CPU thermal paste?

[*Vanguard*]

"do_not_spam_me" said in

I have 2 old jars of Delta Bond thermal epoxy but haven't used them,
except when I needed a putty-like epoxy, and to glue on heatsinks I've
used only silicone rubber RTV, except in the case of large heatsink
and a small contact area, where I used regular epoxy. Thermal epoxy
for a 300 MHz K6-2 is an utter waste of money.

Silicone rubber to affix a heatsink? Geez, why not just thermally
isolate the heatsink with a thick rubber washer and some contact cement?
Why even bother affixing the heatsink at all? It would be better to not
use anything and just press the heatsink and CPU together dry to get as
much metal to metal contact as possible. There are some silicone-only
based thermal pastes for something like 15 cents a blister tube which
are about the same as using double-stick thermal tape (a 1" x 6" strip
good for 2 or 3 uses costs $2). Even the alumina-based thermal pastes
are much better than silicone only (especially compared to what you
mentioned). A single-purchased blister costs all of 65 cents. Skip the
small fries with your next burger meal if you're that hard up for pocket
change. If you really need to use a thermal adhesive and you don't want
to pay the high bucks for the Artic brand (which seems to be the most
popular brand offered almost to exclusivity), make your own
(http://www.overclockers.com/tips683/).

There's a reason no one includes silicon *rubber* in their testing of
thermal compounds in test reports, like
http://www.techwarelabs.com/reviews/cooling/thermal_paste/ and
http://www.thetechzone.com/reviews/roundup/thermal_compound/index.shtml.
Silicone rubber is a thermal *insulator*, not a conductor. Yeah, it
transfers heat 4 times faster than air but that still sucks! Might as
well as use Hard As Nails from a glue gun ("Hey, hold on Joe. While
you're gluing those 2x4's on the basement block, dab a bit of that Hard
As Nails on my heatsink, will ya?"). Thermal paste (and worse for
thermal adhesive) aren't great for thermal transfer but they beat air
and silicone by a long ways. Silicone *rubber* RTV (room temperature
vulcanizing; i.e., it sets at room temperature) to attach a heatsink? I
don't think so. That's the stuff you use as a moisture barrier or
sealant, like around your bathtub, or to insulate wire connections, like
gooping it into a wire nut for an outdoor wire splice. It is NOT to
provide thermal transfer between a CPU and heatsink, or thermal transfer
between any surfaces. You might as well as use double-stick foam tape
(the super dense type). Thermal tape would be better (and less messy)
but is still at the low end. Go read
http://www.azom.com/details.asp?ArticleID=920#_Room_Temperature_Vulcanising.

Thermal transfer:
air ~ 0.035 W/mK
silicone rubber = 0.14 to 0.22 W/mK
3M thermal tape = 0.4 W/mK
silicone/zinc grease or paste = 0.4 to 0.9 W/mK
copper filled silicone grease = 2.0 W/mK
alumina (Artic or generic) = 4.0 W/mK
silver filled silicone grease = 4.2 W/mK
Artic Silver epoxy = 7.5 W/mk
Artic Silver 3 = 9.0 W/mK
Artic Silver 5 = 350,000 W/mK (claimed)

I don't see how Artic can claim their Silver 5 compound, a *paste*, has
a thermal transfer rate that is 814 times greater than solid silver
itself. When I Googled around, no one actually tested this claim and
instead just quotes Artic's own specs. By the way, little known fact
(well, it was to me until now): there was no Artic Silver 4 because "4"
in Japanese is "shi" which means "death" (see
http://www.techaddicts.net/ArcticSilver5/AS5_1.html). Guess they didn't
want a thermal compound with a connotation of "Artic Silver Death".
Yeah, put this on, then your CPU dies. ;->

The biggest problem I see with users applying thermal paste is they goop
on way too much of it. Metal-paste-metal doesn't transfer heat as fast
as metal-metal contact. The paste is only supposed to fill any
microscopic or concavity voids that would otherwise be filled with air.
Of course, if you would rather rely on the most metal-to-metal contact
to provide the greatest rate of thermal transfer, welcome to the world
of lapping your heatsink (and CPU plate which is hazardous).

Gold?

Hmm, just had a thought but no time to think it through. Silicone,
zinc, aluminum, and silver, in that order, give increasiningly better
heat transfer rates. Wouldn't gold be even better? But imagine the
price for gold-filled compounds!

Oops, never mind. Did some searching to find the thermal conductance of
some metals, which are (along with some other materials):

Diamond = 1000 to 2600 W/mK (wow!)
Silver = 430 W/mK
Copper = 390 W/mK
Gold = 320 W/mK
Aluminum = 236 W/mK
Platinum = 70 W/mK
Glass = 1 W/mK
Air = 0.035 W/mK
Styrofoam = 0.03 W/mK

The above are for solid materials, not for pastes filled with micronized
filler using the above materials. Just imagine the cost for
diamond-filled thermal paste! Ouch.

 

[*Vanguard*]

"~misfit~" said in news:[email protected]:

JT wrote:

I have a mate who's into electronics and I remember him using
"thermal transfer compound" on some of the transistors/heatsinks on
his home-made power amplifiers 20+ years ago. I don't kmow what it
was made of but I do remember him telling me it was really toxic.
Zinc oxide? That seems to ring a (faint) bell.

Zinc oxide is used as a filler in some low-grade thermal pastes. It is
not toxic. Zinc oxide is that water insoluable white stuff you see
painted on folks' noses to keep from getting sunburned.

"Zinc oxide is ideal for the nose and ears while swimming"
(http://www.healthcentral.com/mhc/top/001869.cfm).

I don't remember reading any warnings about not touching the stuff. In
fact, it usually needs to be reminded to users to NOT use their fingers
to spread around the paste because of the oil on their fingers.
However, I doubt any of the stuff is suitable for consumption (but
touching it is okay).

Maybe your friend was talking about a thermal adhesive because those do
use solvents that are toxic (but not in the dosage to which you would be
exposed for one heat sink, but maybe if you were on an assembly line).
I mean, hell, SuperGlue is toxic but I don't see folks keeling over that
accidentally got their fingers glued together. Acetone is toxic yet I
occasionally dampen a paper towel with it to remove some glue, residue,
or such, and my fingers are touching that Acetone wetted paper towel.
Grain alcohol is toxic, but are you going to give up your beer? If so,
I'll take it.

 

[~misfit~]

"~misfit~" said in news:[email protected]:


Zinc oxide is used as a filler in some low-grade thermal pastes. It
is not toxic. Zinc oxide is that water insoluable white stuff you see
painted on folks' noses to keep from getting sunburned.

"Zinc oxide is ideal for the nose and ears while swimming"
(http://www.healthcentral.com/mhc/top/001869.cfm).

I don't remember reading any warnings about not touching the stuff.
In fact, it usually needs to be reminded to users to NOT use their
fingers to spread around the paste because of the oil on their
fingers. However, I doubt any of the stuff is suitable for
consumption (but touching it is okay).

Maybe your friend was talking about a thermal adhesive because those
do use solvents that are toxic (but not in the dosage to which you
would be exposed for one heat sink, but maybe if you were on an
assembly line). I mean, hell, SuperGlue is toxic but I don't see
folks keeling over that accidentally got their fingers glued
together. Acetone is toxic yet I occasionally dampen a paper towel
with it to remove some glue, residue, or such, and my fingers are
touching that Acetone wetted paper towel. Grain alcohol is toxic, but
are you going to give up your beer? If so, I'll take it.

Yeah. It obviously wasn't zinc oxide, that just jumped into my head. It was
some sort of metal oxide though I think. I remember he took great pains not
to get it on his skin and there was a red warning label on the tube warning
about transdermal absorbtion and toxicity. It wasn't an adhesive, it was a
thermal goop, to help heat-transfer. He used it 'tween the transistors and
the heatsink before he bolted them on. I still see him very occaisionally,
I'll ask him next time, I'm curious now. He's currently just about finished
making his own helicopter, completely home-made except for the rotors, he
saw those in someone's barn and got them for $20, then built a helicopter to
go with them. Single-seater, powered by a converted 120hp outboard motor
engine, the only other part of it he didn't build himself. (Although he made
an awesome exhaust manifold/extractor system for it which increased it's
power output considerably. It looks like a work of art, tuned according to
some arcane book he has on racing manifolds for two-stroke engines).

 

[JT]

Yep, I know that. I was talking specifically about the PC side of things. I
have a mate who's into electronics and I remember him using "thermal
transfer compound" on some of the transistors/heatsinks on his home-made
power amplifiers 20+ years ago. I don't kmow what it was made of but I do
remember him telling me it was really toxic. Zinc oxide? That seems to ring
a (faint) bell.

One of the best early ones was from 3M and used their newly developed
silicones. Still pretty much the standard for most applications.

JT

 

[VWWall]

Oops, never mind. Did some searching to find the thermal conductance of
some metals, which are (along with some other materials):

Diamond = 1000 to 2600 W/mK (wow!)
Silver = 430 W/mK
Copper = 390 W/mK
Gold = 320 W/mK
Aluminum = 236 W/mK
Platinum = 70 W/mK
Glass = 1 W/mK
Air = 0.035 W/mK
Styrofoam = 0.03 W/mK

The above are for solid materials, not for pastes filled with micronized
filler using the above materials. Just imagine the cost for
diamond-filled thermal paste! Ouch.

Might not be too high. But using industrial grade diamond dust might
not be much better than silicone grease. Diamond's high conductance
relies on the alignment of its crystals, (as does its brilliance).
There is some work being done using "diamond like" layers on silicon
wafers used to fabricate CPUs. Could be a signifigant developement in
the ever increasing heat density of modern CPUs.

Virg Wall

 

[JT]

Gold?

Hmm, just had a thought but no time to think it through. Silicone,
zinc, aluminum, and silver, in that order, give increasiningly better
heat transfer rates. Wouldn't gold be even better? But imagine the
price for gold-filled compounds!

Oops, never mind. Did some searching to find the thermal conductance of
some metals, which are (along with some other materials):

Diamond = 1000 to 2600 W/mK (wow!)
Silver = 430 W/mK
Copper = 390 W/mK
Gold = 320 W/mK
Aluminum = 236 W/mK
Platinum = 70 W/mK
Glass = 1 W/mK
Air = 0.035 W/mK
Styrofoam = 0.03 W/mK

The above are for solid materials, not for pastes filled with micronized
filler using the above materials. Just imagine the cost for
diamond-filled thermal paste! Ouch.

Gold might actually be a good alternative in the form of Gold Leaf.
Relatively inexpensive at hobby shops, and Gold leaf is one of the
thinnest, most malleable materials available. A sheet smoothed onto the top
of the processor would need no paste to hold it, would be soft enough to be
crushed into the little crevices and gaps, and has very good heat
conduction. Gold leaf is only a few atoms thick, so excess is unlikely.
Just have to make sure there was no extra that would overhang and short.
Will have to experiment with it when I get a chance.

JT

 

[~misfit~]

Gold might actually be a good alternative in the form of Gold Leaf.
Relatively inexpensive at hobby shops, and Gold leaf is one of the
thinnest, most malleable materials available. A sheet smoothed onto
the top of the processor would need no paste to hold it, would be
soft enough to be crushed into the little crevices and gaps, and has
very good heat conduction. Gold leaf is only a few atoms thick, so
excess is unlikely. Just have to make sure there was no extra that
would overhang and short. Will have to experiment with it when I get
a chance.

As you said, it's just a few atoms thick. I think the 'sink and maybe the
CPU would have to be lapped.

 

[JT]

As you said, it's just a few atoms thick. I think the 'sink and maybe the
CPU would have to be lapped.

Of course lapped surfaces would be the best, or maybe use '2' sheets .
Hope I get a chance to try this soon. Can just see it now. Arctic Gold

JT

 

[~misfit~]

Of course lapped surfaces would be the best, or maybe use '2' sheets
. Hope I get a chance to try this soon. Can just see it now. Arctic
Gold

If you do I'd really like to know the results. I'm sure others would too.

 

[*Vanguard*]

"JT" said in news:[email protected]:

Gold might actually be a good alternative in the form of Gold Leaf.
Relatively inexpensive at hobby shops, and Gold leaf is one of the
thinnest, most malleable materials available. A sheet smoothed onto
the top of the processor would need no paste to hold it, would be
soft enough to be crushed into the little crevices and gaps, and has
very good heat conduction. Gold leaf is only a few atoms thick, so
excess is unlikely. Just have to make sure there was no extra that
would overhang and short. Will have to experiment with it when I get
a chance.

JT

I had actually thought of using gold leaf (several sheets) and
originally had it in my post about putting *pure* (24 karat) gold leaf
on the heatsink or on the CPU plate (smaller area) because it would be
soft and malleable. Maybe under pressure with the heatsink clipped onto
the CPU the gold leaf would mold into the cavities and microscopic pits.
But then I realized that what looks perfectly flat to your eye would
still probably end up with microcopically huge air bubbles, and while
malleable the gold would probably not ooze into the pits completely.
The only way it might work would be if there was a means of sending a
high electric current just through the gold leaf to flash melt it (to
minimize any generated heat that would conduct back to the CPU's die),
but I doubt the CPU die is electrically isolated from its heat plate,
and somehow I don't think consumers would have the necessary equipment.
Someone would have to find out if this idea worked and start a service
where users could get their selected heatsink flash fused with gold to
their CPU (or buy them that way as a package).

That's a bit of a dream at this time so I've started to look into the
Peltier cooling systems. I don't want to go with liquid cooling (which
still has a fan somewhere or else you listen to pumps and gurgles, have
to deal with leaks, and route all the plumbing [I like my innards very
neat]). ThermalTake has their SubZero solution but I don't like having
to surrender a PCI slot.

 

[Guest]

"do_not_spam_me" said in


Silicone rubber to affix a heatsink? Geez, why not just
thermally isolate the heatsink with a thick rubber washer
and some contact cement?

I don't see how a rubber washer could affix a heatsink to a PC CPU,
and the only common high temperature rubber cement is disk brake
silencer (CRC makes it), which I don't like because it stains the
skin, and I don't know its thermal conductivity.

Why even bother affixing the heatsink at all?

Because even the worst heatsink and thermal adhesive will likely
improve the CPU package's thermal conductivity by a factor of 2-5.

It would be better to not use anything and just press the heatsink
and CPU together dry to get as much metal to metal contact as possible.

Unlikely, unless the surfaces match unusually well. Generally any
liquid will double the heat conductivity.

There's a reason no one includes silicon *rubber* in
their testing of thermal compounds in test reports, like
http://www.techwarelabs.com/reviews/cooling/thermal_paste/ and
http://www.thetechzone.com/reviews/roundup/thermal_compound/index.shtml.
Silicone rubber is a thermal *insulator*, not a conductor. Yeah, it
transfers heat 4 times faster than air but that still sucks!

Where do those links list silicone rubber? The only silicone-based
thermal conductants I found there were made of silicone grease, which,
as you know, isn't like silicone rubber. Where does either mention
that silicone rubber transfers heat 4 times faster than air?

The thermal conductivities of some materials, in W/cm^2/cm/C
("Switching Regulators and Power Supplies," p. 112.):

silver 4.19
copper 4.01
gold 3.39
aluminum 2.25
beryllium oxide 2.08 (best of any electrical insulator, worse at high
temp)
silicon .83
iron .71
stainless steel .16 (type not specified)
Pyrex glass .010
mica .0043
epoxy laminate .0030 (circuit board material)
silicone rubber .0026 (GE brand, type not specified)
air .00024

Water is roughly 1/3 better than copper, oil about half as good as
aluminum, and diamond is much better than anthing, about 20
W/cm^2/cm/C.

This shows silicone rubber being over 10 times as good as air. The
thickness of the material also has to be considered because silicone
rubber insulators tend to perform better than mica insulators, even
when the latter are coated with thermal grease and the former are run
bare (http://users.ece.gatech.edu/~mleach/ or his old Audio magazine
articles). Also because any layer of glue, grease, or tape is so
thin, it's thermal conductivity isn't as important as the heatsink's
thermal resistance to the surrounding air (probably almost 1 C/W for a
Socket 7 heatsink).

Thermal transfer:
air ~ 0.035 W/mK
silicone rubber = 0.14 to 0.22 W/mK
3M thermal tape = 0.4 W/mK
silicone/zinc grease or paste = 0.4 to 0.9 W/mK
copper filled silicone grease = 2.0 W/mK
alumina (Artic or generic) = 4.0 W/mK
silver filled silicone grease = 4.2 W/mK
Artic Silver epoxy = 7.5 W/mk
Artic Silver 3 = 9.0 W/mK
Artic Silver 5 = 350,000 W/mK (claimed)

I don't see how Artic can claim their Silver 5 compound, a
*paste*, has a thermal transfer rate that is 814 times greater
than solid silver itself. When I Googled around, no one
actually tested this claim and instead just quotes Artic's own
specs. By the way, little known fact (well, it was to me until
now): there was no Artic Silver 4 because "4" in Japanese is
"shi" which means "death" (see
http://www.techaddicts.net/ArcticSilver5/AS5_1.html).

I don't know Japanese, but I thought one meaning of "shi" was the
number 4, as in ichi, ni, san, shi (1, 2, 3, 4).

What is the minimum W/mK or W/cm^2/cm/C for you to consider a material
to be a thermal conductor rather than a thermal insulator? Also
pre-formed sheets of silicone rubber has long been used to help cool
semiconductors, including in almost every computer power supply, so I
doubt that it's bad.

I was suggesting a cheap but adequate solution for attaching a
heatsink to a 300 MHz K6-2, for which the type of thermal glue used
will hardly matter, unless it's way too thick, and it's silly to take
extraordinary measures for such a CPU. In this case, "extraordinary"
means more than $2.

 

[Noozer]

Because even the worst heatsink and thermal adhesive will likely

improve the CPU package's thermal conductivity by a factor of 2-5.

Thermal compound is used ONLY to fill the air space where there is NO
contact between the CPU and sink. If the compound is cause less metal to
metal contact then you've used too much.

Where do those links list silicone rubber? The only silicone-based
thermal conductants I found there were made of silicone grease, which,
as you know, isn't like silicone rubber. Where does either mention
that silicone rubber transfers heat 4 times faster than air?

It doesn't... and that's why he included the links. It's the same reason
that nobody uses peanut butter. Silicone rubber is a stupid choice for
thermal compound.

As for 4 times faster than air??? I'd only believe that if the air was not
moving at all. It's a lot easier to circulate air that silicone rubber!

 

[kony]

It doesn't... and that's why he included the links. It's the same reason
that nobody uses peanut butter. Silicone rubber is a stupid choice for
thermal compound.

As for 4 times faster than air??? I'd only believe that if the air was not
moving at all. It's a lot easier to circulate air that silicone rubber!

The interesting part of this discussion is that chips from K6 through P4
do use a silicone rubber interface between the core and head spreader. As
with any other interface the primary consideration is viscosity and
surface flatness since the vast majority of the interface material is
meant to squish out. We should consider it a given that any permanent,
cement-type interface requires the flattest surfaces (reasonably)
possible.

 

[~misfit~]

"~misfit~" said in news:[email protected]:


Zinc oxide is used as a filler in some low-grade thermal pastes. It
is not toxic. Zinc oxide is that water insoluable white stuff you see
painted on folks' noses to keep from getting sunburned.

"Zinc oxide is ideal for the nose and ears while swimming"
(http://www.healthcentral.com/mhc/top/001869.cfm).

I don't remember reading any warnings about not touching the stuff.
In fact, it usually needs to be reminded to users to NOT use their
fingers to spread around the paste because of the oil on their
fingers. However, I doubt any of the stuff is suitable for
consumption (but touching it is okay).

Maybe your friend was talking about a thermal adhesive because those
do use solvents that are toxic (but not in the dosage to which you
would be exposed for one heat sink, but maybe if you were on an
assembly line). I mean, hell, SuperGlue is toxic but I don't see
folks keeling over that accidentally got their fingers glued
together. Acetone is toxic yet I occasionally dampen a paper towel
with it to remove some glue, residue, or such, and my fingers are
touching that Acetone wetted paper towel. Grain alcohol is toxic, but
are you going to give up your beer? If so, I'll take it.

I'm thinking now it was beryllium oxide.

 

[~misfit~]

The interesting part of this discussion is that chips from K6 through
P4 do use a silicone rubber interface between the core and head
spreader. As with any other interface the primary consideration is
viscosity and surface flatness since the vast majority of the
interface material is meant to squish out. We should consider it a
given that any permanent, cement-type interface requires the flattest
surfaces (reasonably) possible.

I pulled a heat spreader off a Celeron Tualatin 1.3Ghz and while the
spreader was held on around the edges with silicone rubber, the compound on
the die itself was something else. Certainly not 'rubbery', more like
dried-out ordinary thermal compound.

 

[*Vanguard*]

"~misfit~" said in news:[email protected]:

I'm thinking now it was beryllium oxide.

While beryllium is toxic
(http://physchem.ox.ac.uk/MSDS/BE/beryllium.html), it looks to be so
when inhaled and not through the skin. It is probably toxic if
ingested, too. Hey, maybe we could send some to the idiot that sued
Sears because the lawn mower he bought from them didn't say it couldn't
be hoisted up and used as a hedge trimmer. As long as it doesn't say
that it couldn't be used for toothpaste, maybe we could get this idiot
listed in the next Darwin Awards (http://www.darwinawards.com/). Ah,
damn, looks like it's not toxic when ingested
(http://www.espimetals.com/msds's/berylliumoxide.pdf). That article
also says it is not toxic through the skin even on an abrasion. It is
toxic for chronic exposure, like if your job was as an assembler that
applies it everyday for voltage regulators or high-power transistors.

Keep trying. Maybe there's a paste with a cyano compound in it. ;-)

 

[VWWall]

"~misfit~" said in


While beryllium is toxic
(http://physchem.ox.ac.uk/MSDS/BE/beryllium.html), it looks to be so
when inhaled and not through the skin. It is probably toxic if
ingested, too. Hey, maybe we could send some to the idiot that sued
Sears because the lawn mower he bought from them didn't say it couldn't
be hoisted up and used as a hedge trimmer. As long as it doesn't say
that it couldn't be used for toothpaste, maybe we could get this idiot
listed in the next Darwin Awards (http://www.darwinawards.com/). Ah,
damn, looks like it's not toxic when ingested
(http://www.espimetals.com/msds's/berylliumoxide.pdf). That article
also says it is not toxic through the skin even on an abrasion. It is
toxic for chronic exposure, like if your job was as an assembler that
applies it everyday for voltage regulators or high-power transistors.

Keep trying. Maybe there's a paste with a cyano compound in it. ;-)

Aside from its not being toxic, I don't think finely divided beryllium
oxide in a silicone like carrier would make a good heat conducting
paste. The main advantage of solid beryllium oxide is that it's a good
electrical insulator while being a good heat transfer medium. It is
very difficult to machine, and the dust is toxic when inhaled. As has
been mentioned, diamond is also a good electrical insulator while having
excellent heat transfer properties. Unfortumately it has to be in
crystaline form and costs accordingly.

Virg Wall

 

[kony]

I pulled a heat spreader off a Celeron Tualatin 1.3Ghz and while the
spreader was held on around the edges with silicone rubber, the compound on
the die itself was something else. Certainly not 'rubbery', more like
dried-out ordinary thermal compound.

Hmmmm. Maybe I'm wrong... it happens. Can't remember now which chip I
last "thought" I saw that on.

 

[~misfit~]

"~misfit~" said in news:[email protected]:

While beryllium is toxic
(http://physchem.ox.ac.uk/MSDS/BE/beryllium.html), it looks to be so
when inhaled and not through the skin. It is probably toxic if
ingested, too. Hey, maybe we could send some to the idiot that sued
Sears because the lawn mower he bought from them didn't say it
couldn't be hoisted up and used as a hedge trimmer. As long as it
doesn't say that it couldn't be used for toothpaste, maybe we could
get this idiot listed in the next Darwin Awards
(http://www.darwinawards.com/). Ah, damn, looks like it's not toxic
when ingested (http://www.espimetals.com/msds's/berylliumoxide.pdf).
That article also says it is not toxic through the skin even on an
abrasion. It is toxic for chronic exposure, like if your job was as
an assembler that applies it everyday for voltage regulators or
high-power transistors.

Keep trying. Maybe there's a paste with a cyano compound in it. ;-)

I might have to go visit him and see if he remembers (He should I'd think)
although it was 20 years ago. It's bugging me now.

 

[ProfGene]

Most fans for that kind of computers have a clip that fits over two notches
on the holder for the CPU. These should be white plastic extensions and the
clip is a tinny piece of metal that slides up and down. You have to hook it
on one of the notches and push it down to hook it firmly on the other.