COPPER CPU SHIMS: The Definitive Answer !!

[rms]

Ok, my gut feeling was that installing a copper shim was a mod that could
provide measurable benefits. My reasoning was as follows: Presently my
Maze3 waterblock is only slightly warm at the edges of the block, indicating
a large thermal gradient between the center water channel and the outer
water channel, and thus plenty of cooling power still available if some heat
could be diverted to the outer portion of the waterblock, instead of simply
the 1cm x 1.5cm cpu core area (non-heat-spreader cpu).

Now, given that the cpu thermistor is on the other side of the substrate on
the botom of the cpu and is registering quite a high temp, and that the cpu
electrical traces (which permeate the substrate) most definitely are
significant thermal conductors, it seems very reasonable to me to conclude
that the opportunity is there to lower temps by this multiplication of the
surface area touched by the waterblock by probably 7x or 8x.

So. On my AthlonXP 1700+ running at 10.5 x 230fsb = 2.4ghz @ 1.9v cpu, I
installed a copper shim from www.1coolpc.com (I bought the XP shim, but it
required a lot of Dremel work to get it to fit -- I suspect this is
actually a Barton shim; but anyway). I was very careful to check that the
shim did not sit higher than the cpu core, to file all edges smooth, and to
check that the shim sat completely flat on the cpu substrate. In addition,
and this is an important touch, I smeared both sides of the shim with
ArticSilver5, a thin coat. Don't get careless and blow off this step as you
will compromise much of the effectiveness of the mod.

The results are extremely satisfying!! Immediately after installation
Prime95 has stabilized at 54C, probably 10C lower than previous, and after
curing will certainly drop even farther. Now, I hear the skeptics scoffing,
and realize that it's likely the thermal characteristics of the copper shim
are not the only factor for the dramatic improvement.

I suspect that other factor consists of the large-surface-area copper shim
forcing the (large and heavy) waterblock to lie completely flat on the cpu
core. Although I've always used a fiber washer, clearly it was not
preventing a slight rocking of the heatsink from one side of the core to the
other, which could be caused by unequal spring pressure or simply gravity.

Both of these factors make the copper shim a must-have addition to a
high-performance Barton or other non-heatspreader-type cpu installation,
IMHO. There is no doubt in my mind that people running very heavy heatsinks
like the SP-97 or big waterblocks will see an immediate temperature
reduction from careful installation of a copper shim.

rms

 

[Michael Brown]

rms wrote:
[...]

The results are extremely satisfying!! Immediately after installation
Prime95 has stabilized at 54C, probably 10C lower than previous, and
after curing will certainly drop even farther. Now, I hear the
skeptics scoffing, and realize that it's likely the thermal
characteristics of the copper shim are not the only factor for the
dramatic improvement.

I know this is a bit more work, but the results would be significantly more
meaningful if you remounted the block several times, both with and without
the shim. I've noticed up to a ~5 deg C variation (as in +/- 3 deg C or so)
of diode temps when remounting a TT Silent Boost, so there's a good chance
the real change is reasonably different from what you got above. Finally, it
would be great if you could find a plastic (not mica, which is also a good
thermal conductor) shim and repeat. Most plastic shims are good thermal
insulators.

I suspect that other factor consists of the large-surface-area copper
shim forcing the (large and heavy) waterblock to lie completely flat
on the cpu core. Although I've always used a fiber washer, clearly
it was not preventing a slight rocking of the heatsink from one side
of the core to the other, which could be caused by unequal spring
pressure or simply gravity.

This would show up in repeated tests. Another thing that could be causing it
is that you didn't crank it as tight when you first mounted it (or it
loosened over time)without the shim. Again, repeated measurements would
demonstrate this.

One thing you might want to try, if you know people with the equipment
(local university or search and rescue, possibly?), is to take thermal
images of the block from above, with and without the shim. If your
temperature is accurate (and assuming water temp ~20 deg C), then a drop of
10 deg C implies that you have about 20% of the heat flowing through the
shim, which should definately show up on a thermal image.

[...]

 

[rms]

If your

temperature is accurate (and assuming water temp ~20 deg C), then a drop
of
10 deg C implies that you have about 20% of the heat flowing through the
shim

I did not imply this. If the difference is simply a result of a skewed
heatsink, the shim would show no temperature increase. I'm saying that a
copper shim gives dual benefits: a flatter more stable platform for the
heatsink, and additional surface area for heat dissipation. Where the
balance lies between those two I can't say.

rms

 

[Alceryes]

A slightly easier way to test would be to put a thermistor on the outer edge
of the heatsinks contact surface both with and without the shim.

 

[Robert Redelmeier]

as follows: Presently my Maze3 waterblock is only slightly
warm at the edges of the block, indicating a large thermal
gradient between the center water channel and the outer water

True if the center water edges are much warmer than the
outside edges. Did the outside edges warm up with the shim?

The results are extremely satisfying!! Immediately after
installation Prime95 has stabilized at 54C, probably 10C

No denying data. I wouldn't call prime95 very severe,
but I'm biased.

I suspect that other factor consists of the
large-surface-area copper shim forcing the (large and
heavy) waterblock to lie completely flat on the cpu core.
Although I've always used a fiber washer, clearly it was
not preventing a slight rocking of the heatsink from one
side of the core to the other, which could be caused by
unequal spring pressure or simply gravity.

Rocking or edge contact with a die-back is extremely
detremental to heat-transfer, even with high-k heat compound.
I always check for this and correct grease spreading by very
slightly rotating the HS after installation. I'm looking
for a very heavy, smooth drag. Any sign of scraping (other
than from the clamp) means trouble.

I suspect this was the biggest effect, but heat spreaders
can be useful (seen by a reduction in center/edge gradients
on the HS)

-- Robert author `cpuburn` http://pages.sbcglobal.net/redelm

 

[BigBadger]

If your 'old' temps were 64C with water cooling you obviously has something
seriously wrong with your installation. Even 54C is not great given that I'm
getting 44C with air cooling at 2640MHz / 1.93V (Abit NF7-S / XP-M 2500+ /
SP-97, no shim).
Shims make no significant difference to cooling performance, any properly
installed heatsink will already be sitting flat on the CPU anyway.

 

[rms]

I suspect this was the biggest effect, but heat spreaders

can be useful (seen by a reduction in center/edge gradients
on the HS)

Heatsinks do perform differently when used with large-area vs.
small-area heatsources, as shown by any of the
http://www.frostytech.com/index.cfm reviews, and it seems evident to me that
waterblocks will always do better with a large-area heatsource.

-- Robert author `cpuburn` http://pages.sbcglobal.net/redelm

cpuburn does indeed raise cpu temps farther than prime95, in my case
about 4C higher. The reason I've always used Prime95 is the error-checking
that I know it does, which gives me a clear indication that I have a
problem. It's not clear to me that cpuburn has any error-checking?

rms

 

[rms]

If your 'old' temps were 64C with water cooling you obviously has

something seriously wrong with your installation. Even 54C is not great
given that I'm getting 44C with air cooling at 2640MHz / 1.93V (Abit NF7-S
/ XP-M 2500+ / SP-97, no shim).
Shims make no significant difference to cooling performance, any properly
installed heatsink will already be sitting flat on the CPU anyway.

A typical lousy off-hand response. I'm running a 2yr-old xp, not a
mobile barton. And I suspect that *most* installations do not conform to
your definition of a 'properly installed heatsink'.

rms

 

[BigBadger]

--
*****Replace 'NAPALM'S with 'bitternut's in the reply address*****

A typical lousy off-hand response. I'm running a 2yr-old xp, not a
mobile barton. And I suspect that *most* installations do not conform to
your definition of a 'properly installed heatsink'.

rms

Suspect what you want.... Maybe most of the ones you build are not properly
installed, in fact I'm pretty certain of it.
Certainly every system that I've ever put together achieves my definition of
'properly installed' (and I've built hundreds) and even the ones with stock
air cooling achieve temps similar to your water cooled ones.

I've also tried shim's and I know for certain they make no difference that
can be recorded .... unless you are comparing to a system built by an ape.

 

[BigBadger]

*****Replace 'NOSPAM' with 'btinternet' in the reply address*****

Heatsinks do perform differently when used with large-area vs.
small-area heatsources, as shown by any of the
http://www.frostytech.com/index.cfm reviews, and it seems evident to me
that waterblocks will always do better with a large-area heatsource.


rms

A shim does not increase the 'heat source' area... The only part of a chip
that produces heat is the core and without increasing the size of the
silicon you cant make the core area larger.
The CPU substrate produces no heat and is also a VERY poor conductor of heat
from the core, therefore the heat path from the core, thru' the substrate,
thru' the shim, thru the air gap (because the shim has a small clearance
between substrate and heatsink) and into the heatsink is virtually non
existant.

 

[rms]

The CPU substrate produces no heat and is also a VERY poor conductor of

heat from the core, therefore the heat path from the core, thru' the
substrate, thru' the shim, thru the air gap (because the shim has a small
clearance between substrate and heatsink) and into the heatsink is
virtually non existant.

Guess you didn't read or didn't understand what I wrote in the original
post, in which case there is no reason to respond further to you.

rms

 

[Robert Redelmeier]

cpuburn does indeed raise cpu temps farther than prime95, in
my case about 4C higher. The reason I've always used Prime95
is the error-checking that I know it does, which gives me a
clear indication that I have a problem. It's not clear to me
that cpuburn has any error-checking?

Yes, burn* do error checking on the calcs they do. I don't
push this feature because I'm not exhaustively checking for
all possible CPU errors (fully exercising the instruction set).

I've had very few reports of error reports except on burnBX
and burnMMX which usually abend from memory/bus errors.

-- Robert

 

[BigBadger]

*****Replace 'NOSPAM' with 'btinternet' in the reply address*****

Guess you didn't read or didn't understand what I wrote in the original
post, in which case there is no reason to respond further to you.

read it, understood it, disagreed with it.

 

[BigBadger]

Now, given that the cpu thermistor is on the other side of the substrate

on the botom of the cpu and is registering quite a high temp, and that the
cpu electrical traces (which permeate the substrate) most definitely are
significant thermal conductors, it seems very reasonable to me to conclude
that the opportunity is there to lower temps by this multiplication of the
surface area touched by the waterblock by probably 7x or 8x.

Just because the substrate gets hot on the back does not mean it's a good
conductor.... You stick a pizza in the oven and it gets hot... does this
mean pizza is a good conductor? All it means is that the pizza or CPU
substrate is in a hot place and therefore warms up... it does not mean that
either is good at transmitting heat from one place to another.

 

[Noozer]

Just because the substrate gets hot on the back does not mean it's a good
conductor.... You stick a pizza in the oven and it gets hot... does this
mean pizza is a good conductor? All it means is that the pizza or CPU
substrate is in a hot place and therefore warms up... it does not mean that
either is good at transmitting heat from one place to another.

Uhm... If you touch somthing and it feels warm/hot then it IS conducting
heat and putting a sink on it will help keep it cool.

....and dont forget that there were some ceramic CPU's that would really
benefit from sinking the whole top of the CPU instead of just the core.

 

[BigBadger]

*****Replace 'NOSPAM' with 'btinternet' in the reply address*****

Uhm... If you touch somthing and it feels warm/hot then it IS conducting
heat and putting a sink on it will help keep it cool.

...and dont forget that there were some ceramic CPU's that would really
benefit from sinking the whole top of the CPU instead of just the core.

yeah, everything conducts heat to some extent but not everything is a 'good
conductor'..... Athlon XP's (which is what we are referring to) do not have
ceramic substrates. The substrate on an XP is some form of resin, while I
accept does conduct heat it does not do it very well. I'd hazard a guess
that it conducts orders of magnitude less well than the copper/aluminium of
the heatsink that is in direct contact with the CPU and therefore the
quantity of heat that would take the substrate>shim>heatsink path in favour
of the easier direct path to the heatsink is very very small.

 

[David Maynard]

Uhm... If you touch somthing and it feels warm/hot then it IS conducting
heat and putting a sink on it will help keep it cool.

I'm afraid that what you're trying to imply isn't true. Yes, the chip
carrier gets hot and yes, you can cool IT. Unfortunately, that doesn't do
squat for cooling the CPU: the thing of concern.

...and dont forget that there were some ceramic CPU's that would really
benefit from sinking the whole top of the CPU instead of just the core.

An impossible 'comparison' since you can't cool 'just the core' of an
enclosed ceramic CPU. But you'd be miles ahead of the game if you could.

 

[David Maynard]

A typical lousy off-hand response. I'm running a 2yr-old xp, not a
mobile barton. And I suspect that *most* installations do not conform to
your definition of a 'properly installed heatsink'.

rms

I'm pretty sure that what he means by 'properly installed' is the heatsink
sitting flat on the core and he would be correct.

 

[Phil Weldon]

"Ok, my gut feeling was that installing a copper shim was a mod that could
provide measurable benefits...."

Is this a shim to help keep the water block level or a shim to spread the
heat from the core to a larger area? If it is a copper plate to spread the
core heat to a larger area, what good is it unless your water block is not
copper, and then it should be 10 mm or so thick.

I think you have some question but not a definitive answer.

--
Phil Weldon, pweldonatmindjumpdotcom
For communication,
replace "at" with the 'at sign'
replace "mindjump" with "mindspring."
replace "dot" with "."

 

[Phil Weldon]

"... I'd hazard a guess that it conducts orders of magnitude less well than
the copper/aluminium of the heatsink that is in direct contact with the CPU
and therefore the quantity of heat that would take the
substrate>shim>heatsink path in favour of the easier direct path to the
heatsink is very very small."

The resin is probably more than THREE orders of magnitude less conductive,
and the quanity of heat taking that route is thus very very very very very
very very very small B^)

--
Phil Weldon, pweldonatmindjumpdotcom
For communication,
replace "at" with the 'at sign'
replace "mindjump" with "mindspring."
replace "dot" with "."