I was concerned about the direct contact of the heat pipes with the CPU
because a heat pipe cooler relies on the Liquid inside to heat ..rise...cool
and then return to start the cycle all over and with those copper pipes
touching the CPU they would slowly heat up........all the way to the top and
the liquid would not cool as efficiently.
Looking at the temps achieved I guess it does work......but the charts also
show that the higher the OC the less efective this heatsink is...look at the
faster rise in temps at higher loads/OCing compared to a Thermalright Unit.
Comparing a HeatPipe Cooler to a WaterCooled system is unfair..........very
very few Heatpipe coolers are able to achieve that type of cooling.
If you are willing to spend the $$$$ water cooling is the way to go
if you are just looking to buy a cooler to keep the temps down with moderate
OCing for a reasonable amount of $$$ by all means buy the unit...if you can
find it.
If you cannot find it and/or are aiming for a higher OC..........take a good
look at those Thermalright heatsinks....you supply the fan.
I love Thermalright and have used their coolers for the past 8 years on AMD
as well Intel CPUs.They have a simpleness to them,an easy mounting
system(most of the time)
and your choice of fan.....a very high velocity screamer or a nice and quiet
unit....you get to chose.
peter
"Jack R" <jackr_nospam@msn.com> wrote in message
news:8GTli.717$AV7.642@newsfe06.lga...
>
> "Phil Weldon" <not.disclosed@example.com> wrote in message
> news:V1Tli.6887$rR.374@newsread2.news.pas.earthlin k.net...
>> 'Jack R.' wrote:
>> | Thanks for your thoughtful response, it's appreciated.
>> _____
>>
>> I believe that, for CPU cooling, heat pipes are mainly useful for
>> efficiently moving the heat to a convenient distance where larger
>> dissapation fins and larger, slower fans can be used, and where the
>> tranfered heat can be dumped further from the motherboard, chipsets,
>> memory,
>> display adapter ... That's why you don't see heat pipes used with water
>> cooling for CPUs.
>>
>> Phil Weldon
>>
>>
> Heat pipes (in a narrow range of operation) have an equivalent thermal
> resistance that is a small fraction of even pure copper. The thin wall of
> the heat pipe is an advantage, not a drawback. This is why the thermal
> response can be superior to a block of copper.
> One of the limitations is that of heat capacity..it's limited with heat
> pipes, thus you see multiple pipes being used commonly.
> The huge advantage of a water system is just that: heat capacity, limited
> mainly by how much water you have in the system (and flow, delta-T,
> ability to extract heat, etc.).
> Jack R
>
> I also have reservations about the base design
> of the Ice Age 120 with the gaps there. They say it leaves space for
> excess thermal compound to go, but I think they are there because that
> is the only way to have direct contact of the heat pipes to the heat
> spreader on the processor using their design....:-).
The four pipe design places a gap directly in the center. In the orientation,
where the gap is parallel to the longitudinal axis of the underlying elongated
chip of the dual core processor, this design might actually be detrimental to
efficient heat transfer.
'Ed Medlin' wrote, in part:
| Looking at reviews of both the Ice Age 120 and Thermaltake i1, I
| like the theory behind the Ice Age 120, but am also leerie about the
| FrostyTech method of testing.
_____
The IceAge 120 design is clearly inferior. The heat pipes only work to move
heat longititudinaly, not circumferentialy. The thermal resistance to heat
to the complete circumference of the hot end of the heat pipe is less when
embedded in an aluminum block than in a copper block. Add that to the
higher thermal resistance of the alminum block when conducting heat from the
entire surface of the heat spreader to the heat pipes and you have just
another example of advertizing over substance. There is no 'theory' behind
the 'IceAge 120', just marketing.
Phil Weldon
"Ed Medlin" <ed@edmedlin.com> wrote in message
news:Rj5mi.27746$C96.12360@newssvr23.news.prodigy. net...
|
| "Phil Weldon" <not.disclosed@example.com> wrote in message
| news:JIWli.6922$rR.1826@newsread2.news.pas.earthli nk.net...
| >| Heat pipes (in a narrow range of operation) have an equivalent
| >thermal
| > | resistance that is a small fraction of even pure copper. The thin
| > wall of
| > | the heat pipe is an advantage, not a drawback. This is why the
| > thermal
| > | response can be superior to a block of copper.
| > | One of the limitations is that of heat capacity..it's limited with
| > heat
| > | pipes, thus you see multiple pipes being used commonly.
| > | The huge advantage of a water system is just that: heat capacity,
| > limited
| > | mainly by how much water you have in the system (and flow, delta-T,
| > ability
| > | to extract heat, etc.).
| > _____
| >
| > Wrong way to look at it; the major thermal resistance is silicon /
| > heat
| > spreader and interface, NOT a few millimeters of a copper heat block.
| > Heat
| > pipes only transfer heat, and cannot do better than ambient. With the
| > current Core Duo package, the only counter to that bottle neck is a
| > higher
| > temperature differential - active phase change, Peltier arrays, water
| > chiller ...
| >
| > Even though heat pipes can have lower thermal resistance than any
| > solid,
| > over a few millimeters distance the increased thermal resistance
| > doesn't add
| > up to much, over a few centimenter or more it does.
| >
| > If you will look closely at the 'FrostyTech' review of the IceAge 120
| > and
| > other heatsink/fan combinations you will see that their test
| > arrangement is
| > deeply flawed. 'FrostyTech' is measuring the temperature rise above
| > ambient
| > of the surface of the hot plate, NOT the temperature rise of a silicon
| > chip
| > beneath a heatspreader. The numbers 'FrostyTech' develops, while
| > collected
| > with a snazzy looking instrument, do not represent the numbers the
| > overclocker will ever see.
| >
| > If you are ever able to get an IceAge 120, it will be interesting to
| > compare
| > your results with the ThermalTake i1 I have. The ThermalTake i1 has
| > heat
| > pipes inbeded in a copper block rather than in an aluminum block and
| > the
| > exposed area of the heat dissapation fins is much larger than for the
| > IceAge
| > 120 (and the fan has a four-pin connector to mate with the Intel CPU
| > thermal
| > fan speed control.) My guess is that the ThermalTake i1 will be a
| > better
| > performer.
| >
| > Phil Weldon
| >
| I have seen the 'transient' spikes Jack is speaking of even with my
| water cooled C2D setup. Under load, I have seen spikes jump from 45c up
| to 55-60c and immediately back to 45c using CoreTemp on one core or the
| other but have never seen it happen on both at the same time. These
| spikes last only one cycle (I forget the default timings of CoreTemp's
| readings) and I really don't look at a few m.s. of temperature rise as a
| problem. I don't even know if they are real or just an anomoly of the
| sensors/software so I really don't care...:-).
| Looking at reviews of both the Ice Age 120 and Thermaltake i1, I
| like the theory behind the Ice Age 120, but am also leerie about the
| FrostyTech method of testing. I think an actual test using "real"
| silicon might be better. I also have reservations about the base design
| of the Ice Age 120 with the gaps there. They say it leaves space for
| excess thermal compound to go, but I think they are there because that
| is the only way to have direct contact of the heat pipes to the heat
| spreader on the processor using their design....:-). The TT i1 uses a
| more tried and true method with the copper block and large dissipation
| area that has proven successful for some time. I would also like to see
| a comparison of the two in real life usage.
|
| Ed
|
|
Phil Weldon wrote:
> 'Jack R.' wrote:
> | Thanks for the response.
> | SpeedFan and others show 15 - 20 deg C near-instantaneous
> | transients in
> core
> | temperatures.
> | If these measurements are real, I'm concerned about long term
> | reliability issues due to thermal expansion/contraction with this
> | type of continuous thermal cycling. So, I'd like to see them
> | dampened.
> | My OC testing shows that these transients are magnified greatly by
> | OC'ing. At nominal settings, stock HS, they are in the 5 - 10 deg C
> | range max...probably acceptable.
> | Of course, only time will tell. If the reliability is reduced from
> | 15
> years
> | to 8 years, then who cares?
> | But, if it's 10 years to 10 months, then we've got a problem.
> |
> | Still looking for a source for the Iceage 120...
>
> _____
>
> You are worrying about the wrong thing.
>
> First of all, how would the temperature monitors for the CPU detect an
> 'instantaneous' temperature change? There is a lag in the A-to-D
> conversion that eliminates that possibility. How often do you poll
> the monitor chip for temperatures?
>
> Secondly, the mass of the CPU chip is small compared to the heat
> spreader and very small compared to the heatsink. You may be reading
> the temperature at the internal thermal diode for each CPU, but the
> temperature of the heat spreader and heat sink are very different.
> In fact, contrary to what you believe, a larger heatsink DAMPS the
> temperature swings.
>
> What you NEED to worry about is the heat transfer capacity of the
> cooling solution, and the temperature of the cold side of the
> solution. At the moment, the weak link is heat transfer between the
> CPU chip and the heat spreader. The best you can do about that is
> have the highest reasonable heat transfer between the heat spreader
> and the cold side of the cooling solution. The greater the
> temperature difference, the more rapid the energy transfer.
Exactly. q=delta-T * -k. For the OP, your choices are:
1. Change k (different material)
2. Change delta-T (different fans, different fluids on the other side of the
HS)
Phil Weldon wrote:
> ....the major thermal resistance is silicon / heat spreader and
> interface...the only counter to that bottle neck is a higher temperature
> differential - active phase change, Peltier arrays, water chiller....
We are at the "...tools are being developed to define the surface
channels..." stage. But wait, why can't we develop our own technique
meanwhile. In my next project, I'm going to use an ink eraser, the type
which is slightly tougher than the regular lead pencil eraser, to lightly
score the mirror-polished base plate of my brand new heat sink in like
pattern, X crosses within increasingly smaller, enclosed squares. The
dollop of heat transfer paste will be placed at the centre and I'll let the
heat sink (plus its fan) sit on this gently. There will be no more than the
weight of this twin device doing the actual spreading.
--
Lin Chung.
[Paste ntlworld over the Water Margin to send a private e-mail.]
'Lin Chung' wrote:
| We are at the "...tools are being developed to define the surface
| channels..." stage. But wait, why can't we develop our own technique
| meanwhile. In my next project, I'm going to use an ink eraser, the type
| which is slightly tougher than the regular lead pencil eraser, to lightly
| score the mirror-polished base plate of my brand new heat sink in like
| pattern, X crosses within increasingly smaller, enclosed squares. The
| dollop of heat transfer paste will be placed at the centre and I'll let
the
| heat sink (plus its fan) sit on this gently. There will be no more than
the
| weight of this twin device doing the actual spreading.
_____
That's an interesting article, thanks for posting it.
I have a few ideas myself, and a few comments on the IBM method.
Overclockers have long attempted to enhance heat tranfer through the
CPU/Heatsink interface. Many of us lapped and polished early X86 CPUs to
get flatter and smoother surfaces. Some of the packages, similar to the
heat spreader cap method used on Core Duo CPUs just about REQUIRED some sort
of extra work - the caps were concave with perhaps as much as a millimeter
depth near the center of the cap - and that was on just the VISIBLE lack of
parallelism and flattness. There is also the interface between the silicon
chip and the inner surface of the heat spreader cap. The problem IBM is
addressing is mainly an assembly problem, I think, rather than a surface
treatment problem. The technique for the past five years or so (perhaps
'Coppermine' forward) is to use a thermal pad, perhaps even a pad with a
mesh imbedded, between the thermal spreader and whatever heatsink the system
builder installs. This method is aimed at the cheapest, lowest common
denominator of installation care and skill. The resulting thermal
resistance is much higher than necessary compared to using greater care and
better technique in insuring the interfaces are
smooth
flat
and parallel.
The Intel CPU series (pre Tualatin) that extended the Pentium III up to 1
GHz used package that placed the essentially bare silicon chip on a carrier;
no spreader cap. This CPU was about 9 X 11 mm, or 1 sq. cm in area; the
surface was optically flat (laser sliced and polished to within a micrometer
or so. Heat transfer for this CPU package vastly improved IF the mating
heatsink surface was also
smooth
flat
and parallel. No added thermal resistance from a two surface heatspreader.
However, even though the power dissapation was low by current standards (30
to 50 Watts, even for high overclocks), the heat flux of 40 Watts through 1
sq. cm is GREATER than the heat flux through a like area of a 2500 Watt
electric stovetop element. BUT the extremely smooth and flat silicon
surface, when mated with high parallelism with a similarly flat and smooth
heatsink surface greatly reduced the thremal resistance. NO THERMAL
COMPOUND AT ALL was nearly as good as the best thermal compound then
available (and unsalted butter worked well also - except for the tendency to
go rancid B^)
To my mind, DIRECT impingment of a cooling fluid onto the CPU die itself is
the way to go. Either through channels cut into the silicon die or through
surface channels cut into a mating surface. Proper treatment of the silicon
die can handle any chemical effects of the cooling fluid; water had no
effect on the top surface of the Intel Pentium III 1 GHz bare die, for
example. I will seriously consider removing the thermal spreader cap on my
Core 2 Duo E4300 and trying such a cooling method when it retires from its
present 2.7 GHz @ 1.250 volt duty.
Other things to consider:
Diamond lapping compound; diamond has the highest thermal conductivity of
any material, it can exceed the thermal conductivity of silver by as much as
a factor of 5, though the polycystaline form has a thermal conductance more
like twice that of silver. (Thermal conductivity of the filler material is
not the only concern, of course, witness the very marginal advantage of
Arctic Silver over RadioShack zinc oxide filled thermal compound.
Increasing the temperature difference between the heatsink and the silicon
die; increased delta t works just as well as decreased thermal resistance -
and is much more extensible. The difficulty in greater delta t is the
cooling-below-ambient method. Peltier arrays are one solution, but the
total system thermal dissapation is greatly increased because of the Peltier
array inefficiency (~ 250 Watts electrical power input to a Peltier array
that can pump 80 Watts of heat accross a 30 C temperature rise [CPU side of
the Peltier array 30 C BELOW heatsink side of the Peltier array]). Greater
demand for small, efficient phase change active coolers (efficency gain of
10 X compared to the Peltier array can reduce the expense of such cooling to
reasonable levels. Sonic cooling has been demonstarated (and used in
orbital experiments); sonic cooling requires no moving parts (other than
the quivelent of piezo-accoustic transducers.)
Ultimately there is no choice but to move to such cooling because, if for no
other reason, increasing the speed of calculation REQUIRES reduction in size
and increase in element density - reducing leakage and operating voltage can
only proceed so far.
As far as your project; I think that the starting point should be
elimination of the heat spreader and direct heatsink/silicon die interface -
otherwise the effect of etching the heatsink surface will be lost in the
noise of the much higher thermal resistance at the silicon die/heatspreader
interface.
(Diamond lapping compound, in a dozen or so graded particle sizes, is
cheaper than you might think. McMaster-Carr, an industrial tool supply
company, offers the compound in various grades, in amounts sufficient for
several interfaces, for a cost in the $15 to $60 US range.)
(Peltier arrays have dropped in price, and can be had for ~ $25 US each for
100 Watt tMax 40 mm X 40 mm sizes. The difficulty here is coupling 50 sq.
cm. of Peltier array surface to ~ 1 sq. cm silicon die { possible by using a
heat spreader with heatpipes or active fluid circulation to tranfer the heat
between the silicon die outside the system.})
At a certain point, of course, the performance gain is less expensive to
realize by purchasing a Core 2 Extreme and overclocking with convential
means B^)
Phil Weldon
"Lin Chung" <lin.chung@the.Water.Margin.com> wrote in message
news:3nbni.34108$nE2.13998@newsfe3-win.ntli.net...
| Phil Weldon wrote:
| > ....the major thermal resistance is silicon / heat spreader and
| > interface...the only counter to that bottle neck is a higher temperature
| > differential - active phase change, Peltier arrays, water chiller....
|
|
|
| "Researchers unveil details of chip cooling breakthrough"
| http://www.physorg.com/news93798065.html
|
| We are at the "...tools are being developed to define the surface
| channels..." stage. But wait, why can't we develop our own technique
| meanwhile. In my next project, I'm going to use an ink eraser, the type
| which is slightly tougher than the regular lead pencil eraser, to lightly
| score the mirror-polished base plate of my brand new heat sink in like
| pattern, X crosses within increasingly smaller, enclosed squares. The
| dollop of heat transfer paste will be placed at the centre and I'll let
the
| heat sink (plus its fan) sit on this gently. There will be no more than
the
| weight of this twin device doing the actual spreading.
|
| --
| Lin Chung.
| [Paste ntlworld over the Water Margin to send a private e-mail.]
|
|
|
|
> I will seriously consider removing the thermal spreader cap on my
> Core 2 Duo E4300 and trying such a cooling method when it retires
> from its present 2.7 GHz @ 1.250 volt duty.
Oh, what might you be planning to replace it with?
The guys over at the xtremesystems.org forums say the E4xxx series
all use thermal paste under the heat spreader. Carefully cutting with
a razor blade the rubbery cement around the periphery of the heat
spreader reportedly is all that is required to remove it. This could
certainly account for the higher temperatures generally incurred by
the E4xxx series for equivalent clock speeds. The E6xxx series are
reported to use a low temperature solder which melts around 100° C.
I am currently entertaining theories which explain the higher default
voltages of the Allendale core CPUs. I get the feeling that Intel may
have taken measures to deliberately impair the overclockability of the
less expensive CPUs.
Problems which must be faced by those removing heat spreaders
include the fact that the CPU retention mechanism clamps down on
the edges of the heat spreader. If the outer portion of the heat spreader
were retained, the CPU core would now be lower than the retention
mechanism, requiring modification to either the heatsink or the retention
mechanism (or both). Of course, the heatsink clamping pressure would
be less because of the absence of the heat spreader thickness, and
crushing of the tiny core also becomes a distinct possibility. I have read
that some have removed the retention mechanism entirely and depend
upon the clamping force of the heatsink to hold the CPU in place!
'Fishface' wrote, in pieces:
| Phil Weldon wrote:
|
| > I will seriously consider removing the thermal spreader cap on my
| > Core 2 Duo E4300 and trying such a cooling method when it retires
| > from its present 2.7 GHz @ 1.250 volt duty.
|
| Oh, what might you be planning to replace it with?
_____
Price, timing, what the future brings, and how far downstream my current
EVGA 680i motherboard will soldier on.
If the EVGA 680i will handle 45 nm CPUs, then the problem would be what to
run the modifed E4300 ON. Otherwise, at the point a new motherboard is
required, the E4300 has a future with plastic surgery. With 2.7 GHz stable
overclock and 1200 MHz stable memory, I don't feel a lot of NEED for speed,
but curiosity may make me twitchy. At any rate, unless I get a zonbu or
minimac, I intend to keep the E4300/680i up and running.
| I get the feeling that Intel may
| have taken measures to deliberately impair the overclockability of the
| less expensive CPUs.
_____
I think Intel is just using a less expensive heatspreader system on their
less expensive CPUs. The new and meaner Intel, I think, is using the
overclocking buzz to kick sand in AMDs face, and wouldn't go out of the way
to reduce overclocking potential. After all, the vast majority of Intel
CPUs are never in any danger of being overclock, and the buzz helps Intel
more that compute power differentiation among the Core Duo series CPUs.
| I have read
| that some have removed the retention mechanism entirely and depend
| upon the clamping force of the heatsink to hold the CPU in place!
_____
Just glue the sucker down! In the near future I see a Intel producing a
card about the size of the old Slot II processor card that has the CPU/s,
chipset, and memory hardwired in, the whole bathed in liquid coolant, with
connector fingers on one edge and coolant fittings on the other. System
cases shrink to shoebox size with the power-supply/coolant-chiller in a
similar sized box.
Phil Weldon
"Fishface" <invalid@ddress.ok?> wrote in message
news:lTdni.5809$Wh4.1332@trndny06...
|
| Phil Weldon wrote:
|
| > I will seriously consider removing the thermal spreader cap on my
| > Core 2 Duo E4300 and trying such a cooling method when it retires
| > from its present 2.7 GHz @ 1.250 volt duty.
|
| Oh, what might you be planning to replace it with?
|
| The guys over at the xtremesystems.org forums say the E4xxx series
| all use thermal paste under the heat spreader. Carefully cutting with
| a razor blade the rubbery cement around the periphery of the heat
| spreader reportedly is all that is required to remove it. This could
| certainly account for the higher temperatures generally incurred by
| the E4xxx series for equivalent clock speeds. The E6xxx series are
| reported to use a low temperature solder which melts around 100° C.
| I am currently entertaining theories which explain the higher default
| voltages of the Allendale core CPUs. I get the feeling that Intel may
| have taken measures to deliberately impair the overclockability of the
| less expensive CPUs.
|
| Problems which must be faced by those removing heat spreaders
| include the fact that the CPU retention mechanism clamps down on
| the edges of the heat spreader. If the outer portion of the heat spreader
| were retained, the CPU core would now be lower than the retention
| mechanism, requiring modification to either the heatsink or the retention
| mechanism (or both). Of course, the heatsink clamping pressure would
| be less because of the absence of the heat spreader thickness, and
| crushing of the tiny core also becomes a distinct possibility. I have
read
| that some have removed the retention mechanism entirely and depend
| upon the clamping force of the heatsink to hold the CPU in place!
|
|
Re: CPU Heatsink info. 
Linear Mode
