Water cooling Idea

[gone_fishin]

I want to know if anyone a has tried this.
A water block on the Die
Then a pelt on top of the water block
and finally another water block on top of the pelt.
The system would use two independant resevoirs.

 

[Iron Hawk]

I have read about something like that before, however I believe it was with one res. I think that the general agreement was that it will not be any more effective.

Let me see if I can dig that thread up...

 

[Iron Hawk]

Well i think this was it, however Its not exactly what you had in mind:
http://forum.oc-forums.com/vb/showthread.php?s=&threadid=53685

But it is a thread you may want to read over.


The system that you describe might work. But i have little experiance with pelts so i guess im not really the one who to give you the final answer.

 

[gone_fishin]

Well i think this was it, however Its not exactly what you had in mind:
http://forum.oc-forums.com/vb/showthread.php?s=&threadid=53685

But it is a thread you may want to read over.


The system that you describe might work. But i have little experiance with pelts so i guess im not really the one who to give you the final answer.

Thanks for the reply, I'm off to read that discussion.

Before I do , I can see a safety advantage here. If the pelt fails or if the first waterblock system goes down, then the CPU will still be cooled. More thoughts people.

 

[Tecumseh]

IF you want a safety advantage put two pelts side by side
on a good coldplate.

 

[gone_fishin]

IF you want a safety advantage put two pelts side by side
on a good coldplate.

well, safety isn't the main goal I am after but only an afterthought.

I guess basically you could call this settup a water cooled coldplate.

Here's a little sketch I drew up of the first water block (acting as coldplate). The center is solid copper which would transfer the heat up through the pelt which would be applied directly above it. The base would be a light thickness of maybe an eighth of an inch and the top as thin as possible, total thickness 3/4". The water channels around the center core would take care of the spreading heat as it transfered out from the contact point and notice it surounds the copper core. This would have the effect of drawing heat out of five sides of the center core of copper. Remember there would be a pelt and a second water block such as a mazeII on top with it's own water resevoir, possibly chilled with an evaporator coil. The first water system could be a standard in case type settup.

The center looks like a cat so it could be called the "Meow block"
Someone must have thought of this before.
So, tell me why it will or won't work.

 

[Kaese]

Another wonderful idea once again, to be hacked off at the knees by simple thermodynamics. The problem you have is twofold. There is not enough surface area to transer the heat effectively, regardless of how you make your waterblock. Secondly the waterflow is going to be too fast for any real cooling to occur. You would be better off trying to cool a reservoir.

Here is a link to an article with a setup that is far beyond what you are trying that resulted in very little cooling.

http://www.procooling.com/articles/html/kev_explores_pelt_water_chille.shtml

 

[Tecumseh]

What you have proposed is a sort of in-place chiller.
If you review the articles on chillers on various sites
you will see a common experience that they don't
work too well. I believe this has to do with the transit
time of the exchange fluid (water).

Your idea of having two loops could help. The circulation
thru the outer WB can be faster and thus more effective.

I think the real issue here is this: Is it better to have
a WB under a pelt or simply a coldplate? I believe that
the best, by far, is to use a coldplate. This will achieve
the lowest overall temps on the die.

 

[gone_fishin]

The idea here is NOT I repeat NOT to chill anything with the first waterblock but to remove the heat that spreads out through the first block so it doesn't reach its max load of heat. The main heat transfer is through the center core of solid copper up through the pelt and taken away by the second waterblock which would have a chilled resevoir of sub zero fluid. This second resevoir would have a 25 gallon capacity and cooled by an evaporator. The size of the evaporator I have in mind would reduce the resevoir to a 20 gallon capacity.
See my parts list here:
Parts list

The pelt would be at least 180watts. Can you get larger than this?

 

[gone_fishin]

An edit to the block design. It would have a raised central core extending 1/4" beyond the enclosed top for the pelt to make contact with. This would also serve for more space for insulation.

 

[rossd]

Water circulation won't help the heat conduction to the second cold plate. In fact any water in the way will hinder it. Also you'll have heat leakage into the cold reservoir and pipes, which will degrade performance.

I can't see how this could help things.

My $0.02,
rossd.

 

[drunkmonkey]

Just about ur waterblock, it needs waterflow directly over the core.

 

[gone_fishin]

Water circulation won't help the heat conduction to the second cold plate. In fact any water in the way will hinder it. Also you'll have heat leakage into the cold reservoir and pipes, which will degrade performance.

I can't see how this could help things.

My $0.02,
rossd.

Please elaborate, I am very interested in your analysis.

 

[gone_fishin]

Just about ur waterblock, it needs waterflow directly over the core.

The dangerden mazeII on top does have waterflow directly over the copper core which is directly in contact with the core. Please elaborate on your deduction. Remember this is a pelt draw of heat to the second block.

 

[rossd]

As I understand it, you're talking about a water block instead of a cold plate between the die and the TEC. You have this block on a separate circuit.

Now you can't cool the new block with a radiator since it's operating at below ambient temperature. All you can do is have a reservoir and a pump.

You'll need to insulate the new circuit from its surrounds, since it's operating below ambient. Any heat leakage into it will hurt performance.

So what does the second circuit buy you ? It doesn't increase heat conduction from the core to the TEC. In fact it will probably hinder it since there's a longer path length with water in the way.

I guess it will provide thermal inertia so if your TEC fails, the entire system will heat up slower.

Apart from that, I can't see how it will help.

Regards,
rossd.

 

[gone_fishin]

"In fact it will probably hinder it since there's a longer path length with water in the way. "

No the water is not a longer path in the way. The cube of solid copper in the center is the only contact to the pelt and CPU. And yes there would be a radiator with the first block.

Please elaborate more.

 

[Voodoo Rufus]

Since the block will be cooled by ambient air and the pelt is below ambient on the cold side ideally, I don't think the cold will reach the CPU. At best it'll be somewhat better than a straight H2O ambient system. It's best to have a pelt on a thick coldplate then a waterblock on top of that. I think you'll get much better temps.

 

[gone_fishin]

Since the block will be cooled by ambient air and the pelt is below ambient on the cold side ideally, I don't think the cold will reach the CPU. At best it'll be somewhat better than a straight H2O ambient system. It's best to have a pelt on a thick coldplate then a waterblock on top of that. I think you'll get much better temps.

"It's best to have a pelt on a thick coldplate then a waterblock on top of that"
That's exactly what this is.

I see your point on the first block only being cooled by ambient air. But if you think about it, the water in block one is storing the residual heat, some of which is transfered away through a radiator. The storage capacity of the cold plate is now huge compared to a plain sheet of 3/8" copper. As I now understand it heat will transfer to water from a metal much more quickly than heat will transfer out of water to a metal. If this is so then the center core of copper is being insulated by the water and not drawing heat out of the water very efficiently, which is fine. Would this make the pelt draw heat directly from the CPU core more efficiently?

 

[gone_fishin]

So far I have heard a few good arguements against this, but none that have really knocked it off the shelf yet. I am hoping some of the senior experts could weigh in here.

I do have the heavy equiptment to do this, all I would need is a couple of milling bits. I have a one horse 15 speed heavy floor stand drill press with a sliding cross vise and a band saw for cutting copper stock.
Maybe someone else with more experience would like to try this.

 

[Voodoo Rufus]

I think it may be something to try and see how it works. I think that the block on thr CPU and under the pelt will remove heat from the CPU fine, But the returning water will be warmer than the cold side of the pelt and serve to warm it. Overally the CPU will be colder, but as compared to a dedicated pelt/coldplate/CPU system, I don't know.