K
Kev
Is thermal paste really needed for a P3/600/2.05v cpu? Or can I just
go with the heatsink on top?
go with the heatsink on top?
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
J
jeffc
Kev said:Is thermal paste really needed for a P3/600/2.05v cpu? Or can I just
go with the heatsink on top?
As opposed to the heatsink on the bottom? Or no heatsink if using thermal
paste? No. Use thermal paste (or pad) and a heatsink. Period.
W
w_tom
If you are the classic computer builder who really never
learned basic principles, then I too would tell you to use
thermal compound. That is what Intel, AMD, etc say so that
even the naive will be less likely to cause problems. But
let's assume you want to understand the underlying principles
- which is the primary reason for building a computer.
Thermal resistance increases (thermal conductivity is worse)
with each change of medium. The best thermal conductor is CPU
direct to heatsink - 2 medium. This explains why thermal pads
and thermal tape are less conductive solutions. A properly
machined heatsink mounted bare and directly on CPU is more
than sufficient - especially for your lower power CPU.
CPU temperature can be reduced by single digit degrees if
applying the most minimal amount of thermal compound. IOW so
little thermal compound as to only fill microscopic holes in
the heatsink to CPU interface.
Now to measure quality of that heatsink. If thermal
compound causes more than single digit temperature reductions,
then either heatsink is not properly machined or human did not
properly install heatsink. Note, the properly machined
heatsink is not perfectly flat. It may be tapered so that
pressure applies mostly in that little center section where
virtually all heat is transferred from CPU to heatsink.
1) Best thermal conductivity is the direct CPU to heatsink
connection. This interface achieved either by no thermal
compound or by the most minimal amount of thermal compound.
No thermal compound is quite acceptable for any Pentium III
CPU.
2) Too much thermal compound causes increased thermal
resistance because interface is three medium - CPU to thermal
compound to heatsink. IOW if using thermal compound, then
compound should never spread to the outside half of mating CPU
/ heatsink surface. If thermal compound squeezes to outer
half, then way too much thermal compound was applied.
3) Above test, using thermal compound, verifies the quality
of that heatsink interface. Many heatsinks are not even
machined to reduce costs. They assume you will fix the
problem with thermal compound, thermal pad, etc. If thermal
compound results in more single digit temperature reduction,
then suspect a bad heatsink. Furthermore, if heatsink does
not loudly claim a 'degree C per watt' number, then assume
worst from that heatsink manufacturer. Inferior heatsinks
fear to provide that all so important specification -
especially when they don't even bother to machine the
interface surface.
4) Difference between thermal compounds is trivial. To
maximize profits, some thermal compounds have the naive
promoting their Arctic Silver product line. Why pay up to ten
times more for a product that is only equivalent? Its called
hype or junk science typically promoted without supporting
numbers.
5) Better designers don't even use a heatsink fan assembly
for that CPU. Such 'system' designs use basic principles such
as airflow and the most important parameter for all heatsink
assemblies - 'degree C per watt'.
6) Do the temperature test with and without thermal
compound. You are building a system because you want to
learn. That test should be very informative.
learned basic principles, then I too would tell you to use
thermal compound. That is what Intel, AMD, etc say so that
even the naive will be less likely to cause problems. But
let's assume you want to understand the underlying principles
- which is the primary reason for building a computer.
Thermal resistance increases (thermal conductivity is worse)
with each change of medium. The best thermal conductor is CPU
direct to heatsink - 2 medium. This explains why thermal pads
and thermal tape are less conductive solutions. A properly
machined heatsink mounted bare and directly on CPU is more
than sufficient - especially for your lower power CPU.
CPU temperature can be reduced by single digit degrees if
applying the most minimal amount of thermal compound. IOW so
little thermal compound as to only fill microscopic holes in
the heatsink to CPU interface.
Now to measure quality of that heatsink. If thermal
compound causes more than single digit temperature reductions,
then either heatsink is not properly machined or human did not
properly install heatsink. Note, the properly machined
heatsink is not perfectly flat. It may be tapered so that
pressure applies mostly in that little center section where
virtually all heat is transferred from CPU to heatsink.
1) Best thermal conductivity is the direct CPU to heatsink
connection. This interface achieved either by no thermal
compound or by the most minimal amount of thermal compound.
No thermal compound is quite acceptable for any Pentium III
CPU.
2) Too much thermal compound causes increased thermal
resistance because interface is three medium - CPU to thermal
compound to heatsink. IOW if using thermal compound, then
compound should never spread to the outside half of mating CPU
/ heatsink surface. If thermal compound squeezes to outer
half, then way too much thermal compound was applied.
3) Above test, using thermal compound, verifies the quality
of that heatsink interface. Many heatsinks are not even
machined to reduce costs. They assume you will fix the
problem with thermal compound, thermal pad, etc. If thermal
compound results in more single digit temperature reduction,
then suspect a bad heatsink. Furthermore, if heatsink does
not loudly claim a 'degree C per watt' number, then assume
worst from that heatsink manufacturer. Inferior heatsinks
fear to provide that all so important specification -
especially when they don't even bother to machine the
interface surface.
4) Difference between thermal compounds is trivial. To
maximize profits, some thermal compounds have the naive
promoting their Arctic Silver product line. Why pay up to ten
times more for a product that is only equivalent? Its called
hype or junk science typically promoted without supporting
numbers.
5) Better designers don't even use a heatsink fan assembly
for that CPU. Such 'system' designs use basic principles such
as airflow and the most important parameter for all heatsink
assemblies - 'degree C per watt'.
6) Do the temperature test with and without thermal
compound. You are building a system because you want to
learn. That test should be very informative.
A
Adam S
Thermal resistance increases (thermal conductivity is worse)
If the surface of the CPU die and the heatsink should be flat and next to
each other, and presumably the CPU makers know this, why do they print the
CPU identification on the surface? The ink _must_ have a thickness to it and
this will prevent a _totally_ flat surface. I always use a very thin layer
of thermal paste to compensate for the manufacturers messing up my flat
surface with ink.
Adam S
with each change of medium. The best thermal conductor is CPU
direct to heatsink - 2 medium. This explains why thermal pads
and thermal tape are less conductive solutions. A properly
machined heatsink mounted bare and directly on CPU is more
than sufficient - especially for your lower power CPU.
CPU temperature can be reduced by single digit degrees if
applying the most minimal amount of thermal compound. IOW so
little thermal compound as to only fill microscopic holes in
the heatsink to CPU interface.
If the surface of the CPU die and the heatsink should be flat and next to
each other, and presumably the CPU makers know this, why do they print the
CPU identification on the surface? The ink _must_ have a thickness to it and
this will prevent a _totally_ flat surface. I always use a very thin layer
of thermal paste to compensate for the manufacturers messing up my flat
surface with ink.
Adam S
S
somebody
If you are the classic computer builder who really never
learned basic principles, then I too would tell you to use
thermal compound. That is what Intel, AMD, etc say so that
even the naive will be less likely to cause problems. But
let's assume you want to understand the underlying principles
- which is the primary reason for building a computer.
Thermal resistance increases (thermal conductivity is worse)
with each change of medium. The best thermal conductor is CPU
direct to heatsink - 2 medium.
Only way you'll ever achieve this, is by copper soldering the
copperbase directly to the cpu. (This will destroy the cpu BTW). In
all other cases, you have a thin third medium involved. Only
difference a *properly machined* , that is very smooth and flat,
heatsink will make, is to make this third medium layer thinner. And
thinner is certainly better. But you will never achieve 2 medium. The
two 'contact' surfaces only touches each other on a number of points,
that will increase in number with increased pressure, but which for
nondestructive pressures in our case, will never amount to more than a
very, very small part of the total area. Its always 3 mediums. The
purpose of paste is to replace air as the third medium.
This explains why thermal pads
and thermal tape are less conductive solutions.
A properly
machined heatsink mounted bare and directly on CPU is more
than sufficient - especially for your lower power CPU.
I'm aware that Intel is the subject matter here, but I'm looking out
for the general value of information.
Thermal pads solve a couple of problems, for particularly AMD socket A
cpus. These cpus do not have any integrated heatspreader, like P4 or
Athlon64. This means they depend upon the base of the heat sink to act
as heat spreader and soak up heat from almost microscopic hotspots.
There must be good continuous contact all over, with no air pockets.
Mounting a heatsink without pad or paste on a socket A cpu, is just
stupid, - and you won't see the real problem on any temperature
reading either. And that is to some extent true about P4s and A64s
too. Without paste, you might have a bigger problem than mere
temperature readings might suggest. Also remember, P4s don't really
get hotter, they get slower...
And frankly, I don't really see the point in mounting a heatsink on
_any_ cpu, without paste or pad. Why is paste or pad so inconvenient?
What do we hope to gain by not using it?
Yes, the pads are thicker and slightly less efficient heat transfer.
But they eliminate those air pockets and shortages from major
spillovers outside the chip face, that both might result from inept
application of paste. So it's a more reliable solution. And reliable
is good.
CPU temperature can be reduced by single digit degrees if
applying the most minimal amount of thermal compound. IOW so
little thermal compound as to only fill microscopic holes in
the heatsink to CPU interface.
Now to measure quality of that heatsink. If thermal
compound causes more than single digit temperature reductions,
then either heatsink is not properly machined or human did not
properly install heatsink. Note, the properly machined
heatsink is not perfectly flat. It may be tapered so that
pressure applies mostly in that little center section where
virtually all heat is transferred from CPU to heatsink.
1) Best thermal conductivity is the direct CPU to heatsink
connection. This interface achieved either by no thermal
compound or by the most minimal amount of thermal compound.
Well, as I've already hinted, I would put it in different words. The
interface we're looking for, is the thinnest possible 3'd medium
layer. But it's better that layer is air-free, than as thin as
possible. (Thus pads). The fact that even with very smooth surfaces,
without paste, temperatures rises by 7-8 deg C, is a strong testimony
to both that we indeed always have 3 mediums, not 2, and to the
immense insulating properties of air. To overcome even that
infinitesimal thin layer air, it still takes that considerable
increase in temperature differential to drive the same amount of
energy across the gap.
Powerful argument indeed to use paste. ...or pad.
ancra