Improve the TIM joint?

[logosmani]

don't see why it wouldn't work, but it would probably _have_ to be mounted vertically in order to stop air forming in the wrong place, presuming of course that u can't actually remove all the air b4 sealing it.

real question is of course, would there be any point to it + what risks would any liquid used in this fashion pose to the die.

[Cossey3]

ammonium hydroxide silver nitrate and any aldehyde
and it would be almost impossible to use this method to bond the block to the cpu. what happens is the silver nitrate is reduced to silver so it comes out of solution and forms a very thin film on all the surfaces. if a concetrated solution was used it would make a think enough layer to fill all the gaps then lap it back to get to a flat surface.

[murray13]

Originally posted by Groth
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.

Actually on their data page they list thermal impedance (in metric) as "0.006-0.013 °C-cm2/W". Using those numbers and the 80W example above this TIM joint would be between .48 and 1.04 deg C. Much better than well applied grease, by 8x. If I'm overlooking something please correct me.

BB2K: Ok, integrate the WB into the silicon directly. I think I read somewhere that Agilent (used to be part of HP) is looking into doing that using inkjet technology.

[nicozeg]

If water is the best short time TIM material, we only need to find another liquid that don't evaporate or degrade over time.

That's easy: Synthetic motor oil

As a test I mounted my GPU heatsink about a month ago with a drop of oil, and it's working OK. The only purpose of this was to see long term stability, as I dont have a way to know GPU temps, and didn't want to risk my CPU on the first try.

Maybe it's time to go for the CPU, but It's not going to be very conclusive, as my mobo is very crappy with temp readings. :shrug:

Oil running away because the vertical position is not a real issue, surface tension make that impossible given the microscopic distance between surfaces.

[Cossey3]

if we could get the silver thing to work it would give a tim temp gradient of alot less than <0.1oC. so with one of cathars blocks it should give thems of 6oC above water. so 31oC baised on 25oC water wouldnt be bad

[BO(V)BZ]

Well, I just put in a request for a sample of the Thermagon stuff. I have no idea whether they will actually send me anything, but if they do, I'll tell you guys about the results.

BO(V)BZ

[Groth]

Originally posted by murray13
on their data page they list thermal impedance (in metric) as "0.006-0.013 °C-cm2/W".

Hmmm. They also say 18 W/mK and 0.148 mm thick. So, 0.0148 cm divided by 0.18 W/cmK is 0.082 °C(sq. cm)/W or 0.0127 °C(sq. in)/W.

18 W/mK, 0.148 mm, and 0.01-0.02 °C(sq. in)/W are at least mutually consistant - I've greater faith in them.

[bigben2k]

Originally posted by Cossey3
ammonium hydroxide, silver nitrate, and any aldehyde
and it would be almost impossible to use this method to bond the block to the cpu. what happens is the silver nitrate is reduced to silver so it comes out of solution and forms a very thin film on all the surfaces. if a concetrated solution was used it would make a think enough layer to fill all the gaps then lap it back to get to a flat surface.

Sounds feasible. Could I ask you to outline a procedure?

Originally posted by Cossey3
if we could get the silver thing to work it would give a tim temp gradient of alot less than <0.1oC. so with one of cathars blocks it should give thems of 6oC above water. so 31oC baised on 25oC water wouldnt be bad

Ok, but I'm still not convinced that we can improve much on the default 5% contact area. We get 5% with lapping, so how would we get more here?

If it's any help, Bill and N8 discussed the aspects of smoothness and flatness, because they really go together to make a good joint.

[Boli]

Silver powder? bound with some sort of adhesive/jelly mould stuff. you mix it together spread it all over your CPU shove the Water block on and wait for it to "set" once (and indeed IF you can taker it off you'll have the perfect imprint of your CPU die and waterblock).

I have no idea how you could do this, but another idea in the mix doesn't sound like a bad idea. .

~ Boli

[hara]

Originally posted by Boli
Silver powder? bound with some sort of adhesive/jelly mould stuff. you mix it together spread it all over your CPU shove the Water block on and wait for it to "set" once (and indeed IF you can taker it off you'll have the perfect imprint of your CPU die and waterblock).

I have no idea how you could do this, but another idea in the mix doesn't sound like a bad idea. .

~ Boli

AS3 epoxy?

[jaydee]

Originally posted by hara
AS3 epoxy?

You mean this: http://www.arcticsilver.com/arctic_s...l_adhesive.htm

[satanicoo]

Volenti should know if the P4 IHS has termal past, he removed it for his on-die wb.

[freeloadingbum]

Like hara, I tried water as a TIM with positive results. After 4 remounts the water was 2 to 3C better than AS3 in all cases. This was using a palamino@1833mhz/1.95volts

Considering the hazards of water, I thought of using Fluorinert as a substitute. Some of the variations have a low viscosity and will not evaporate untill it's boiling point is reached.

One of the potential problems of a low viscosity liquid is it may be more prone to "pumping" compared to pastes and may not last for long periods. This could easily be solved by flooding the entire area around the die, using an "O" ring or a thick non-conductive grease for a border. This may also transfer some secondary heat off the cpu package to the block.

[satanicoo]

Still this makes me think that direct-die can be much, much better then normal wb.

There's always the problem of it lasting litle, i know...

[jaydee]

Originally posted by satanicoo
Still this makes me think that direct-die can be much, much better then normal wb.

There's always the problem of it lasting litle, i know...

If I were to do it I would do it like mentioned in the Cascade thread. Drill holes through the base of the block and make jets aimed right at the holes so the inpengment is right on the core of the CPU Die itself. That would rule ass, but I am not sure the base would hold up under pressure without collapsing as the base would have to be pretty thin to get the maximum impengment effect. With holes in a thin base with 25lbs or so of mounting pressure you can do the math.... Then there is that sealing issue. If they didn't put any of those blasted little electronic parts ontop of the CPU you could get around the colapsing issue by making a foundation on the top of the CPU. Bah! To much work....

[bigben2k]

Originally posted by jaydee116
If I were to do it I would do it like mentioned in the Cascade thread. Drill holes through the base of the block and make jets aimed right at the holes so the inpengment is right on the core of the CPU Die itself. That would rule ass, but I am not sure the base would hold up under pressure without collapsing as the base would have to be pretty thin to get the maximum impengment effect. With holes in a thin base with 25lbs or so of mounting pressure you can do the math.... Then there is that sealing issue. If they didn't put any of those blasted little electronic parts ontop of the CPU you could get around the colapsing issue by making a foundation on the top of the CPU. Bah! To much work....

Actually, the baseplate in Cascade can be just about anything: it won't affect the first nor the second inpingement. Personally, I thought about using thin blades, crossed over the pattern, to reinforce the support. Heck, even SS blades would work well.

The main concern though, is if the inpingement effect on a direct die works, what kind of flow rates would be required? I mean, would there be a minimum flow rate where the CPU becomes endangered?

In any case, there would still be a need for a TIM, for the remaining copper, and that's the topic here

[bigben2k]

A link of interest:
http://www.bergquistcompany.com/thermal_materials.cfm

[sevisehda]

So if one was to try and chemically build a layer of silver onto there die then why not build up about 1/4 of an inch then just mill the block straight from the silver.

[bigben2k]

Originally posted by sevisehda
So if one was to try and chemically build a layer of silver onto there die then why not build up about 1/4 of an inch then just mill the block straight from the silver.

There might be an issue with applying a tool to the plated die: the vibration might destroy it.

Of course by putting shapes into place, we might be able to build a Cascade style baseplate, right on the die...


But I don't think that the plating (actually "precipitating") process lends itself to building up such a thick layer. Cossey3?

[logosmani]

Originally posted by bigben2k
A link of interest:
http://www.bergquistcompany.com/thermal_materials.cfm

nice one, being looking for UK suppliers of Gap fillers for ages :]

[jaydee]

I am confused as usual. Whats the point of growing this silver layer again? Seems to me you would have to lap the silver layer down and then lap the block's base down to a perfect flatness to use it effectivly. If this is the case why not just lap the Core of the CPU and the base? Why the silver layer? Seems to me the more layers youi add the worst off you will get.

And then whats it going to cost to get the stuff to grow this silver layer? How long will it take? Is it worth it for just a one off deal? How many people would even consider this?

[satanicoo]

It would be perfectly atached to the core, with no termal paste, and near perfect heat conductivity.

But still ppl, this is something imposible i think, the temperature would destroy the processor no?

[logosmani]

must confess i thought the idea was to grow a TIM between the hs + the core, not introduce a silver heat spreader with a TIM of its very own

[jaydee]

Originally posted by satanicoo
It would be perfectly atached to the core, with no termal paste, and near perfect heat conductivity.

But still ppl, this is something imposible i think, the temperature would destroy the processor no?

This doesn't make any sence though. You still have a joint in between the block and the silver thats going to be on top of the processor die? Whats the difference than just putting the block on the processor die as usual? Unless that silver is moleculary bonded to the Die I don't see anything that would resemble a significant gain here. You still have a TIM joint between the silver and the base of the block right? or is this silver supposed to BE the TIM joint somehow?

[jaydee]

Originally posted by logosmani
must confess i thought the idea was to grow a TIM between the hs + the core, not introduce a silver heat spreader with a TIM of its very own

Thats what I originally thought to, but the last several posts didn't seem to support that.