Improve the TIM joint?

[bigben2k]

Following what I posted in Cathar's "Cascade" thread here, I'd like to revisit the TIM joint question.

Specifically:

For an 80W CPU, the theoretical limit for a "perfect" waterblock is ~4.4C better than the White Water at 10lpm flow rates.

For example, for an 80W 100mm^2 heat source on the White Water with 10lpm flow rates, the total temperature rise is ~14.0C. Of that 14.0C, 8.0C is due to the thermal paste, 1.5C is due to the copper base-plate itself, and 0.1C is the amount that the water would be heated up by. Those are all values that can't be altered. So the 4.4C that remains is a measure of the efficiency of the block at 10.0lpm.

For the Cascade, that 4.4C is being brought down to around 2.5C, and it's taken a LOT of refinement using the most efficient known way to use water to cool (impingement), to get it that far.

I've come to see the Cascade as the final destination in my water-cooling journey.

Can we put our heads together, and see how we can improve this? Theories and wild ideas are welcome !

[Cossey3]

the wild idea first, chemically grow a layer of silver over the cpu to fill all the irregularities, then lap it until is is flat as possible. then dont bother with thermal paste. if the mounting is done properly the 2 surfaces should have decent contact. also use some standoff form the mobo so that the block sits better of the cpu and dont bother with springs on the mounting.

[bigben2k]

Originally posted by Cossey3
the wild idea first, chemically grow a layer of silver over the cpu to fill all the irregularities, then lap it until is is flat as possible. then dont bother with thermal paste. if the mounting is done properly the 2 surfaces should have decent contact. also use some standoff form the mobo so that the block sits better of the cpu and dont bother with springs on the mounting.

Sounds like a pretty wild idea. Is this feasible?

In another thread, also started by Cathar, we're discussing graphite, as a TIM material:
http://forums.procooling.com/vbb/sho...&threadid=7095

[jaydee]

What is there "realistically" and "reasonably" is there to do to gain anything from the TIM joint. People at amdmb are swearing by this shin etsu stuff saying it is up to 6C better than AS3. http://www.amdforums.com/showthread....hreadid=209018 But maybe it has some truth to it.

Anyway I don't see there being to many options especially being we are going to have to start working with heat/load spreaders in the near future anyway. Once those start becoming the norm we can directly cool the spreader itself. Should be plenty of room on it to seal it from the top of the spreader.

Graphite is not going to work simply because you need the molucules to go in the right direction. That would take premade sheets from a factory made to fit on a CPU. I don't see it working any other way. If you mix it into a paste form you can't line up the molucules right. Has to be in a solid 1 layer form designed to move the heat strait from the core to the HS. Therfore it has to be solid, and therefor it would probably shatter if put under the presure of a CPU as graphite is brittle? And then there is this cost issue. How many are going to pay $30 or so for a one off mount to gain maybe a C or 2? I know a few of you freeks will but I am past that point.

Anyway carry on. All I can get in now. Gotta get some work done here today at work.

[hara]

Water is by far one of the best TIMs. Try it out and you'll be amased. For TIMs, thermal conductivity doesn't matter much since the distance is nearly microscopic. What matters most is what material most adheres to the surface (thermally), deeply filling all gaps and which would be as thin as possible when compressed.

Forgot to mention that it dries up quickly. Try it only for a few minutes/hours.

[bigben2k]

You gotta love links to "TIM roundups" that don't clean the block and CPU with anything else than a piece of cloth, then congratulate the manufacturer for "ease of cleaning"! LOL!

Putting aside the incredible difficulty in measuring temps, I'd like to drop that temp gradient from 8 to 4 deg C. The problem I see is that there's some material between the die's temp diode, and the (clean) surface of the die, and specifically, what the temp gradient is between those points.

[Cossey3]

Originally posted by bigben2k
Sounds like a pretty wild idea. Is this feasible?

In another thread, also started by Cathar, we're discussing graphite, as a TIM material:
http://forums.procooling.com/vbb/sho...&threadid=7095

yes and no, silver is not very reactive so there are several reactions which will precipitate a very fine silver layer. one of these is used as a chemical test by most chemistry students at school. it deposits a very fine silver mirror efect on the inside of the test tube. if you did this over the core it would deposite the same silver layer. if done carefully and slowly it is possible to build up a very high purity layer. then you would need to carefully machine lap it flat. it would take a while and you would have to be careful not to get any impurities in the layer.

[bigben2k]

Originally posted by Cossey3
yes and no, silver is not very reactive so there are several reactions which will precipitate a very fine silver layer. one of these is used as a chemical test by most chemistry students at school. it deposits a very fine silver mirror efect on the inside of the test tube. if you did this over the core it would deposite the same silver layer. if done carefully and slowly it is possible to build up a very high purity layer. then you would need to carefully machine lap it flat. it would take a while and you would have to be careful not to get any impurities in the layer.

"Precipitate" a silver layer? Do you really believe that the result could produce a silver layer that's well bonded to the die?

We've discussed silver plating, for waterblocks, and came to the conclusion that electroplating silver to a copper base would yield little to no improvements, if any, so I don't see how precipitating silver on the die would be any better :shrug:

It's a wild idea, but I'd like to beat it up some more: there maybe something usefull to get out of it.

If the next step would involve very fine lapping of the silver, couldn't I just very fine lap my copper base instead? I could lap the CPU's die too, would that help?

From my understanding of the joint, there's only about a 5% area of contact, from lapped surfaces, without a TIM joint. The TIM joint helps fill in the crevices, and enables a better contact, but the heat must travel through the thermal paste.

Do you know of any way to improve the contact area, without using a thermal paste?

One thing I never quite understood, is if the thermal paste fills in those crevices, weren't they filled with air, before that? If so, then what happened to the air, once the HSF/WB is layed onto the CPU? If it's still there, isn't it still impeding performance?

[jaydee]

I guess you could attempt to mirror lap both the base of the block and the CPU core and mount then direct with no TIM joint. I actually laped a block to a nice polish and mounted it once forgetting to apply the thermal compound and it actually worked pretty good. I imaging if the base was professionaly laped "flat" and the core was lapped "flat" and they both where perfectly smooth it would work well. In fact I might try a half assed version of this again. I got plenty of spare CPU's laying around. I will try to get the core as flat as I can and the base of the block as flat as I can and polish them and then mount them with no compound. I will a do a before and after without compound and see what happens. I will do it on my ECS being it is setup for the Thermocoupler. This might be more realistic then anything else I can think of.

You know the little writting on the older AMD CPU dies and then the little AMD trade mark symbol on the newer AMD CPU dies adds a considerable gap when mounting the HS. I am always able to see the imprint of the symbole and writting. That needs to be removed.

[utabintarbo]

I suppose the ultimate TIM joint answer would be to precipitate or "grow" it between the die and waterblock. Theoretically possible, but impractical?

Bob
Specialist in crazy ideas

[bigben2k]

Originally posted by utabintarbo
I suppose the ultimate TIM joint answer would be to precipitate or "grow" it between the die and waterblock. Theoretically possible, but impractical?

How?

[jaydee]

Originally posted by utabintarbo
I suppose the ultimate TIM joint answer would be to precipitate or "grow" it between the die and waterblock. Theoretically possible, but impractical?

Bob
Specialist in crazy ideas

How long would it take to "grow" this though. Would your CPU be out of date by the time it was ready to use? What exactly would we be growing?


[not seroius]
Would you have to water it constantly and add fertilizer? j/k.

Maybe we can get AMD to make a waterblock and the die of the CPU all one peice. [/not serious]

Alright sorry, we can get back to being serious again.

[Cossey3]

the growing is a liquid that precipates the solid slowly so it builds up a layer than has no gaps in it. and ben with this reaction in chemistry we have to throw away the test tube half of the time because it is almost impossible to get off. this is not like plating the silver is built up molecule by molecule so the layer is a single metal crystal. if done properly it will be incredably durable

[utabintarbo]

Originally posted by utabintarbo
I suppose the ultimate TIM joint answer would be to precipitate or "grow" it between the die and waterblock. Theoretically possible, but impractical?

Bob
Specialist in crazy ideas

Originally posted by bigben2k
How?

a la "plating" them together.....????:shrug:

Bob
First Sub-God of Silly Thoughts

[Cossey3]

oh yeah ben to make lapping really worthwhile both sides need to be lapped to maximise contact area.

[utabintarbo]

Originally posted by Cossey3
the growing is a liquid that precipates the solid slowly so it builds up a layer than has no gaps in it. and ben with this reaction in chemistry we have to throw away the test tube half of the time because it is almost impossible to get off. this is not like plating the silver is built up molecule by molecule so the layer is a single metal crystal. if done properly it will be incredably durable

Yeah! What he said!

[bigben2k]

Ok, I think I'm beginning to understand the "precipitating" process a little better.

So if we go that route, and let's explore it, because I think that there's something to it, do we:
1-(Cossey3) "precipitate" on one surface (CPU) or both the CPU and the waterblock? Let's stick to AMD, for now.

or

2-(Utabintarbo) Would it be possible to "precipitate" the silver in a mounted assembly?


What chemicals would one need to do this?

[murray13]

Actually the best TIM joint, NONE! If we could get Intel and AMD to integrate a WB as part of the heat spreader the TIM joint would be a thing of the past.

BUT how soon do you think they will do that?

So has anyone ever tried or seen someone use this as a TIM? I've been thinking of asking for a sample but have never gotten around to doing it.

[bigben2k]

Last I checked, there's a TIM between the Intel die, and the IHS, no?

[logosmani]

thermal conductivity of mercury is what now?, perhaps an idea for a decent tim?

[bigben2k]

Originally posted by logosmani
thermal conductivity of mercury is what now?, perhaps an idea for a decent tim?

No. (You must come from OC ). It's toxic, and that makes it impractical.

I really don't want to get into the Intel IHS either. If ya'll have one, just pop it off (at your own risk), so that we can resume this discussion.

[Groth]

originally posted by murray13
So has anyone ever tried or seen someone use this as a TIM?

Based on their stated '0.015ºCin2/W or less thermal resistance' and a die size of 100 sq. mm, you'd have a TIM thermal impedance of ~0.1 °C/W. About the same as well applied thermal paste.

Edit: Don't type so fast you moron!

[logosmani]

maybe something to think about however, perhaps there is a way to use a liquid form tim in some manner + contain it _safely_

[bigben2k]

Originally posted by logosmani
maybe something to think about however, perhaps there is a way to use a liquid form tim in some manner + contain it _safely_

Hum... possible. The advantage with the paste (or solid, aka tape) over a liquid is that it won't leak, and since the CPU usually sits vertically, that's an issue. Maybe containing it, as you suggest, with silicone, or RTV?

[jaydee]

Originally posted by bigben2k No. (You must come from OC ). It's toxic, and that makes it impractical.

LOL. I am going to remember you said that next time I get blasted for it.

I really don't want to get into the Intel IHS either. If ya'll have one, just pop it off (at your own risk), so that we can resume this discussion.

Problem. By the time anything comes of this thread we will all be using AMD Athlon 64's/Optrons/Hammers with IHS's on them. From the looks of the pics I am not so sure they will just pop off either. But carry on. Good stuff so far. Will test the core and block lapping starting tonight after I get home.