Changing the thermal compound on a Prescott 650 stock Intel Heatsink?

[Brian]

Temperatures are still hovering with a load in the mid- 50'sC in a ~74F
room after adding an air duct and must find a way to lower them even
further when summer arrives. Are there any other tips, warnings, or
precautions besides running the computer for ten minutes prior to
uninstalling the heatsink?

And how often must the Arctic Silver compound be reapplied before it
dries out?

 

[badgolferman]

And how often must the Arctic Silver compound be reapplied before it
dries out?

Where did you hear this? I have never had to reapply thermal compound
because it dried out. Perhaps you are not applying it correctly or
your heat sink/fan combo are inadequate. Blowing all the dirt out of
the heat sink fins does wonders for cooling.

 

[Paul]

Temperatures are still hovering with a load in the mid- 50'sC in a ~74F
room after adding an air duct and must find a way to lower them even
further when summer arrives. Are there any other tips, warnings, or
precautions besides running the computer for ten minutes prior to
uninstalling the heatsink?

And how often must the Arctic Silver compound be reapplied before it
dries out?

That is not the worst processor to have to cool. It is only 84W.
http://processorfinder.intel.com/scripts/details.asp?sSpec=SL7Z7

In terms of the ultimate air cooling solution, it would be
an inlet duct leading to a wind tunnel heatsink (with heatpipes
to spread the heat better into the soldered fins) following by
an exhaust duct leading outside the case. Thus, the processor
uses outside 25C room air, instead of a mixture of 25C room
air and 40C stagnant case air. A good question for such a design,
would be where to stick the fan - on inlet or exhaust ? The
reason designs like that don't exist, is the difficulty of
adapting to every motherboard - case combo out there. And the
rest of the computer components still heed an air flow, so
the conventional fan on the back is still required.

There are some small computer cases (Shuttle sized), that use
a technique like that. The air flow path for the processor
is tightly controlled, and processor exhaust doesn't get a
chance to heat up the rest of the computer. Air flows from one
side of the case to the other, with a duct guiding the air all
the way.

If you insist on better thermal performance than can be achieved
with a Big Typhoon or a XP-120, there is always water cooling,
or if you are the devil-may-care kind of person, a Peltier+water
block. While a Peltier is a highly inefficient device, it
could shave a few more degrees off the temp the processor
is seeing. Of course, a big Peltier probably could use its own
ATX power supply, and the pump on the water cooling system might
appreciate a private power source as well. (There are computer
cases with room for two power supplies, so it can be done.)

This Peltier+water_block is 12V@18A and really could use its
own power supply. You can see in the plots of coolant temp
versus processor temperature, that you can achieve a processor
temperature of 20C, when the water flowing past is still at
42C. Now, at 216W of power for the Pelt, from a 70% efficient
power supply, where I live it would cost $0.42 a day to power
it, or $154 a year (and my power is cheap).

http://www.swiftnets.com/products/mcw5002-775T.asp

As for the Arctic Silver, try the following:

1) Install AS, accordng to the instructions on the AS web site.
2) Run the computer for several days. This gives the particles
in the AS a chance to "set" in their final position, and the
temps will fall a bit over the period of those several days.
3) On the third day, measure case air temperature and measure
CPU temperature. The difference is delta_T . Say you measure
55C under load on the CPU, and the case air temperature is
35C, then the delta is 20C.
4) A year from now, repeat the experiment in step 3. Use the
same program and same conditions, to apply a load to the
processor as you did in step 3. (Something like Prime95
will suffice). Now, depending on whether it is "summer"
or "winter" in your room, you are covered, as you are
measuring the delta_T and not the absolute temperature.
If it is a summers day, and CPU is 60C and case air temp
is 40C, the delta_T is still the same 20C.

You don't have to change the AS, until the delta degrades by
5C-7C or more. If the original installation managed a delta_T of
20C, and you measure it today and it is 27C, then change the
AS. In other words, let the performance of the cooling
solution, guide you in when to do maintenance.

HTH,
Paul

 

[kony]

Temperatures are still hovering with a load in the mid- 50'sC in a ~74F
room after adding an air duct and must find a way to lower them even
further when summer arrives.

How much further? It'd be good to know if you're expecting
a need to shed 20C more or just 5C, etc... at stock speed it
should be stable at least through the 60-70C range, but is
"with a load", a full load?

Are there any other tips, warnings, or
precautions besides running the computer for ten minutes prior to
uninstalling the heatsink?

Well unplugging the fan the last 30 seconds might help a bit
more, but that's about all, just remove it while it's still
warm.

An alternate approach is to just take a hairdryer to the
sink, where the fan was to heat it up.

Excess thermal material will wipe off most easily with
petroleum solvent.

And how often must the Arctic Silver compound be reapplied before it
dries out?

Arctic Silver and any other entirely synthetic based
compounds should not dry out, that was generally the most
common problem with silicone based compounds on small open
(flipchip) cores without a heat spreader.

Just put a tiny bit of AS in the middle of the core, don't
frost a cake when applying it since even a tiny bit will
spread out once the 'sink pressure is on it. It may take a
few days and thermal cycles to reach the stablized minimum
temp, dropping by single-digits until then.

 

[w_tom]

Thermal compounds are so cheap and don't dry out. You would
never know this from those who blindly hype Arctic Silver
(that is grossly overpriced). Since thermal solutions are so
inexpensive, you can bet Intel has already installed the best
solution.

CPU temperature in 50sC in a 74 degree room means CPU in the
70sC in a 100 degree F room. Now, what does an Intel spec
sheet say for maximum temperature? Not what you 'feel' is max
temperature. As long as CPU in a 100 degree room remains
below that max temperture, then CPU is in pig's heaven -
perfect and happy.

So first - what are the numbers? What does the Intel data
sheet demand? Will that CPU exceed Intel specs when room
temperature rises above 100 degree F? If not, then we have
Don Quixote chasing wind mills or hype promoted by Arctic
Silver. First get the data sheet numbers. And ignore hype
promoted by Arctic Silver advocates.

 

[Paul]

Thermal compounds are so cheap and don't dry out. You would
never know this from those who blindly hype Arctic Silver
(that is grossly overpriced). Since thermal solutions are so
inexpensive, you can bet Intel has already installed the best
solution.

CPU temperature in 50sC in a 74 degree room means CPU in the
70sC in a 100 degree F room. Now, what does an Intel spec
sheet say for maximum temperature? Not what you 'feel' is max
temperature. As long as CPU in a 100 degree room remains
below that max temperture, then CPU is in pig's heaven -
perfect and happy.

So first - what are the numbers? What does the Intel data
sheet demand? Will that CPU exceed Intel specs when room
temperature rises above 100 degree F? If not, then we have
Don Quixote chasing wind mills or hype promoted by Arctic
Silver. First get the data sheet numbers. And ignore hype
promoted by Arctic Silver advocates.

What the OP is trying to avoid, is the throttling temperature,
not the absolute max. 70C is the temp at which the processor
slows down internally, and that is why people try to improve
the cooling of recent Intel processors. The processor will
shut the computer off, if it actually gets too hot. Shutdown
on the latest processors is at (throttle+20C). And, as this
article explains, the throttle temperature is not even a
constant, and that is why the recommendation is to stay
below the throttle temperature stated in the Intel datasheet.

http://www.xbitlabs.com/news/cpu/display/20040224072411.html

There is a program called Throttlewatch, that a user can
use, to determine if the processor is hot enough to
throttle down internally. That is explained in this article.

http://www.xbitlabs.com/articles/cpu/display/oc-guide_7.html

Throttlewatch can be downloaded here.

http://www.panopsys.com/Downloads.html

Paul

 

[kony]

Thermal compounds are so cheap and don't dry out. You would
never know this from those who blindly hype Arctic Silver
(that is grossly overpriced). Since thermal solutions are so
inexpensive, you can bet Intel has already installed the best
solution.

I agree that Arctic Silver is overpriced but silicone based
compounds can be a problem, especially on flip-chips without
a head spreader. Below is a picture of an Athlon XP2000
which ran fine for (I forget the period at the moment, think
it was about 12-16 months) then system began crashing. Most
difficult to diagnose was that it seemed an island of dried
compound remained over the area of the core with the thermal
diode in it as the overall temps were not excessively high.

After removing the compound and applying fresh compound the
crashing was gone, even with a few degrees higher reported
temp.

http://69.36.189.159/usr_1034/burnt_athlon.jpg

 

[Brian]

That is not the worst processor to have to cool. It is only 84W.
http://processorfinder.intel.com/scripts/details.asp?sSpec=SL7Z7

In terms of the ultimate air cooling solution, it would be
an inlet duct leading to a wind tunnel heatsink (with heatpipes
to spread the heat better into the soldered fins) following by
an exhaust duct leading outside the case. Thus, the processor
uses outside 25C room air, instead of a mixture of 25C room
air and 40C stagnant case air. A good question for such a design,
would be where to stick the fan - on inlet or exhaust ? The
reason designs like that don't exist, is the difficulty of
adapting to every motherboard - case combo out there. And the
rest of the computer components still heed an air flow, so
the conventional fan on the back is still required.

Have currently replaced the side panel with a piece of cardboard and
managed to add an air duct directed at the heatsink. This is just a
temporary solution till I can figure out the exact placement of any
additional fans if they are needed. A friend installed the processor
and heatsink and I don't know if he used the furnished pad or applied
Arctic Silver. So my next step will be removing the heatsink and
lopping the area that makes contact if required. The exhaust duct that
you mentioned sounds good but I don't know if it could be utilized in
my case.

There are some small computer cases (Shuttle sized), that use
a technique like that. The air flow path for the processor
is tightly controlled, and processor exhaust doesn't get a
chance to heat up the rest of the computer. Air flows from one
side of the case to the other, with a duct guiding the air all
the way.

The case that I chose seems to have a problem getting enough air into
the case. Air flow is restricted by 75 holes in the front assembly and
is blocked by an ornamental baffle.
http://www.newegg.com/Product/ShowI...CurImage=11-119-030-12.JPG&Description=COOLER
MASTER Wave Master TAC-T01-EK Black All Aluminum Alloy ATX Mid Tower
Computer Case
http://tinyurl.com/9tjkw

If you insist on better thermal performance than can be achieved
with a Big Typhoon or a XP-120, there is always water cooling,
or if you are the devil-may-care kind of person, a Peltier+water
block. While a Peltier is a highly inefficient device, it
could shave a few more degrees off the temp the processor
is seeing. Of course, a big Peltier probably could use its own
ATX power supply, and the pump on the water cooling system might
appreciate a private power source as well. (There are computer
cases with room for two power supplies, so it can be done.)

This Peltier+water_block is 12V@18A and really could use its
own power supply. You can see in the plots of coolant temp
versus processor temperature, that you can achieve a processor
temperature of 20C, when the water flowing past is still at
42C. Now, at 216W of power for the Pelt, from a 70% efficient
power supply, where I live it would cost $0.42 a day to power
it, or $154 a year (and my power is cheap).

http://www.swiftnets.com/products/mcw5002-775T.asp

As for the Arctic Silver, try the following:

1) Install AS, accordng to the instructions on the AS web site.
2) Run the computer for several days. This gives the particles
in the AS a chance to "set" in their final position, and the
temps will fall a bit over the period of those several days.
3) On the third day, measure case air temperature and measure
CPU temperature. The difference is delta_T . Say you measure
55C under load on the CPU, and the case air temperature is
35C, then the delta is 20C.
4) A year from now, repeat the experiment in step 3. Use the
same program and same conditions, to apply a load to the
processor as you did in step 3. (Something like Prime95
will suffice). Now, depending on whether it is "summer"
or "winter" in your room, you are covered, as you are
measuring the delta_T and not the absolute temperature.
If it is a summers day, and CPU is 60C and case air temp
is 40C, the delta_T is still the same 20C.

You don't have to change the AS, until the delta degrades by
5C-7C or more. If the original installation managed a delta_T of
20C, and you measure it today and it is 27C, then change the
AS. In other words, let the performance of the cooling
solution, guide you in when to do maintenance.

HTH,
Paul

Thanks for the advise that was detailed in this portion, I greatly
appreciate the help.

 

[Brian]

How much further? It'd be good to know if you're expecting
a need to shed 20C more or just 5C, etc... at stock speed it
should be stable at least through the 60-70C range, but is
"with a load", a full load?

Actually it's more like a minor load being described, just about ~20%
of the resources as reported by Windows Task Manager. I've read reports
of this processor reaching 70-75C and still being stable, perhaps I am
just worrying too much and should let Intel's warranty handle the
problem if the processor burns. It would be nice to have the CPU hover
in the mid-40's performing this same task and peak at ~59C during a
thirty minute Prime95 Torture Test, but I don't think it's possible
with a Wavemaster case:

http://www.newegg.com/Product/ShowI...CurImage=11-119-030-12.JPG&Description=COOLER
MASTER Wave Master TAC-T01-EK Black All Aluminum Alloy ATX Mid Tower
Computer Case
http://tinyurl.com/9tjkw

A perforated intake grill behind a baffle and one 80mm exhaust does not
circulate enough air.

Well unplugging the fan the last 30 seconds might help a bit
more, but that's about all, just remove it while it's still
warm.

An alternate approach is to just take a hairdryer to the
sink, where the fan was to heat it up.

Excess thermal material will wipe off most easily with
petroleum solvent.



Arctic Silver and any other entirely synthetic based
compounds should not dry out, that was generally the most
common problem with silicone based compounds on small open
(flipchip) cores without a heat spreader.

Just put a tiny bit of AS in the middle of the core, don't
frost a cake when applying it since even a tiny bit will
spread out once the 'sink pressure is on it. It may take a
few days and thermal cycles to reach the stablized minimum
temp, dropping by single-digits until then.

Read about this single drop approach that eliminates the need to spread
it around with a credit card or fingers in a plastic bag.

Thanks for responding, I appreciate any and all help -

 

[Brian]

Thermal compounds are so cheap and don't dry out. You would
never know this from those who blindly hype Arctic Silver
(that is grossly overpriced). Since thermal solutions are so
inexpensive, you can bet Intel has already installed the best
solution.

CPU temperature in 50sC in a 74 degree room means CPU in the
70sC in a 100 degree F room. Now, what does an Intel spec
sheet say for maximum temperature? Not what you 'feel' is max
temperature. As long as CPU in a 100 degree room remains
below that max temperture, then CPU is in pig's heaven -
perfect and happy.

So first - what are the numbers? What does the Intel data
sheet demand? Will that CPU exceed Intel specs when room
temperature rises above 100 degree F? If not, then we have
Don Quixote chasing wind mills or hype promoted by Arctic
Silver. First get the data sheet numbers. And ignore hype
promoted by Arctic Silver advocates.

In late September, room temperatures were hovering in the mid-80's and
the processor was averaging 62C without an air duct during normal
operation. Intel states on their website that 66.6C is the maximum
operating temperature, so that point would easily be reached here in
Hawaii. Tasks such as Encoding a DVD have pushed temperatures up
to 67C just a couple days ago but it's not something which is done
everyday.

Thanks for bringing up the point of "chasing windmills", if the
addition of an air duct, changing the thermal compound, adding another
fan or purchasing another heatsink at last resort does not bring down
the temperatures, Intel's 3 year warranty will be the ultimate
solution.

 

[w_tom]

66.6 degree C? Which processor is this? That sounds rather
low?

An air duct should increase airflow through the heatsink.
Increased airflow causes the degree C per watt number to
decrease. But increasing airflow results in an exponentially
diminishing improvement. It typically takes something like
quadruple airflow, to only double the efficiency of that
heatsink - these only being ballpark numbers. A mild increase
in a gentle airflow causes significant cooling improvement. A
massive increase on an already large airflow causes very
little thermal transfer improvement.

 

[kony]

Actually it's more like a minor load being described, just about ~20%
of the resources as reported by Windows Task Manager.

I'm not talking about resources, rather an application
pegging CPU at 100%... run Prime95's Torture Test, large
in-place FFTs test for a half hour to determine a (near)
full load temp. Consider that temp the baseline to add any
additional summer ambient temp changes onto, for final
judgement of whether target temp was achieved.

I've read reports
of this processor reaching 70-75C and still being stable, perhaps I am
just worrying too much and should let Intel's warranty handle the
problem if the processor burns.

Sure, it's quite likely it could be stable at 75C, but
individual specimens may be more or less stable around this
limit... it's too high to be considered a safe upper ceiling
for operational stability but it is cool enough that no
permanent damage should result from intermittent operation.

So that's not the threshold for damage, just instability.
It's supposed to throttle back to lower speed temporarily,
before the damage threshold was reached.

Intel's warranty should not cover you if you let the CPU
overheat. They "might", but they shouldn't, cover it
because the warranty is within context of proper use,
including system designed to accomodate needed airflow
including use in higher ambient ranges. With the intel
retail heatsink installed properly and in a properly
ventilated chassis it should not be getting high enough to
be damaged.

It would be nice to have the CPU hover
in the mid-40's performing this same task and peak at ~59C during a
thirty minute Prime95 Torture Test, but I don't think it's possible
with a Wavemaster case:

http://www.newegg.com/Product/ShowI...CurImage=11-119-030-12.JPG&Description=COOLER
MASTER Wave Master TAC-T01-EK Black All Aluminum Alloy ATX Mid Tower
Computer Case
http://tinyurl.com/9tjkw

A perforated intake grill behind a baffle and one 80mm exhaust does not
circulate enough air.

Don't worry too much about the 80mm exhaust, it's not
optimal for noise reasons but the case is sufficient to keep
the CPU cool enough... I think a 59C peak is not an
unreasonable target if ambient temp stays moderate... I
can't guesstimate about summer temps since we have no idea
just what increase that'd be.

It might help some to change the state of the intake grill,
but I can't tell you how much, I dont' have that case.

Read about this single drop approach that eliminates the need to spread
it around with a credit card or fingers in a plastic bag.

There is no need to use a credit card or plastic bag to
spread compound. It's not hazardous to skin and if you had
a lot of dirt on your hands you could simply wash them
first. Geeks like to spread these crazy ideals about how
things should be done (and in some cases, I probably do it
too in other topics) but there's no need for some like of
spreading tool at all... IF it was a good idea to frost it
like a cake but it isn't since there isn't even an assurance
that both the spreader and the 'sink themselves were
perfectly flat, and further that this will result in too
much compound. Due to the head spreader, all the excess
wouldn't be forced out, and it could even cause trapped air
in the middle unless WAY too much compound were applied.

If all else fails then get a different heatsink... if your
summer ambient temps are going to be significantly higher
that would be the clearest solution.

 

[kony]

66.6 degree C? Which processor is this? That sounds rather
low?

An air duct should increase airflow through the heatsink.
Increased airflow causes the degree C per watt number to
decrease. But increasing airflow results in an exponentially
diminishing improvement. It typically takes something like
quadruple airflow, to only double the efficiency of that
heatsink - these only being ballpark numbers. A mild increase
in a gentle airflow causes significant cooling improvement. A
massive increase on an already large airflow causes very
little thermal transfer improvement.

What a duct commonly does (varies per particular case and
duct used) is provide cooler intake air temp. It does so by
avoiding air pre-heated by the rest of the system which had
entered the front of the case, but even as significant, if
the duct comes down in close proximity to the heatsink's fan
intake then it reduces recirculation of the heatsink's
heated exhaust, more than having only the rear case exhaust
fan.

It doesn't necessarily make it a good idea for some case
designs because many rely on a passive intake in the front
to cool the drives, but with the case the OP has there are a
pair of forward mounted fans pushing through the drive rack.

 

[Paul]

Thanks for the advise that was detailed in this portion, I greatly
appreciate the help.

I still think you should use Throttlewatch, while running Prime95
or a similar 100% load. You want to see whether you are being
robbed of performance by this thermal problem. I mean, if
you don't really need the full performance of your processor,
you could always turn down the processor clock a bit. If you want
to get the performance you paid for, then Throttlewatch or a
similar program, can tell you whether the processor is really
that hot or not. It could be a temperature readout error, after
all.

The processor throttling feature is supposed to limit the
temperature rise. If you were expecting a dramatic rise
in temperature, when a load is applied to the processor, the
processor will resist the load, by slowing down execution.
Throttlewatch is a program that displays the status of this
temperature-based throttling feature, and it tells you whether
the processor has really hit ~70C and started throttling back.

Any system should be characterized at full load. Say, for
example, the computer hangs in a tight loop, while you are
away from the computer. You want your cooling system to be
able to run 24/7 at 100% computing load, and maintain a
usable temperature range. A program like Prime95 (mersenne.org)
and the torture test option, is a good way to get a 100% load.

Paul