Interesting observation with CPU cooling

I have a couple of CF-30's (K and F models) and noticed that when they are sitting on my lap (bare legs) they get scorching hot but only in specific spots along the bottom plate. Figuring that there might be a way to slightly improve the heat transfer so the entire bottom plate heats up and CPU temperature decreases I started to mess around with the heat transfer compound on the bottom plate where it attaches to the CPU chip.

There is a greyish, crumbly brick of thermal compound that will either stick to the plate or the chip, (or sometimes to both). It is fairly thick, even in it's compressed state. I set up a little test to find out what the normal CPU temperature was before altering things;

Now it is warm in my house (I do not run the air conditioning) so right now the temperature is 28.3 C (83 F). To put a consistent load on the CPU I am running BONIC with two projects running on a Windows 7 OS at 100% CPU utilization. The front of the laptop is elevated 4" above a flat surface so there can be a little bit of air flow under the computer. There is a fan moving the air around in the room.

After 2 hours of letting temperatures stabilize the CPU temperature was 76 C. Running my hand along the bottom of the case there are hot-spots that would be uncomfortable against bare skin for more than a few minutes.

I then removed the bottom plate and scraped off the thermal compound and set it aside. Wiped down the mating area on the heat sink and chip with a cotton swab and a little bit of acetone.

Then I applied thermal interface pads (Dell PN UP755). To fill the gap it took ten pads stacked so the CPU chip and the thermal plate were in contact when the two screws on the center of the bottom plate was tightened down.

I had a use for the old thermal compound; I unscrewed the two heat pipe clips at the ends, the one bent pipe with the clamp at the middle and the entire heat pipe assembly. Under the clamp areas you can see where the heat pipes are supposed to come in contact with the magnesium alloy bottom plate. I rolled the left-over compound out, like you would with pizza dough, until it formed a small stick about six inches long. I broke that in to three pieces and put the material between the bottom plate mating surface and the heat pipe. Each of the three pieces were about 2" long.

Bolting it all together I fired up the laptop and re-started BONIC and "open hardware monitor" to watch the temperatures.

At first it did not look good; the temperature jumped up to 84 C and over the next hour it slowly climbed to 100.0 C. Right when it hit 100.0 C something happened with the compound; the temperature very quickly fell to 62 C where it remains right now.

The entire bottom plate is pretty uniform in temperature, no hot-spots. The resulting temperature is 14 C than what it was before I started to change things around.

1). The gap between the CPU chip and the heat sink is ridiculous. I am wondering if I shim up the heat sink it will close the distance to the chip. I have plastigauge so I can measure the distances and determine how much shimming I would need to do on the heat sink to close that gap up. Decreasing the amount of thermal compound will make cooling a bit more efficient.

2) Adding a heat sink compound to the ends of the thermal pipe really does improve the heat transfer characteristics to the bottom plate. If you are going to pull the bottom plate off for any reason that can be a five minute fix with some thermal compound, even if it is not the expensive stuff (like Arctic Silver).

3) There is some information out there that thermal compounds go through a state change when they are temperature cycled. I will be interested to see if 62 C is the new "normal" or if it will lower the temperature by a few more degrees.

I did this not because I am overclocking the machine but I do some CPU intensive things that will make it run hot over a several hour period (radio propagation analysis stuff). I do not want to burn my lap.

So,
From: 76 C
Before curing or "set" 100.0 C
After curing or "set" 62 C

Materials used: Thermal transfer pads and reusing the old compound to improve heat transfer elsewhere

 

There have been other posts about thermal pads and about putting material under the "arms"
I have never heard about the temperature change after the thermal pad cycles.
Great info.

Here is a copy of an old post of mine. The CF 29 and Cf 30 are very similar. Here is a link to the entire thread. http://forum.notebookreview.com/threads/toughbook-thermal-pads.523950/


Actually the cf29 uses a putty that comes in a sheet form. I opened a parts cf29 and I took the thermal material off and rolled it into a ball.
Fujipoly is what I came up with to use.

Fujipoly

150-XPE
150 means 1.5mm XPE is the material.
The XPE has a thermal conductivity of 11 W/mK. Which from what I found out is really high performance. Thermal conductivity of 6 W/mK is considered high performance.
Here is a link to one source of the 6 W/mK. The 11 W/mK is a bit harder to acquire.

Fujipoly_Premium_System_Builder_Thermal_Pad Frozencpu.com

Premium Thermal Pads FrozenCPU.com

Here is another post of mine with information and links.
http://forum.notebookreview.com/thr...ust-bought-a-cf-19.763310/page-3#post-9804027


Fujipoly / ModRight Ultra Extreme System Builder Thermal Pad Blister Pack - 60 x 50 x 0.5 - Thermal Conductivity 17.0 W/mK

http://www.frozencpu.com/products/1...rmal_Conductivity_170_WmK.html?tl=g8c487s1797

I think 1.5mm is correct for a CF 30. I know it's correct for a CF 29.
Also get some 0.5mm thickness and put it under the arms of the heat sink......

 

Have you thought of using a copper shim between the heat sink and CPU? I have also thought myself about finding another heatsink with more pipes to try out.

 

That is exactly what I am thinking.

This morning I repeated the same process with the second CF-30 that is the K model (the first was the F model) and I think I can add a copper shim between the surfaces. I have a whole bunch of 4" wide copper strap (about 200 linear feet of it) that is used for grounding at radio communications sites. Since I have another project going on at the same time ( a lightning protector for a HF vertical antenna ) that needs me to snip up some copper strap this might be a good reason to do both at once.

I let the first machine bake in last night until around midnight, the temperature remained stable at 62 C. This morning I started up the first machine and at this moment it is 64 C. I think that the excessive gap between the two parts and the relatively low thermal conductivity of the compound (when compared to copper and a thin layer of something like Arctic MX-4) can improve cooling.

Other than the silly distances between the heat sink and the CPU chip Panasonic did an excellent job in designing the Toughbooks. I have found little that I could criticize.

I do not not want to get obsessed over cooling; I went in to this with a specific goal of not burning my lap, but if I can slightly lower temperatures on the CPU die that is OK too.

 

Insert a extra finned heatsink or a micro//nano fan.

 

Something interesting about the heat pipe block.

On the F model laptop it is an aluminum casting, just like your picture shows.

On the K model laptop it is a stamped steel piece with little key-guides.

I noticed it this morning but did not comment upon it. It seems that Panasonic made a bunch of really small changes to certain parts in the K model.

 

If I recall, they are a different height also.

mini heatsink

 

the part i am hesitant about (in this) is the flex of the MB and case bottom .
i bet that the panasonic boys were not saki-soaked when they designed that grand canyon gap between the two .
while most of us here are cautious in our handling of equipment ...
there is always the chance that some well-meaning but brain dead cretin (read ground crew grunts) decides to be helpful and drops the '30 smack dab on a 9/16 stainless steel flange bolt .

i agree that steel is not the best when it comes to thermal conductivity .
perhaps lapping the mating/contact surfaces of the (shown) aluminium holder/coupler is in order .

rf strapping ...
yeah ... i have ripped out many feet of wire (some of it aluminum !) at sites .
in some instances i have drilled holes behind a transmitter (through the floor) and drove ground rods ...
a few times (with tight installs) i took long ground rods to a friend and had him cut/machine/braze a fitting on the ends so i could screw them together as i drove them down .
(i had a oxy/acet torch to braze them after tightening) .
i guess they never heard of the "surge impedance" of *something* when excited by a pulse with a bandwidth of dc-to-daylight .
that said (and medications to help ease the horror-visions) ...
the small wrinkles and such can be minimized by due application of flat surfaces and a #2 ball-peen hammer .
this is a game of "small returns" ... each item in the chain needs to perform as best as is possible .

 

CWB32.
I agree with the copper shims//flex issue that you speak of. I knew there was a reason we discussed long ago as to why not use solid shims. I forgot what I knew. You reminded me.
I still think that a small heatsink or fan "inside" the standard heatsink is the way to mod the cooling.
I am working on a fan that would go where the wwan sled is. I don't own a CF30 anymore so that project is just for the challenge.

 

I was the latest one that brought up the idea of attempting to use copper shims in my CF-29 some time ago, and I was told it was already discussed before that. We decided against it for this reason. Simply put, the thick thermal pad also protects the CPU/GPU from damage caused by external shocks to the bottom of the case.

 

Great thread, I've said several times that I think something is not right with the
CF51 MK3 cooling and I just completely tore one down wondering if there are
any improvements I could make.
It gets VERY hot on the bottom especially in the corner of the RAM cover near
the processor, what do you know, the Northbridge chip is there and there is no
heat sink at all on it. But there is about a 1/4" thick "pillow" on the Southbridge.
Usually the Southbridge does not have a heatsink and the Northbridge does. I
believe that the pillow should be on the Northbridge. There are some other
very odd things about it also.

I'm going to start a thread on this with pictures.

The Thinkpads from the same era, have the Northbridge share a heatpipe
with the GPU, and when there is no GPU the heat pipe is just for the Northbridge.
There is nothing on the Southbridge. The data rates through the Northbridge are
VERY high going to the RAM and PCIe ports, and the power is proportional to
switching speeds.

I've had many similar thoughts to yours over the years thinking about why it runs so hot.

 

I bought an assortment of copper shims for CPU/GPU's off of eBay. I spent around $10 for enough to do at least a dozen CPU's (just the way they were sold).

How I measured to find out what the appropriate shim thickness needed to be;

1. On a laptop that has the caulk-like grey sticky stuff I shut it down and let it completely cool to room temperature
2. Pulled the bottom plate and carefully peeled the grey sticky stuff off in one solid piece.
3. Put the grey stuff flat on something, used a razor knife to cut it cleanly in half
4. Went through the assortment of shims to find one that was closest to the thickness of the grey sticky stuff.
5. Cleaned and prepped the top of the CPU chip and the heat sink surface with a cotton rag with a touch of methanol until they were both shiny.
6. Applied five dots of "Arctic Alumina" to the top of the CPU chip.
7. Placed the copper spacer on to the CPU chip, gave it a slight twist to spread the heat sink compound around.
8. Applied five dots of Arctic Alumina to the top of the copper spacer
9. Removed the heat pipe and put dots of arctic alumina where the aluminum block bolts to the base plate and where the heat pipe comes in intimate contact with the backplate (near the middle of the bend on one leg, near the ends on both legs).
10. Checked and tightened the hardware for the heat sink block and the two heat pipe tubes that are attached to the bottom plate with screws in six places.
11. Put the base plate back on and lifted it off again quickly to be sure that the heat sink compound was making contact between the copper shim and the aluminum heat transfer block.
12. Put the plate back on, this time bolting it down, especially tight on the two middle screws that hold the heat sink down tight against the CPU.

It seems that the computer CPU temperature is much more responsive to rapid cooling changes on the bottom of the base plate. I am going to let it thermal cycle on and off for a few days to see what the numbers look like. There aren't any hot spots on the bottom any more, the base plate just feels evenly warm.

I know that one of the concerns about reducing the spacing is that if the bottom of the baseplate gets bashed really hard that the blow would go right up in to the CPU and maybe damage the main board. I looked at how the heat sink is mounted to the chassis when it is sandwiched between that and the bottom plate. There is not much mechanical room as the heat sink rests on a set of two posts where the center bottom plate comes through. Unless someone does something really extreme, on par with a blow that would crack the chassis.. the reduction in spacing is not going to hurt anything.

Having copper in the middle instead of grey goo is a much better heat sink. the thermal transfer characteristics of copper is about 50 times greater than the best numbers you will get for any thermal compound. So, copper to CPU is the best, all any thermal compound is going to do is to smooth out any surface irregularities on what should be very flat and polished metal surfaces.

 

To make sure the added copper stays in place have you considered on one side only using the thermal paste that is also a glue? That way it's no way it can get dropped hard and come off?

 

That is something to consider; I would probably bond between the chip and the copper shim. The size is 15 mm X 15 mm X 1.5 mm thick....

The more I think about it, the more I like the idea, thanks much.

 

I would glue it to the base. That way if you ever decide or need to remove the shim, it's as easy as buying another base plate.
If it is glued to the cpu it is pretty much there forevermore.

 

I'm having the same thoughts to glue a copper shim as a heat spreader on
the CF-51 Northbridge. Bought the glue many months ago.

 

That makes sense but isn't the thermal paste designed to work under pressure with
as little material remaining as possible. I would think that the largest heat spreader
glued on to the processor would be best, with very thin thermal interface pads on
the otherside to provide pressure and conduct across the gap.

 

for the best ability to transfer heat , the metal surfaces should be "lapped" for the best mechanical contact .
lapping each piece in a metal assemble with its mating surface is the best bet .
think about how an engine is "blueprinted" ... a bit laborious but the results are worth it .

i have seen metal surfaces lapped so well that a single drop of water put on one surface and the other surface then put in contact with it ...
i could not pull them apart .
i think the demonstration devices were about 3 square inches .
a friend did some lapping of his cpu and heatsink ... he had trouble getting them apart when they were dry !

a thought ...
if the surfaces are that "good" ... could one simply use a drop of a good grade of ATF or other oil to take the place of heatsink compound ?
the reasoning is that the zinc oxide or silver used would have to be finer that the spacing/gaps of the surfaces in order to keep from lowering the conductivity .

 

http://www.ebay.com/itm/SAMSUNG-830...179955?hash=item4b06d52833:g:DGIAAOSw65FXu27D
5 days 10hr running time ...
i am buying one .
there are a couple left .

 

mistake ... thought i was on another page ... please remove this and my previous post .
(although it is a good buy on a ssd)

 

I can move it if you like.just let me know where you want it...but you are right that is a very good price. Will not last long for sure

 

ehhh ... that was actually a private message to shawn .
i have sent him the info .
there is one left .

 

Anyone know of the best thermal pad?
Fujipoly has ones that are 17W/mk, 2 to 3 times better than typical.
anything better?

 

That's the best I know of.

 

To the OP, what program are you using for temperature measurements?
Anything for fan speeds?

 

I would give Shawn "rep" points but alas the "system" won't let me.

I would be most happy to spread it around....if we were getting answers.

This is probably not worth the hassle but I'm old enough now.....to not care.

IOW I give my points to whom I chose.

 

Here is how GETAC beefed up the cooling without a fan...Copied from this link

http://www.ruggedpcreview.com/3_notebooks_getac_b300.html

Below you can see the original B300's thermal design solution (left), and the significantly beefed-up solution of the new B300 (right) that allowed operating a 2GHz Core i7 processor, and now an even more powerful 2nd generation 2.3GHz Core i7 CPU, without fan. That is remarkable.

 

i'm guessing that the original getac design was on the marginal side to start with .
usually , with successive "generations" or drops in photolithographic/architectural sizes , there is a drop in oranges-for-oranges power consumption .
even allowing for the increase in speed (2.0 versus 2.3 GHz) the increase in power consumption would be offset by virtue of the smaller construction techniques .
of course , this depends on the exact processor(s) used ... and there are enough of them to make a lawyer giddy with excitement :
https://en.wikipedia.org/wiki/List_of_Intel_Core_i7_microprocessors

ps ... slightly aside from the gist of this thread ... did you ever notice that somehow or another that a desktop with the supposedly same processor as a laptop seemed to respond quicker ?
it may not just be your imagination ...
a close look at exactly what cpu is used in what may reveal a bit of what i would call "advertising sleight of hand" .

as i write this , i am using a desktop comp that uses supposedly same processor as one of my toughbooks (in the broad sense of that statement) ... sure seems faster at performing tasks
barring the ol' "throttled back for your protection" trick , methinks there is some hocus-pocus going on here .