e-mail address removed), and said: