Water cooled CPU backside?

[pissboy]

bigben.. you know if you *really* want to make a good block with tiny ass holes in it.. I could sneak a block into a machine at work and drill holes for you. Those brass tubes are used in drilling (EDM=electron discharge machining=we never touch the part, just hit it with a high frequency high power spark)

We need the tubes to be hollow so we can use the water to flush out any eroded metal. Copper would be tough as its very conductive, but I think I could work with it. I can do holes as small as .010" (with a .008" electrode) but I think best bet for me to get them done before someone sees it is if I made them like .014-.015" diameter. I've actually been tempted to play with the idea of making a block in them for a while, just haven't really had much time. you could probably fit a 1/8" thick block down there, or a little less if you use a thermal pad on one or both sides of it. I have electrodes to make holes as big as about .072", but a .020" or .015" hole would be ideal for me (I already have parts I'm drilling with holes that small, can just sneak your part in and pop a bunch, heh.) I would need 2 pieces or your final part and something I can use as a scrap piece to figure out the burn settings I'll be using on the final. Holes that are completely through the part are easiest for me, if you want I could drill a bunch and then 2 bigger holes to act as inlet/outlets in a cross drilled pattern. You'd need to solder a cap on it though.

just a thought, maybe we can see this become a reality

[redleader]

Ben:

As long as you're back there, heres a better idea. Reinforce the back of the die and then increase the clamping pressure. Higher pressure means a better interface between the CPU/TIM/Block. Cut into that 8C gap mentioned in your other thread.

[bigben2k]

Thanks Pissboy, but I think I'll stick with the copper stock parts.

Redleader: I don't see how the core would become less crushable?

[airspirit]

If you insist on this insanity, how about you remove the zif socket altogether and solder the bastage to the board? The plastic piece is only an installation guide afterall, and this will allow you to use laterally thin but vertically tall tubing ... kind of like a tapeworm. It could happen ....

They do this on some ECS boards with durons, btw.

[redleader]

Originally posted by bigben2k
Thanks Pissboy, but I think I'll stick with the copper stock parts.

Redleader: I don't see how the core would become less crushable?

I assume if you evenly increase the mounting pressure the cermaic would start to bow and the core would crack. I've never tried it so I could be wrong.

[pissboy]

[quote]Originally posted by bigben2k
[b]Thanks Pissboy, but I think I'll stick with the copper stock parts.

I'm offering to cross drill a block for you, copper or whatever. Just was saying I can put very small holes in it.

[bigben2k]

Originally posted by pissboy
I'm offering to cross drill a block for you, copper or whatever. Just was saying I can put very small holes in it.

I might take you up on that, but it would be for a TEC/Peltier chiller, not this. The holes wouldn't need to be small.

[Cptn. Foo Foo]

This just came out. I seen the news bit on it at AMDZone.

Someone here said it could never be done. Well...some company pulled it off apparently. We will have to wait for some reviews to be sure though.

http://www.upgradeware.com/english/p...ptmc/xptmc.htm

Sounds like making a socket extender could work.

[Boli]

It doesn't show how thin it is though... Personally I recon you could get the socket about 5mm raised and just by upping the voltage a bit you can negate the additional resistance.

Though you don't get anywhere without trying it. I don't have the capabilities but maybe you could see HOW high you can move it off the socket whilst still retaining stability.

Using a non-conductive fluid you could make a hollow riser and cool all the pins as well as the back side (though making the pins would be a pain).

Whatever you do choose it is going to be a toss up between making the CPU colder for OCing and the rising stability issues involved.

[bigben2k]

Thanks Capt, very nice find!!!

[Cptn. Foo Foo]

Thanks bigben2k.


Actually I had always thought cooling the back of the CPU would be a good thing to try. I mean where else could the heat go but into the water....right.

I had put some thought into this idea some time ago. I thought what if you could extend the CPU pins and then bent out the center to allow for a pipe or tube to get in there.


Here is a rough pic I made just now to illustrate my idea...



That of course is just the basic idea. That does NOT show how to make the extender solid or secure so you could put pressure on it when mounting the water block on the CPU. Also, all of the indifidual pin extensions would have to be kept isolated from each other if they were just bare wires.



Anyways, this my $0.02 on the subject

[Boli]

Apologies for the crudeness of the drawing...

Anyhow the pins of the CPU are sloted into "extention pins" that go through the waterblock.

The extention pins can be small plastic tubes with another metal pin the CPU pin makes contact wioth inside or something involving non conductive fluid/ coating.

Even with widening the pins you should have a nest of longer pins with gaps between them all.

The water goes through the gaps at high speed and impacts on the small Copper heatsink in the center picking up its heat before exiting.

the Copper heatsink makes contact first with a non electrical conductive pad before it touches the back side of the CPU

I know it doesn;t look or sound very clear but I hope you can figure out what I am getting to.

NB: I know blockages are going to be a problem matic but without cutting into the back of the Mobo I think this could be te best solution I desides to use the nest of pins to an advantage rather than try to work round them.

~ Boli

[flyingass]

Originally posted by Cptn. Foo Foo
...Someone here said it could never be done. Well...some company pulled it off apparently. We will have to wait for some reviews to be sure though.
http://www.upgradeware.com/english/p...ptmc/xptmc.htm

My earlier post probably did overestimate the negative effect such a socket extender would have on a system, but the fact remains that any modification of the signal traces' length will change the signal handling properties of the system, possibly out of the range it was designed to handle. Any modification of trace length will affect the system's maximum overclock.

At these frequencies, resistance is the least of the issues signal engineers have to deal with - Dan's data responded to a letter a while ago which brought up some of these issues:
http://www.dansdata.com/io011.htm

Trace length isn't the only factor. There are capacitance and resonance and resultant signal settle time issues as well; when you're dealing with high speed digital systems you end up fiddling with a lot of analogue variables.

Realistically, these issues would counteract any possible benefit backside cooling would offer. Socket extensions are simply not viable if you're seriously concerned about an overclock.

I'd still like to see backside cooling put to use though. Best of luck, bigben, if you're going to try implementing it.