how thin is too thin?

[The Overclocker]

i know this have been covered before, and i know the disadvantages of a thin waterblock base (less heat spred but quicker heat dissapation into water) and a thinker base (longer heat retention but better heat spread) but what i want to know is, how thin is too thin?

 

[Hoot]

I've just seen the result of this the past few days. With blocks that gush the inlet water in the center, right over the core, thinner is better. Too thin would be if the retention pressure causes the base to cave in where it sits on the core. With spiral type designs, you can go very thin and not risk this as the walls of the spiral strengthen the thin base. The Maze-2 is an example of a block that could benefit from being milled deeper. The MCW-462, is an example of a thick base, though it has some stippling holes drilled into it. It got beat by a thinner base spiral block that weighed 1/4 as much as it did.

Hoot

 

[BillA]

might want to ck that Hoot
I have a bunch of data going the other way

you started on a center feed, then shifted to an internally 'fined'

saying (I guess ?) that thinner is better for all ?

apart from the different 'types' of wbs,
have you looked at ONLY changing the bp thickness ?

I have, and the data would not support your observations

are we going to be heading 'back' to Aesik's thread ?

be cool

 

[nihili]

I've wondered about using an uneven thickness. Thin at the center and thicker as it goes out to the edge. Of course the base would be flat, so basically I'm talking about slanting channels.

nihili

 

[Aesik]

Why would you want it thinner in the middle and thicker on the edges? Remember that thermal conductivity is directly affected by the cross sectional area of the solid body.

 

[nihili]

Well, I take it that Hoot's comments, and the whole direct die movement are motivated by the idea that you want the thinnest possible base so that the heat transfers across it quickly and efficently. Thick bases are motivated by the idea that the base acts as a sink, a holder for the heat, and it dissipates it across a wider area.

By using a graduated base, I would hope to attain the best of both worlds. I would get very efficient transfer at the center while still allowing the heat to spread outward and allow a greater area for cooling. Additionally the graduated design should allow me to get thinner at the center because I would still have the structural support of the somewhat thicker area at the edge of the core.

My thought is to try a multi-armed spiral, with one arm beginning at each of four corners converging to a central chamber just above the core. I'd probably run the flow from the edge to the center in order to take advantage of (some process whose name I can't recall, but basically where you use the coldest water to cool the cooler parts and as it warms you use it to cool the hotter parts).

Keep in mind this is entirely theoretical. I'm not an engineer, nor do I play one on TV. And I don't even own a water pump (yet).

nihili

 

[Aesik]

Hmmm....I'm sitting here pondering how exactly a thicker perimeter but a thinner base would perform. The only advantage I can see to having the thicker perimeter would be to act as an energy sink. Unfortunately, as soon as the system reaches equilibrium the effect is virtually negated.

I need to ponder some more before giving up on the idea

 

[BillA]

perhaps looking at the question incrementally:

at what distance from the edge of the die will there no longer be any contribution to heat transfer ?
(meaningful/significant/measurable)

how does changing the bp thickness affect this distance ?

if more is better, how much more ?
or
if less is better, how much less ?

so
start thick, go thin ?
or
start thin, go thick ?

now this question is also quite related to the aspect ratio (width to depth) of the folded channels in a labyrinth type wb
(but the shared walls make the analysis iffy)

no answers today, only questions

be cool
(Hi nihili)

 

[nihili]

Hmmm....I'm sitting here pondering how exactly a thicker perimeter but a thinner base would perform. The only advantage I can see to having the thicker perimeter would be to act as an energy sink. Unfortunately, as soon as the system reaches equilibrium the effect is virtually negated.

I need to ponder some more before giving up on the idea

Well, as I understand it, having a thicker piece of metal beyond the edge of the die will allow heat to be conducted further away from the die as the energy sink reaches capacity. This means that I have a larger useable area on which to place water channels than I would if I had a uniformly thin base. So I get more useable transfer area.

nihili

***Edit***
Unfortunately, while I recognize BillA's questions as good ones, I haven't any idea what their answers would be.

(Hi BillA)

 

[Owenator]

This is an interesting discussion because I have use two similar water blocks one with a 1/8" base that had a 'core spraying' inlet and open cavity and another maze type with a 1/16" base. I found that the thinner base worked better for the maze as Hoot mentioned. I notice that swiftech also uses a 0.320" thicker base on their newer 462 and state that it has a 25% higher heat capacity. as I understand it a thicker base can retain more heat and unless it is removed effectively that heat will remain. I'm sure like all engineering problems there is a point or 'sweet spot' where the thickness and flow correspond to the best cooling. I recall seeing a tiny waterblock about the size of the core that worked but it wasn't the best out there. The amount of turbulence and surface area inside the block to remove heat will be the deciding factor on base thickness. The spiral walls can remove heat more efficently so a thin base is fine because the water is cooled rapidly as it flows through, less internal volume is ok in this case. I am sure flow rates also make a difference because the longer the water is in the waterblock it can absorb more heat but asit warms it also absorbs less so there is a point of diminishing returns. From what I have seen we really need a person to do a 3D water block simulation that combines a conduction model for the block and a fluid model for the water. I saw an article in the American Society for Mechanical Engineer magazine that had a 'engineer' designed passive water cooler. That would be interesting research to look and to see their results. I'm sure it is published somewhere, I'll try to find it.

O

BTW I used to be an Engineer but I don't use that knowledge that much any more so feel free to tell me I'm all wet

 

[Wicked Klown]

Ok this may sound stupid but here it goes. How about a waterblock that fits all around the core. Like have alittle place in the bottom thats fits the core. Kinda like this.

 

[Tecumseh]

This is all a good discussion. To answer the original
question: I think anything less than 3mm for the base
is going to require an internal design that will keep the
base flat and in contact with the CPU core.

 

[Owenator]

OK, I dug out my text book on the fundamentals of heat transfer and it looks to me like the thicker the bottom of the water block the greater the drop in temperature at the inside surface. The heat tends to spread out more in a thick bottom and spread less in a thin bottom. This suggests that thr internal design is the determining factor on thickness as I said earlier. An 'open' block with lots of mixing like the swiftech one benefits from the more even surface temperature due to the thick base, like a cast iron skillet. The spiral blocks have more surface area to transfer the heat so they can handle higher temps at the inside surface, which is what a thin base does. I used to have some thin bottomed pans and whenever I burned some soup (I'm not the best cook) It would stick more to the middle than the outside because of this same effect. Now I have to make some more water blocks to test this theory. Sorry I'm so longwinded!

O

 

[The Overclocker]

now to reveil my plan - i cannot get thick copper easilly here, so i must get sheet copper which is around 2-3mm thin, for the base it will be a doube layer with the upper layer covered in multipull holes to increase turbulence by making water 'swirl' around in the holes - abit like the koolance waterblock. the block may also use fins or a spiral design comprised for cut segements of copper pipe. the 'skin' of the waterblock will be plexiglass or a aluminum project box.

 

[Aesik]

Just one question for you...

How do you plan to attach the two bottom plates together so as not to get too high of a thermal resistance at their junction? If they aren't perfectly mated, you are going to have a serious problem getting the energy from the plate in contact with the cpu to the plate with the holes drilled in it.

 

[The Overclocker]

i will probably lap both sheets then solder them together or use artic silver epoxy and put them in a vice

 

[Tecumseh]

No, I think that is a mistake. By the time you are done
you would have been better off with aluminum.

Getting thick copper cannot be that difficult.

 

[Aesik]

Anytime that you don't have a continuous piece of material, there is going to be a pretty big temperature drop across the interface. No matter how much you lap, press, etc. it will never create and new grain structure across the interface. Thermal conduction depends on the grain/molecule structure to transfer energy efficiently.

Like Tec said, I too think you would most likely be better off using aluminum then trying to jury-rig up seperate pieces of copper.

 

[Barryng]

Aesik has been giving you guys excellent advice. To answer the original question you have to first acknowledge that the purpose of the block is to be a heat exchanger, NOT a heat sink. The ONLY purpose of the block is transfer heat into the water. I would guess that the amount of heat transfered off the surfaces of the block to the air in the case is negligible. Even this is hindered as the walls of the block get thicker.

It serves no real purpose to have a thick massive block. In fact, the thicker the block the harder it is to transfer heat to your primary heat transfer medium, the water. The thickness of the block that is ijn contact with the CPU should be no thicker than that necessary for an acceptable pressure boundary and to prevent mechanical damage.

Heat is transfered from the CPU to the water via conduction through the bottom surface of the block. Conduction requires a temperature difference to work. Therefore the inside of the block must always be colder than the outside. The thicker it is the bigger the temperature difference. Practically this means that the CPU gets hotter and the inside/water stays about constant. As you can see thinner means a cooler CPU.

Since water is the intended heat transfer medium, it is advantageous to get the heat into the water as quick and efficiently as possible. The rate of heat transfer increases with decreasing thickness of the bottom surface, larger surface area of the bottom surface/CPU, and better conducting materials (copper, silver).

 

[rossd]

Has anyone considered an "active" water block ?

Thermal conductivity of copper is high but water and air are comparitively _very_ low. The heat flow in the water is therefore mostly accomplished by turbulence. Mixing the water up rapidly causes heat to travel into the middle of the water stream.

So one strategy is to create a block with a long winding channel and a thick base. The base efficiently distributes the heat to the surface of the water, where it is taken away by turbulence.

Another strategy is to inject the water onto a small thin base right over the core. The usable channel surface area is only as big as the core, but you don't get any temperature drop through the copper. Direct-die cooling is a variant of this.

The optimum is no-doubt somewhere between the two strategies above.

I wonder if anyone has considered an "active" water block. That is, instead of using turbulence in the water to create mixing, you force it to happen mechanically.

e.g. Have a rotating metal "squeegee" blade directly on the core. This is a direct-die setup. The squeegee rotates rapidly, forcing the water on the surface of the core to be recirculated every few milliseconds.Thus the water on the core is always fresh and guaranteed to be at reservoir temperatures.

You could power the squeegee directly from your water flow (just like a lawn sprinkler). No need to have a motor right on the block.

Such a system would have to mate with the core to a very fine degree of accuracy (within a micron), so you'd probably want to lap the core . The squeegee could be sharp enough to do this automatically when it is first installed.

I think this system could beat the pants off any conventional direct-die or water block system. If done properly, the temperature of the surface of the core will be virtually the same as the reservoir temperature.

Comments ?
rossd.