Heat Sink Compound Necessary?

[Talal Itani]

I want to replace the Intel stock fan on my 478 Pentium 4, because I need to
overclock heavily. Is heat sink compound needed, recommended? Thank you.

Talal Itani

 

[Paul]

I want to replace the Intel stock fan on my 478 Pentium 4, because I need to
overclock heavily. Is heat sink compound needed, recommended? Thank you.

Talal Itani

You need a good thermal path from the top of the CPU, into the heatsink.
So, yes, thermal compound or a thermal pad should be used. The
thermal compound displaces air, and air is a good insulator.
Only a thin layer is required, and you don't want to use so much,
that major quantities squirt out the side of the joint.

Paul

 

[DaveW]

You ABSOLUTELY have to use thermal paste or compound, or your CPU will
overheat and fry.

 

[Skeleton Man]

You ABSOLUTELY have to use thermal paste or compound, or your CPU will

overheat and fry.

I think that's exaggerating it a little.. their CPU would certainly run
hotter than it should, but to the point of destruction ? I have my doubts..

Chris

 

[kony]

I think that's exaggerating it a little.. their CPU would certainly run
hotter than it should, but to the point of destruction ? I have my doubts..

Chris

It's not absolutely necessary to use thermal compound at all
either, (three?) years ago I reported on a test where I took
a ridiculous amount of time to lap a heatsink *VERY* flat
and had practically identical temps even on a flipchip
Athlon XP. Today's heat spreaders aren't as flat as the
flipchip cores bare, were, but if they were lapped very well
the result should be equally good. However, the main point
might be what was written above, "a ridiculous amount of
time", when otherwise all one has to do is spend a few
seconds applying thermal compound to achieve the same
results. Plus, you can't lap a CPU w/o voiding the
warranty.

 

[Plato]

You need a good thermal path from the top of the CPU, into the heatsink.
So, yes, thermal compound or a thermal pad should be used. The
thermal compound displaces air, and air is a good insulator.
Only a thin layer is required, and you don't want to use so much,
that major quantities squirt out the side of the joint.

Agreed...

 

[larry moe 'n curly]

You ABSOLUTELY have to use thermal paste or compound, or your CPU will
overheat and fry.

How does a Pentium 4, which has built-in thermal protection, fry,
unless you apply a torch to it?

 

[Skeleton Man]

It's not absolutely necessary to use thermal compound at all

either, (three?) years ago I reported on a test where I took
a ridiculous amount of time to lap a heatsink *VERY* flat
and had practically identical temps even on a flipchip
Athlon XP. Today's heat spreaders aren't as flat as the
flipchip cores bare, were, but if they were lapped very well
the result should be equally good.

Interesting, I've head of lapping the heatsink, but not the cpu. Guess it
makes sense to have metal to metal contact rather than metal to plastic
coating.

What is the coating on them anyway ? some kind of thermal compound or just a
protective plastic ?

However, the main point
might be what was written above, "a ridiculous amount of
time", when otherwise all one has to do is spend a few
seconds applying thermal compound to achieve the same
results. Plus, you can't lap a CPU w/o voiding the
warranty.

It would be an interesting exercise.. I doubt you could do much (if any)
harm to the heatspreader.

btw, when a temperature reading talks about cpu case, does that mean the
temperature of the heatspreader ?

Chris

 

[kony]

Interesting, I've head of lapping the heatsink, but not the cpu. Guess it
makes sense to have metal to metal contact rather than metal to plastic
coating.

There's minimal lapping necessary on the flipchips, only a
slight raised area from the laser product etching on earlier
athlon XP / Palomino, none on the latter XPs unless there
were some surface irregularity. As for lapping CPUs with
heat spreaders, there were people doing it back in the
PII/Celeron era too as the Celerons in particular had a heat
spreader with the outer border raised a bit higher than the
middle.

What is the coating on them anyway ? some kind of thermal compound or just a
protective plastic ?

I don't know what the coating on them is, but immediately
below that I though it was just a silicon sheet.

It would be an interesting exercise.. I doubt you could do much (if any)
harm to the heatspreader.

btw, when a temperature reading talks about cpu case, does that mean the
temperature of the heatspreader ?

It might depend on the context, normally I would think not
but that would also depend on the sink interface, IMO die
temp is a much better number to use if it's reasonably
measurable (motherboard does it).

 

[kony]

How does a Pentium 4, which has built-in thermal protection, fry,
unless you apply a torch to it?

It's doubtful it would fry w/o thermal compound, but without
any heatsink it can as the die temp rate of increase on some
areas can be higher than on others, it's conducted from the
area generating heat to the rest of the die, so by the time
the thermal shutdown circuit were to respond some areas
could already be hot enough to be damaged.

 

[UCLAN]

There's minimal lapping necessary on the flipchips, only a
slight raised area from the laser product etching on earlier
athlon XP / Palomino, none on the latter XPs unless there
were some surface irregularity. As for lapping CPUs with
heat spreaders, there were people doing it back in the
PII/Celeron era too as the Celerons in particular had a heat
spreader with the outer border raised a bit higher than the
middle.

One other consideration other than surface flatness is the
existence of small, microscopic gaps in the two surfaces.
Individually, these gaps do little to affect heat transfer
between the two surfaces. Collectively however, they can have
an affect. Heat transfer compound fills these gaps and eliminates
their effect.

 

[larry moe 'n curly]

It's doubtful it would fry w/o thermal compound, but without
any heatsink it can as the die temp rate of increase on some
areas can be higher than on others, it's conducted from the
area generating heat to the rest of the die, so by the time
the thermal shutdown circuit were to respond some areas
could already be hot enough to be damaged.

Do you remember this test done by Tom's Hardware back in 2001, where
they yanked the heatsinks off running Athlons and Pentiums?

www.tomshardware.com/2001/09/17/hot_spot/

Both the Pentium III and 4 survived intact, but both Athlons fried.
AMD modified a mobo to prevent this, and Asus/Asrock later claimed
that their Socket A mobos would protect the CPU against any such
failure (but I haven't verified that witn my Asrock).

The only reasons I bought a Pentium 4 instead instead of a faster AMD
CPU was because the latter could burn up and its heatsink mounting
wasn't that solid.

 

[Paul]

Do you remember this test done by Tom's Hardware back in 2001, where
they yanked the heatsinks off running Athlons and Pentiums?

www.tomshardware.com/2001/09/17/hot_spot/

Both the Pentium III and 4 survived intact, but both Athlons fried.
AMD modified a mobo to prevent this, and Asus/Asrock later claimed
that their Socket A mobos would protect the CPU against any such
failure (but I haven't verified that witn my Asrock).

The only reasons I bought a Pentium 4 instead instead of a faster AMD
CPU was because the latter could burn up and its heatsink mounting
wasn't that solid.

The Pentium 4 has THERMTRIP. The Athlon64 has THERMTRIP. The AthlonXP
doesn't, but it does have a thermal diode. My S462 motherboard has
an 8 pin chip, which connects to the thermal diode, and provides the
THERMTRIP function. I think it trips at 85C or 90C or so. Older S462
boards were less protected than that (software mediated shutdown). And
it is possible that older Athlon processors didn't have a thermal
diode - I've have to go digging through the datasheets to see whether
the function was there or not.

I downloaded a late model 0.13u Pentium III datasheet, and it has THERMTRIP
too.

So I guess Tomshardware has run out of processors to burn up.

Paul

 

[Noozer]

At this point everyone should have realized that compound was NOT necessary.
If Dave says it, you know the opposite is true.

Now, it would be stupid to run your CPU without paste, but as long as the
heatsink wasn't improperly mounted you should be OK.

 

[kony]

Do you remember this test done by Tom's Hardware back in 2001, where
they yanked the heatsinks off running Athlons and Pentiums?

www.tomshardware.com/2001/09/17/hot_spot/

Both the Pentium III and 4 survived intact, but both Athlons fried.
AMD modified a mobo to prevent this, and Asus/Asrock later claimed
that their Socket A mobos would protect the CPU against any such
failure (but I haven't verified that witn my Asrock).

The only reasons I bought a Pentium 4 instead instead of a faster AMD
CPU was because the latter could burn up and its heatsink mounting
wasn't that solid.

There were also people who started up P4 systems without a
heatsink installed, to find the head spreader pops off and
the chip still dies. Key in the P4 protection is the rate
of temp increase, it will do it's job as intended in a
typical situation where there was dust buildup, a fan
failure, etc, but it's not a reliable way to save a CPU in
the same test Tom's ran.

 

[Plato]

How does a Pentium 4, which has built-in thermal protection, fry,
unless you apply a torch to it?

What if the entire pc was filled with dust from a textile factory?