Refrigerant touching chip?

[ak_47_boy]

Will propane, R134a, or oil eat any part of a cpu or socket? I would like to try some DIRECT die cooling :D

[n00b 0f l337]

Your going to braze to an IHS? Too thin, will buckle.

[AlphaChicken]

how could you braze to an IHS anyway without toasting the chip?

[n00b 0f l337]

Remove it first ;0

[ak_47_boy]

I was thinking about removing the ihs and connecting the cpu to motherboard via about 6" of 30awg kynar wire. 1550 solder joints :(

I think just dipping it in LN2 would be alot easy then building a box around the cpu for phase. I forgot about the 775 wire holes in the box that would need to be sealed :D

[n00b 0f l337]

Think about your designs before you do em or mainly post em.

[PhilippF]

Besides it takes a little time to do 775x2 solderings: There are high frequencies involved, so I think any lengthening of the signal pathway between CPU and for example RAM would cause a lot of problems. The designs of the motherboards are very tricky and create appropriate shielding of the signals by positioning them between grounded layers.

I can think back to times where an engineer told me "More than 200 MHz on a CPU will not be possible because of the necessity to conduct these HF signals away from the CPU". Now he was proven wrong obviously, but I think the point is that the HF signals will get very fuzzy when pathways get too long.

I had some similar thoughts a while ago when playing with watercooling but came up with no promising idea how to realize it. One guy in a forum did glue a box on top of an Athlon XP (without IHS), so that the water could cool the die directly. It showed worse temps than a good waterblock, which is, in my opinion, the result of the reduced heat exchange surface.

Regards,

Philipp

[Pete]

How about an evap IHS.

The STD IHS would pop under the presure!

So make a cop of it but with a thicker base but already mounted to the rest what makes the evap up.

Then it's direct die

[ak_47_boy]

True, 30awg would be to thick for 4ghz. But then again its only a few inches. Inductance measurements would have to be taken to be able to calculate the independence.

I don't think making a evaporator out of the ihs would be much better than a conventional evaporator. Removing the ihs would probably yield the same results. I think cooling the whole chip front and back would defiantly help.

It can be done, it would just take a hell of a long time to solder.

So i am assuming refrigerant or LN2 will not eat silicon, the glue, smt parts, or the pcb?

[|RickY|]

use soft solder or some power glue :D

[[XC] gomeler]

Or... wait for it.. remove the IHS, make a thin tape shim, and directly mount the evaporator to the die! Welcome to 2002 (atleast for me).

[Stigma]

ok, 2 problems here...

First of all you can't braze or even solder anything to the chip itself, or it will fry and become useless.

You could pop of the IHS and braze that on, but in that case, how is that differetn from just using a regular evap and taking off the IHS and doing direct-die cooling? In both cases you get 1 layer of metal between you and the core, so both will be equally effective in principle. An IHS dosn't actually DO anything with the heat after all, its just a layer of metal there to lessen the chance of dammage to the CPU by preassure.

-Stigma

[ak_47_boy]

No, i mean having the cpu(without ihs) actually inside the evaporator. This way the entire surface of the cpu(PCB and all) would be cooled.

Or have the CPU actually submerged in LN2.

[[XC] gomeler]

Problem is surface area. The surface area of the CPU die is infinitesimal compared to the potential surface area of an evaporator. By maintaining contact with the die the evap utilizes the greater surface area to more effectively remove the heat. Just pouring LN2 or refrigerant over the chip isn't going to be as effective as using a well designed evap.