OT: Intel stock vs heatpipe cooler

[RxK]

Sorry for being a bit off-topic but, surely there's lots of builders in
here.
I've built lots of boxes in the past and never really dabbled with "super"
cpu coolers i.e. stock heatsink and fan always did a sufficient job.

A few months ago I upgraded my faithful old XP2600 Barton Skt. A to a D935
Conroe cpu, on a "Conroe" motherboard, and during lots of digging on the
web, at that time, noticed observations that the Pentium D's ran "hot"
compared to others. My "D" core always idled at approx. 38 degrees | 42
degrees under load. Yesterday I fitted a cheap Akasa triple heatpipe cooler
and noticed that the thermal "paste" was quite soft. When I removed the
stock Intel heatsink it appeared that the thermal compound was quite thick,
(obviously it had "cured"), but it hadn't spread out very much, suggesting
that it was perhaps too thick, or "aged," when I originally fitted the
thing.

I'm not going to try it just to satisfy my curiosity but, do you think that
if I cleaned that up, and refitted the stock h/s with a good quality thermal
compound, that it would run a lot cooler ?

regards, Richard

 

[VanguardLH]

Sorry for being a bit off-topic but, surely there's lots of builders in
here.
I've built lots of boxes in the past and never really dabbled with "super"
cpu coolers i.e. stock heatsink and fan always did a sufficient job.

A few months ago I upgraded my faithful old XP2600 Barton Skt. A to a D935
Conroe cpu, on a "Conroe" motherboard, and during lots of digging on the
web, at that time, noticed observations that the Pentium D's ran "hot"
compared to others. My "D" core always idled at approx. 38 degrees | 42
degrees under load. Yesterday I fitted a cheap Akasa triple heatpipe cooler
and noticed that the thermal "paste" was quite soft. When I removed the
stock Intel heatsink it appeared that the thermal compound was quite thick,
(obviously it had "cured"), but it hadn't spread out very much, suggesting
that it was perhaps too thick, or "aged," when I originally fitted the
thing.

I'm not going to try it just to satisfy my curiosity but, do you think that
if I cleaned that up, and refitted the stock h/s with a good quality thermal
compound, that it would run a lot cooler ?

regards, Richard

What do you consider "a lot"? Is 4C a lot to you? It isn't to me but
then I am not overclocking and looking to squeeze out every Celsius
degree in reduction possible because I'm pushing the temperatures up
towards their maximums.

Most home jobbers gob on way too much thermal paste. It should be
transparent. The idea is to fill in the gaps, not insulate the heatsink
from the die cover plate on the CPU. Air is bad for heat transfer.
Metal-to-metal is best. Thermal paste is somewhere in between: not as
bad as metal-to-air-to-metal but definitely not as good as
metal-to-metal. You want to fill in the microscopic pores which contain
air, not try to use the thermal paste for where metal-to-metal contact
would've been made. After applying the thinly transparent layer of
thermal paste, twist the heatsink atop the CPU to make sure you spread
it out evenly before clamping it down. If you remove a heatsink and see
a ridge of extra goop around an indent for the where the CPU cover was
pressed, you used too much. You should just make out an outline of
where the cover pressed against the heatsink. If you have a warped
heatsink that doesn't mate well with the CPU cover, you'll have to lap
the heatsink.

Some pastes are deliberately made thicker to prevent home jobbers from
gooping on too much, squeezing it out when clamping down on the
heatsink, and having the excess squeeze out onto the CPU and dripping
elsewhere to contaminate those other parts. In fact, you have to be
careful with the silver metal pastes since they are conductive. Some go
through a transition (phase change) when heated that changes their
viscosity (which is no longer important after the surfaces have been
mated since the paste won't get where there is metal-to-metal contact).
Thermal pads are like that, too: they are in an amorphously fluid state
that go through a phase change under pressure at elevated temperature to
make them less viscous and allow reflow into the gaps. Phase-change
here means the material changes viscosity, not that a liquid changes to
a gas as is used in some heatsinks that use heatpipes. See
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/26951.pdf.

Different pastes have different transfer rates. The differences can
amount to a few Celsius in temperature change, like maybe 4C to 6C. The
silver paste might get you 2C more than the ceramic paste which gets you
maybe 2C more than good general paste which gets you maybe 1C to 1C
better than a thermal pad. Lapping the heatsink might get you another
1C to 2C. Some heatsinks come with a rather rough surface so lapping
helps. Few come with a highly polished (minimal pores) and nearly flat
base. If they are warped (where you see the paste is thicker in some
spots than others after removing the heatsink and inspecting the pasted
area), you need to lap the heatsink. You want as much metal-to-metal
contact as possible rather than relying on paste to replace air.

 

[RxK]

Thanks for your response. I was aware of much of what you say, ...and
completely agree with what I was't aware of !
....btw, I am an ardent ANTI-overclocker - much of what is in the discussion
forums i.e. people deluding themselves that they've achieved something
marvellous by overclocking, comments such as "...and very stable..." etc.
<ROFL>

I prefer the manufacturers specs., ...and was querying, as I see it,
unsatisfactory Intel stock heatsink compound behaviour, in relation to
maximising the lifespan of the cpu. Specifically, I suspect it was too
thick, or not "fluid enough" for the excess to squeeze out, as normally
happens. I've always applied the stuff thinly, and the excess that squeezes
out is not excessive, if you see what I mean.

anyhooo, to clarify, the stock thermal paste that was on the Intel stock hs,
and cpu "lid", had obviously cured to some degree and had not "spread"
enough to allow any metal to metal contact at all, as far as I could see,
which one would expect to happen. ...and I suggest it's not sensible to
remove the thing to check that it spread out satisfactorily - sort of
defeats the object of a satisfactory fitting !

I should have mentioned in my OP that bios / and Speed Fan reported the core
temperature, at idle, as 38 deg.C ...sometimes up to 45 deg. C under load.
Since replacing the Intel stock hs with an Aksa heatpipe cooler, it's 26
deg. C at startup, 30 deg. C atm - mostly idling.

regards, Richard

 

[Andy]

Sorry for being a bit off-topic but, surely there's lots of builders in
here.
I've built lots of boxes in the past and never really dabbled with "super"
cpu coolers i.e. stock heatsink and fan always did a sufficient job.

A few months ago I upgraded my faithful old XP2600 Barton Skt. A to a D935
Conroe cpu, on a "Conroe" motherboard, and during lots of digging on the
web, at that time, noticed observations that the Pentium D's ran "hot"
compared to others. My "D" core always idled at approx. 38 degrees | 42

According to
<http://global.msi.com.tw/index.php?...o=1342&maincat_no=1&cat2_no=170&cat3_no=#menu>
and
<http://en.wikipedia.org/wiki/List_of_Intel_Pentium_D_microprocessors>
the D 935 has a TDP of 95 watts compared to 65 watts for the Core2
Duo.

 

[M.I.5¾]

Sorry for being a bit off-topic but, surely there's lots of builders in
here.
I've built lots of boxes in the past and never really dabbled with "super"
cpu coolers i.e. stock heatsink and fan always did a sufficient job.

A few months ago I upgraded my faithful old XP2600 Barton Skt. A to a D935
Conroe cpu, on a "Conroe" motherboard, and during lots of digging on the
web, at that time, noticed observations that the Pentium D's ran "hot"
compared to others. My "D" core always idled at approx. 38 degrees | 42
degrees under load. Yesterday I fitted a cheap Akasa triple heatpipe
cooler and noticed that the thermal "paste" was quite soft. When I
removed the stock Intel heatsink it appeared that the thermal compound was
quite thick, (obviously it had "cured"), but it hadn't spread out very
much, suggesting that it was perhaps too thick, or "aged," when I
originally fitted the thing.

I'm not going to try it just to satisfy my curiosity but, do you think
that if I cleaned that up, and refitted the stock h/s with a good quality
thermal compound, that it would run a lot cooler ?

There are various types of thermal compound around these days. The most
common type looks a bit like thick white or silver grease and will always
appear to be soft even after many hours of operation. There are cheap
variations on this type that dry out with use and are not to be recommended.
There is also a newer type known as phase change thermal compound. This
comes in various forms including pre applied to a heatsink. It is quite
firm to the touch, but the idea is that as it warms up it changes phase and
flows into all the gaps between the processor and heatsink/pipe. This
latter type is gererally considered the best type and some processor
suppliers won't honour a warranty if any other type is used.

It should be remembered that thermal compound of any type is not actually
particularly thermally conductive and should not be liberally plastered over
the processor or the heatsink. The idea is to apply just enough to fill the
tiny air gap that would otherwise exist between the processor and the
heatsink as the compound conducts heat better than air. I bough a 50ml tube
several years ago and have barely used a fifth of it despite overhauling
many power amplifier circuits.

 

[RxK]

Hello thanks for your repsonse.

You covered what I've been postulating about !

A couple of days ago I replaced the Intel stock heatsink and fan - fitted
with cpu at same time approx. 6 months ago, and noticed that its'
pre-applied thermal paste, had evidently not "spread out," very much at
all. It had sort of remained quite thick, and was covering just the centre
third area of the "cpu lid," resulting in the compound forming almost a
complete "barrier" between cpu and heatsink - instead of, as is desired, as
much metal to metal contact as possible, and any uneven-ness between those
surfaces, assisted by a very thin and only the required of compound, if you
see what I mean.

Anyhow, after cleaning that off, and fitting the Akasa heatpipe cooler, its'
pre-applied paste was obvously much softer and upon fitting it, the initial
excess quickly "squeezed out," ...this was not excessive but, was obviously
a thermal paste with a superior and more carefully designed "viscosity,"
than the stuff that was on that stock Intel heatsink.
....and at idle, core temp. is 31 deg. C with Akasa, compared to the 38 deg.
C. it had always settled at with Intel stock h/s, when idling.
When making the cpu do lots of work, now with Akasa on it, core temperature
stays significantly below the 48 deg. Cent. is used to go up to when the
Intel h/s was on it.

....leaving me with the suspicion that all I had needed to do, (if I had been
aware that the stock Intel h/s was not providing the level of cooling
performance it could have been providing, if had been fitted with a half
decent thermal compound), was to replace the thermal compound, instead of
replacing the whole h/s / fan assembly !!

regards, Richard