Copper Spacer?

[Guru]

I have thought of bilding a copper spacer for my FC-PGA P3-700@980 1.65V!
What size should it bee (No jokes please)?

------------------
Join the MURC SETI team! | SETI @ MURC

[ALBPM]

Most that I've seen for sale on the net are 40 X 40MM square outside demensions. The cut out is generous a bit larger than the chip and I thought that somewhere I read that they are ~0.5mm thick.
No one seems to give a thickness for their shims. I imagine that some hand lapping is required for the perfect fit. You want to make 100% contact to the chip and most shims are a bit thinner mainly for support to keep the heatsink from tilting and flat on the processor.

Paul

[Guru]

Thanx for the info!

------------------
Join the MURC SETI team! | SETI @ MURC

[xortam]

First I heard of using copper shims. What is being accomplished by this? If its just thermal transfer from the chip to the HS, I would think that silver compound is superior.

[Holerith]

Silver is a better electrical conductor but I'm not sure about thermally. Beryllium Oxide would be excellent but it is *VERY* toxic!

------------------
Dean
-----------------------------------------------------------------------------
PDP-11, Dec-writer & ZD-11 Terminal Unit, RSTS-OS

[DuRaNgO]

the shim is to keep the heatsink perfectly flat on the slug (wider surface)
jim

[ALBPM]

Now are we talking shims like this

These are very thin and really aren't intended for heat transfer as much as to keep the heatsink from rocking around and chiping the Coppermines.


Or, is the spacer your talking about a solid
"Cold Plate" to put between the processor and heatsink. If you want a cold plate the copper should be a minimum of 1/8' thick, preferably 1/4" is better.

This would act to draw more heat from the processor and transfer it to the heatsink. That's why Alpha bonds the copper plate to the bottom of their aluminum heatsinks.
If this is what your doing than the dimensions can be greater than 40 X 40mm. this all depends on the size of your heatsink base.

Paul

[This message has been edited by ALBPM (edited 30 September 2000).]

[Greebe]

xortam, solid copper is superior to all granular metal in conducting thermal energy. Altho silver is slightly better than copper at conduction, it's not by much.

Copper shims were first desired to help conduct heat out of the cpu. This they do a great job at. The second reason for having them was due to the tilt factor. But if tilt was the only reason then someone smart would use just plastic.

Holerith, Your darn tootin there! Beryllium Oxide is a great thermal conductor, but that deadly dust just won't cut it for plain folk.
We had a TWT break in our recycle bin once and had to have hazmat do a clean up. What a pain in the arse! Scrub down head to toe, every nook and cranny til your raw! Not fun (and glad for once to be working outside on a plane when it happened

[xortam]

Thanks for the replies. After looking at the thermal transfer article that Paul pointed out, it confirms my understanding that you want to minimize the number of interfaces between the conducting materials. Ideally, it would seem that you want to get the CPU slug and solid metal HS to exactly mate and be held together via a bracket and/or bonding material to maintain the full contact. Lacking a perfect mate between the surfaces and dealing with the porousness of the subject materials, use of a highly conductive interface material such as silver thermal compound makes sense. Regardless, I would think the HS could be secured to the CPU with a bead of epoxy surrounding the diameter of the slug (or better, a secure removable adhesive). Wide, heavier HS's may need support stands to keep them from stressing the slug. Are the copper shims just such a bracket and actually don't fill the space between the CPU and HS (they surround the chip)? I'd be surprised if they were effective if they were used as an interface material between CPU and HS along with a compound. It seems the best affordable solution is to mate a solid copper HS to the CPU if it could be done securely. Is there such a beast? The Hedgehog apparently isn't a solid piece.
Let me simply ask, ignoring stability, wouldn't a high quality aluminum HS, properly mated and using a silver thermal compound be superior to using the same HS with a intermediary copper spacer and compound? Perhaps a micron thin copper foil (or better, silver) alone between the HS and CPU would be superior!?

[Holerith]

Greebe, yep powdered BeO2 is definitely something you don't want to inhale or ingest. Or even get too much on the skin.

I encountered it in doing radio work as some Eimac high power ceramic-metal vacuum tubes use a BeO2 plate as a heat transfer pipe between the plate fins and the heatsink.

It's is a royal pain to clean up if you crack a plate.

------------------
Dean
-----------------------------------------------------------------------------
PDP-11, Dec-writer & ZD-11 Terminal Unit, RSTS-OS

[Greebe]

xortam, simple answer, nope. Biggest problem with this conjecture is we're only talking about "cost effective" solutions. If spending $200+ for a cooling solution then many of these normally cost prohibitive designs / materials could be used.

Problem with epoxies is they exhibit much poorer thermal conductivity than normal HS compounds. The cpu "slug" directly made to a copper HS (electrolysis possibly after the cpu is finished?) would be the best for air cooling. We could use silver as the HS material, but it's not much better based on expenditure.
Artic Silver is the best HS compound I've seen to date. Highly recommended but not a cure all when larger gaps are concerned.

Here the best point to make as for copper being the best material of choice based on cost/thermal conductivity (lowest to highest)

Thermal Conductivity
Aluminum 2.18 (pure)
Copper 3.94 (pure)
Silver 4.08 (pure)

From this you can see how little of a difference there is between copper and silver, but yet undeniably see why copper is the material of choice.

Copper spacers do remove alot of additional thermal energy due to delta/T. Where basically it calls for a higher thermal gradient towards the corners. Thus adding the spacer will improve the overall sink area.
(yes, I know, a very rough cut, well bite me

I have used sterling silver spacers on my Athlon to get the proformance levels I've achieved (70% boost over baseline clock speed). These where provided by ALBPM as Athlon L2 spacers but were then fashioned to fit my needs.

If water cooling was finely integrated into the case by the manufacture, then this would be by far the most cost effective way to cool our HOT cpu's. Peltier assisted or not.

Speaking of which, Univ. of Michigan research has produced an enhanced forumula for making Peltier devices and licensed it to Tellurex Corp. Preliminary info suggests these new devices will be capable of thermal gradients of upto 200c compared to only 70c of those currently in use. (remember this one Jord?)

All I can say is WooHoo send me some!

[xortam]

Greebe ... don't like my use of the term slug eh? I would normally have said substrate but I was trying to employ the commonly used slang found on MURC. (personally, it irks me every time I see the term 'puter).

I understand that silver isn't cost effective for its marginal conductivity over copper, but can we agree that its still worth considering when very little material is needed (e.g. foil or compound or even spacers).

Can we agree that the term shim is used to describe a square washer that surrounds the chip and does not come between the chip and HS surfaces?! The shim would only lend to support the HS as the only contact with the CPU is the thin edge surrounding the circumference. Also, lets agree that the term spacer is used to indicate a thin (1/16-1/4") sheet of material which is placed directly between the CPU and HS. The use of a bead of adhesive would only insulate the outer top edge of the chip and the contact with the HS. Perhaps that's sufficient of a penalty where a shim is superior, even with the use of a compound. Certainly makes removal easier.

Can you clear up some of your points? Is the term delta/T used to indicate a factor in energy flux? Where thermal conductivity is defined as "the proportionality constant that when multiplied with the ratio of the temperature change to the change in distance from the zero plane, will result in the value of the energy flux (Fourier's Law)." (Thermal conductivity is rated in W/mK. What are these units?) If my interpretation of delta/T is correct, I take it that a larger delta/T is the same as when you say "higher thermal gradient towards the corners." Mr. Larsen (author of the referenced article) stated that "Aluminum will have a higher temperature difference between the point of contact and the heatsink extremities.": Therefor, it could be said that aluminum has a higher delta/T than copper. Aren't you relying on the higher thermal conductivity of copper to pass the energy through to the HS and not it actually having a lower thermal gradient?

The point of my previous post was that not only do you have to be careful in selecting cost effective materials but you need to minimize the number of interfaces to minimize the inherent contact thermal resistance. The thermal conduction benefit of using a copper spacer has to be countered with the penalty of adding an additional interface. There must be a minimum height of the copper spacer to overcome the penalty and yield a net benefit. The big question is ... what are the minimal and optimal dimensions of the spacer when considering cost benefit? If your using a pure copper HS, it seems you would be better of not using a copper spacer.

[Greebe]

Xortam, I didn't have a problem what so ever with your use of the term slug. Yes substrate is correct but heck these are the forums.

delta/T has to do with thermal losses within a medium. Not with the medium itself. HS are based on this and other factors like surface area, mass, dimensional characteristics etc. Within a cpu heat spreads outward from the substrate and the edges of it's casing is why the cpu spacer improves energy dissipation laterally.

Copper cpu spacers only encompass the slug. Typical thickness of these are 0.5 mm. Actually it's the thickness of the slugs height, subtracting the additional thickness of a properly applied thermal compound. There are two big problems with using these spacers tho. One is being with it's thickness and the other being with it's electrical conductivity. The first, if the spacer is even the slightly higher than the slug then a gap would form between the heatsink base and substrate. A minor gap of even 0.001 would be enough to either render the HS virtualy useless and or destroy the cpu.

Using epoxy to bond the outer edge would be a manufacturing nightmere. Even if done by hand is a waste. One of the important aspects to consider when mentioning this is thermal expansion. Sealing a liquid within a critical gap and not taking this into consideration would be a major mistake. If any were to ever be force out by any means our gap problem would seriously impair it's proformance. I see no reason to do it.

Copper foil is worthless due to it inabiltiy to spread the heat longitudinally (delta/T again). A copper plate between the cpu slug and a copper HS is of very little use and may impair dissipation due to the additional interface surface. But on an aluminum HS this extra contact surface is negligible due to the improved heat spread across it's base.

I colaborated with Gabrael Rouchon over at Swiftech on many ocassions before he introduced his own variety of spacers or water cooling setups. Discussing thermal requirements to get optimum cooling, common findings of OC'd Athlon cpu's, to selling my Eclipse gfd through him exclusively.

One of our common finding was that all cpu's made today have that extra glop of adhesive surrounding the slug. Even tho thin, this is enough in many cases to cause just this kind of gap induced failure.
Also in the case of Duron/Tbirds where the electrical connections on top could be shorted by the spacer. In this case ALBPM informed me that Gabe no longer is selling spacers due to damaged customer cpu's. I haven't talked to him in a few weeks but think this why he pulled them.


------------------
Designer of the first commercial Athlon mod.

beta system
3000 mips, 1.172 gflops @ 900 mHz 1:3 latency w/ active cooling
CPUmark99 75
FPU Winmark 4890

G400Max
3Dmark 99 Max (8x6 16, 16, triple) PD5.41, Athlon@850mHz
8000
15234

now running a G450

[Brian R.]

Anyone know the ranges of thermal conductivities of sterling silver alloys? Since they are alloys, I'll bet pure copper is superior in general. Sterling alloys are only 92.5% silver or better.

[This message has been edited by Brian R. (edited 03 October 2000).]

[Himself]

Bit unrelated, but I have a chrome orb cooler for my Duron, it has 2 pads and in the center is a square patch with a gummy substance under it, I removed the sticker but I am wondering what the patch is for besides protecting the core and whether it conducts heat properly.