Improve the TIM joint?

[satanicoo]

Its supposed to be the tim joint, and ultimately, we build the block on that grown silver. (i mean, machinise it)

Quite impossible ein?

[Cossey3]

my idea was to make the core as flat as possible thus making the contact area across the tim as large as possible. this will not give as good results as a system with no tim but i dont think there is a practical way for use to remove the tim altogether. the silver precipitation method can only build paper thin layers at a time so it would be very difficult to buld an entire hs. my idea was to make a heat spreader with a very flat top surface that was grown onto the die so while there is still a tim it is now between 2 surfaces that can be properly lapped so the contact area will be much larger.

[jaydee]

Originally posted by satanicoo
Its supposed to be the tim joint, and ultimately, we build the block on that grown silver. (i mean, machinise it)

Quite impossible ein?

If not impossible, certainly way out of reason. I am pretty sure this originated, from reading earlier posts, that they were going to try an grow the silver inbetween the Die and the block so it naturally fills the gaps and what not. Then it wen to growing silver on top of the die and then milling it. Guess thats where I got confused. This is all pretty unrealistic anyway.

[jaydee]

Originally posted by Cossey3
my idea was to make the core as flat as possible thus making the contact area across the tim as large as possible. this will not give as good results as a system with no tim but i dont think there is a practical way for use to remove the tim altogether. the silver precipitation method can only build paper thin layers at a time so it would be very difficult to buld an entire hs. my idea was to make a heat spreader with a very flat top surface that was grown onto the die so while there is still a tim it is now between 2 surfaces that can be properly lapped so the contact area will be much larger.

So what is pressure was applied a little off on a corner? Would the grown on silver pop off? How strong is the bond between the aluminum Die and the gown on silver?

[sevisehda]

Basically someone through out the idea of chemically plateing the cpu die with silver. Its actually a simple chemical reaction that results with Silver being deposited onto just about anything around.

If I was going to try this I'd mask off the bridges and anything else on the CPUs surface leaving as much room around the cpu open. Then I'd chemically buld up a thin layer of silver onto the die and surrounding areas. After it go thick enough I'd start to electroplate more silver onto the existing silver until I built up the amount I wanted. This would make the best heatspreader possible. If you were able to build a WB into the silver then you'd have no TIM and probobly the best WB known to man.

Possible Problems...
Silver may not stick to the die
Some of the chemicals used may eat at the die or the CPU
Lapping/maching the new silver may destroy the die

[Cossey3]

Originally posted by jaydee116
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?

the idea is that the way the silver is put on the die there will be no real tim between the die and the silver. so no we have a head spreader that will have a far flatter top than the die did so better contact area

[jaydee]

Originally posted by Cossey3
the idea is that the way the silver is put on the die there will be no real tim between the die and the silver. so no we have a head spreader that will have a far flatter top than the die did so better contact area

So how well would this bond to the Die. Which is AL from what I understand. If you put to much pressure on one edge of the silver IHS would it pop off or would the Die pop off the CPU. Then what does it take to do such a thing. Can anyone do it?

[Cossey3]

well the silver is too soft for it to pop off the pressure if it were great enough would probably dent the silver or snap the core off. it is unlikely to be any weaker than the normal core.

the reaction is done by mixing silver nitrate solution with ammonia
then a reducing agent is added with will cause the silver to come out of solution and form a film over all surfaces it can come it contact with. then this is repeated until the layer is thick enough.
anyone with a basic knowledge of chemistry can do this and all the cemicals are readily available but silver nitrate is expensive as it contains silver.
the electroplasting idea is a good one because that would be capable of building up a much large amount of silver.

[utabintarbo]

Originally posted by jaydee116
So how well would this bond to the Die. Which is AL from what I understand. If you put to much pressure on one edge of the silver IHS would it pop off or would the Die pop off the CPU. Then what does it take to do such a thing. Can anyone do it?

I believe the silver would be molecularly bonded to the die.

I also believe the process is no more complicated than anodizing aluminum, just using different chemicals. Ask cossey3!

Bob

[Cossey3]

yeah the silver is formed molecule by molecule therefore it will fill all the valleys on the die surface so it wont come off as a single piece.

[jaydee]

So whos going to try it then?

[Cossey3]

i wonder if cathar can do a version of the cascade with a silver base than could be joined to a cpu

[satanicoo]

*burns his cpu and spills hot silver all over his body*

Im ok guys, Really!

*dies after.*

[bigben2k]

I have no problem voluntering to do this, but I need an outlined procedure, because my chemistry classes are more than 10 years behind me

Of course right now, I don't have a CPU either

Tell ya'll what: I can purchase a 1700+ TBredB, and send it to whoever is going to plate it, and then that person can either test it, or let someone else test it.

[Cossey3]

well ive been totting up the cost and this is what ive found:
cascade block to attach to cpu £60
1700+ £45
silver nitrate £30 enough for a few cpus (may take some testing to work out how much is needed)
ammonia £2.50
aldehyde £2.50

so £140 or $230 its prob cheaper over the pond.

[redleader]

How does plating silver help? Theres no reason to think it would be any flatter then the silicon die. In fact since you're just allowing molecules to settle where they will it could be very, very course depending on minute elctrical differences within the silicone chip.

If anything you should be looking at plateing the base of the waterblock, but even then I see no reason to assume you can plate something and not make the existing valleys worse. Anyone ever see those childerns chemistry kits that make staligmits by chemically plating a flat surface?

----

Has anyone ever looked into low temperature solders? Since those form a continuous metallic layer from the die to the water, I think they'd be ideal. I even remember seeing a few that had melting temps in the lower 100sC. Drop that a few degrees and it might be practical on a CPU.

[N8]

You guys always try doing things the hard way.....

I have stated before that you can (high vacuum) vapor plate or sputter just about any material onto any surface (almost). Sputtering is much more controllable and you can control the thickness better than vapor plating. I have two high vacuum vapor plating machines at work that I use to deposit a 1 micron or thinner layer of metals onto different materials. Aluminum, copper, silver, gold, chrome, nickel, etc. are some of the metals that I use to deposit thin layers. All I need is to make a mask to cover up what shouldn't get plated. You can machine a mask out of metals, or just cut some plastic (like overhead transparency plastic) or mylar into a mask.

The problem with doing such a TIM layer (as I have stated in the past), is that both surfaces (CPU and heatsink), MUST both have a matched flatness tolerance with each other less than the TIM layer thickness. You want a thin TIM layer, correct? Well you first have to have the CPU and heatsink prepared. The P4 IHS is NOT flat at all, forget about the P4 at this time. The AMD core is going to be pretty damn flat, and I wouldn't worry about lapping it. You will most likely round the corners off and make it more rounded than flat if you try to lap or polish it. The real gains here are to be made in making the heatsink surface flat. Technically, you only need the heatsink to be flat where it mates with the CPU. Think you guys can get the heatsink flat to less than 1 micron of tolerance? I am not talking about polish either. I am talking about FLATNESS, or deviation from a plane surface. It is hard to do. I can do it, but I have the equipment and 10+ years of job experience.

When you get down to that level of flatness, then you have to worry about surface roughness (polish) affecting the contact and TIM interface. The more polished your surface the less TIM you should be applying, otherwise you will be creating a TIM layer that HAS to be squished out as there is no surface roughness for the excess TIM material to flow into. It can only displace at the edges of the CPU/heatsink interface.

If you can get the FLATNESS tolerance and the POLISH tolerance down far enough (I don't know exactly what that level is), you can forget about using a TIM at all, as it will only be a detriment.

I have tried and done this and the big problem (and I mean BIG problem), is that whenever you mate two near flat and polished surfaces together, they suction to each other, and at that point, you CANNOT try to move them around to fine tune the alignment of motherboard holes, heatsink holes, etc. as the surfaces will literally rip each other apart and create voids, pits, and high spots and conglomeration lumps. They basically surface-weld to each other, which is exactly what we want, but you have to get it correct the very first try.

I don't have it tested yet (I just finished sealing the barbs in last night), but I just made a 'rotor'-style waterblock with the mill instead of a drill press, 1/4" OFHC (oxygen-free high conductivity) copper, o-ring, and a 3/4" polished plexiglass top. On the copper plate I was able to get both surfaces parallel to each other within 1 micron, and the surfaces lapped and polished to approximately 1 wavelength of light. With the bulb I am using, that is flat to within .0000232", or 0.6 micron. I only bothered with a 3 micron surface finish (polish), as I will be using some AS3. I could vapor plate a layer of copper onto the Barton 2500 I just got, but I don't feel like it, or have the time for it. Been running this Barton 2500 (1830 MHz) for 1 week now with the retail heatsink and retail TIM, and I have had it up to 2200 MHz with only 1.8v so far. If the leak testing goes well, it will have chilled water cooling.

Don't ask when I will put up pix, just wait a bit, and I will start a new thread.

[jaydee]

And with that I am just going to stick with Silver compound. Good Luck guys!

[bigben2k]

Thanks for chipping in N8, that's exactly what I've been saying.

I think I'll stick with AS3, unless I find something better. I also ordered a sample of the Thermagon stuff, and the DSM is giving me a hard time about it, because he says it's for "space" application, or where the surfaces are really flat.

[satanicoo]

Originally posted by N8
You guys always try doing things the hard way.....

I have stated before that you can (high vacuum) vapor plate or sputter just about any material onto any surface (almost). Sputtering is much more controllable and you can control the thickness better than vapor plating. I have two high vacuum vapor plating machines at work that I use to deposit a 1 micron or thinner layer of metals onto different materials. Aluminum, copper, silver, gold, chrome, nickel, etc. are some of the metals that I use to deposit thin layers. All I need is to make a mask to cover up what shouldn't get plated. You can machine a mask out of metals, or just cut some plastic (like overhead transparency plastic) or mylar into a mask.

The problem with doing such a TIM layer (as I have stated in the past), is that both surfaces (CPU and heatsink), MUST both have a matched flatness tolerance with each other less than the TIM layer thickness. You want a thin TIM layer, correct? Well you first have to have the CPU and heatsink prepared. The P4 IHS is NOT flat at all, forget about the P4 at this time. The AMD core is going to be pretty damn flat, and I wouldn't worry about lapping it. You will most likely round the corners off and make it more rounded than flat if you try to lap or polish it. The real gains here are to be made in making the heatsink surface flat. Technically, you only need the heatsink to be flat where it mates with the CPU. Think you guys can get the heatsink flat to less than 1 micron of tolerance? I am not talking about polish either. I am talking about FLATNESS, or deviation from a plane surface. It is hard to do. I can do it, but I have the equipment and 10+ years of job experience.

When you get down to that level of flatness, then you have to worry about surface roughness (polish) affecting the contact and TIM interface. The more polished your surface the less TIM you should be applying, otherwise you will be creating a TIM layer that HAS to be squished out as there is no surface roughness for the excess TIM material to flow into. It can only displace at the edges of the CPU/heatsink interface.

If you can get the FLATNESS tolerance and the POLISH tolerance down far enough (I don't know exactly what that level is), you can forget about using a TIM at all, as it will only be a detriment.

I have tried and done this and the big problem (and I mean BIG problem), is that whenever you mate two near flat and polished surfaces together, they suction to each other, and at that point, you CANNOT try to move them around to fine tune the alignment of motherboard holes, heatsink holes, etc. as the surfaces will literally rip each other apart and create voids, pits, and high spots and conglomeration lumps. They basically surface-weld to each other, which is exactly what we want, but you have to get it correct the very first try.

I don't have it tested yet (I just finished sealing the barbs in last night), but I just made a 'rotor'-style waterblock with the mill instead of a drill press, 1/4" OFHC (oxygen-free high conductivity) copper, o-ring, and a 3/4" polished plexiglass top. On the copper plate I was able to get both surfaces parallel to each other within 1 micron, and the surfaces lapped and polished to approximately 1 wavelength of light. With the bulb I am using, that is flat to within .0000232", or 0.6 micron. I only bothered with a 3 micron surface finish (polish), as I will be using some AS3. I could vapor plate a layer of copper onto the Barton 2500 I just got, but I don't feel like it, or have the time for it. Been running this Barton 2500 (1830 MHz) for 1 week now with the retail heatsink and retail TIM, and I have had it up to 2200 MHz with only 1.8v so far. If the leak testing goes well, it will have chilled water cooling.

Don't ask when I will put up pix, just wait a bit, and I will start a new thread.

What do you think, is it possible to make a layer of +-5 mm on top of the cpu?

[N8]

Originally posted by satanicoo
What do you think, is it possible to make a layer of +-5 mm on top of the cpu?

5 mm? Holy crap that is pretty thick. I am guessing you mean 5 micro meters (microns)? You could do that (5 microns) with a sputter deposition setup, but I don't have one with easy access.

The other problem that could occur once you start growing/depositing/layering metallic layers is that they could de-bond or flake off and some point in the installation or removal and then you have a problem with a very conductive particle(s) on your CPU/mobo.

[hara]

Originally posted by bigben2k
Thanks for chipping in N8, that's exactly what I've been saying.

I think I'll stick with AS3, unless I find something better. I also ordered a sample of the Thermagon stuff, and the DSM is giving me a hard time about it, because he says it's for "space" application, or where the surfaces are really flat.

Why don't you try this shin etsu stuff.

Another TIM idea I had was to use gold leaf between the CPU and cooler.

The reasons are simple: Gold is a very mallable material and with compression *could* make a good TIM. The leaf is only a few microns thick.

Gold's thermal conductivity isn't bad at all. It is better than aluminium but not quite as good as copper. If you calculate the thermal resistance for that short distance, the difference between it and silver would be negligable IMO.

I saw a thread on [H] once about using al foil. Some people got quite good results since the foil is adaptive. That is why I'm suggesting gold leaf now.

[Cossey3]

Originally posted by N8


The problem with doing such a TIM layer (as I have stated in the past), is that both surfaces (CPU and heatsink), MUST both have a matched flatness tolerance with each other less than the TIM layer thickness. You want a thin TIM layer, correct? Well you first have to have the CPU and heatsink prepared. The P4 IHS is NOT flat at all, forget about the P4 at this time. The AMD core is going to be pretty damn flat, and I wouldn't worry about lapping it.

Just how flat is the amd core are we talking a few microns or <1 micron
and if the amd core is flat enough if we were to get the base of the heatsink or waterblock to a similar flatness we could do without the paste.

[sevisehda]

OK back to the original topic, improving the TIM. A big problem evryone has is too much TIM. The flatter we get the die and block the harder for the TIM to slip out. Every TIM I've run across has been very thick. Is there a way to thin it up with alcohol or something so the extra can more easily flow out of the gap?

[Cossey3]

thers someone on the ocuk forums who tries using alcohol to thin down the as3 and it knocked a couple of degrees of so if we can get the surface flat enough that may be the answer.