Why Doesn't a Fractal Heatsink exist?

[Twigsoffury]

This is my thinking:

• Massive amount of coupling, hopelessly impractical, multiplying leak risk many-fold
• Massive increase in flow resistance, would need a much higher flow rate. That means a much higher water pressure, and a massive increase in leaks... see point 1
• Enormous blockage risk (though I appreciate that this may be mitigated)
• Uncleanable (can't visually inspect either)
• Unmachinable (would need layers of laminated material therefore reduces thermal conductivity anyway)
• Thermal conductivity in copper is so high, warmer spots a few mm apart are removed within moments anyway, mitigating the need for this design (see last point)
• ...Why not just machine a very large number of straight passages through a thin block? There's an industrial process used for print rollers that drills extremely narrow holes to create a layer of air between the paper and the roller, I've always thought a very large number of small parallel holes like that would be the best option...

http://t3.gstatic.com/images?q=tbn:A...z5jbglolAQ&t=1

Dr.walrus strikes again.

[dr.walrus]

http://t3.gstatic.com/images?q=tbn:A...z5jbglolAQ&t=1


Dr.walrus strikes again.

[Blibbax]

It could work in a fanless setting, though.

No, firstly for all of those excellent Walrus based points above, and secondly because genuinely passive cooling (as in not even a case fan) still only works by moving air.

If the air around the heatsink was stationary, it'd eventually just become incredibly hot and the CPU would overheat. But the heat from the heatsink creates a convection current, which moves air over the heatsink. This heatsink would stifle that convection current, and indeed much of the heatsink would never be reached by the passing air.

[Twigsoffury]

No, firstly for all of those excellent Walrus based points above, and secondly because genuinely passive cooling (as in not even a case fan) still only works by moving air.

If the air around the heatsink was stationary, it'd eventually just become incredibly hot and the CPU would overheat. But the heat from the heatsink creates a convection current, which moves air over the heatsink. This heatsink would stifle that convection current, and indeed much of the heatsink would never be reached by the passing air.

the air necessarily doesn't need to touch the inner parts of the heatsink since the opposing cold flow cools off the hot side coming from the other direction. and its internal structure is basically a bunch of interconnected bent radiator fins welded together thats in and of itself surrounded by water that is surrounded by a larger array of fins to dissipate the heat generated....just like how heatpipes work.

Sure looks like that system would burn up from a outside perspective doesn't it just like your thinking with the inside of my heatsink.? But since physics says heat ALWAYS travels to cold and cold never travels to hot the inner most core of my heatsink would only be slightly warmer then the layers towards the outside.



And i was thinking it would look something more this this that sits outside the computer.

and not something that fits nicely inside your computer case baking in the heat. But the device up there is just a big empty tube inside.

Not alot of metal on water physical contact with the heatsink is there when you think about it....


I'm saying make a device with a set amount of channels for each side that follows a sort of fractal based design for maximum surface area against the water in as small of a space as possible inside.

And make the hot sides enter from a opposing side and always have at least 2 of those surfaces against a cold side running in the opposite direction for maximum thermal transfer between the hot and cold.


Heres a single loop version a friend came up with with the classic reservoir

Physics just says a square has a better capacity then a circle does (less wasted space using squares then using circles) so thats why my design was cubed and not cylinderical like the one up above since I'd be using a ton of ultra bent tubes or channels inside and not a empty space. Guess i need to MSpaint CAD me a side angle view of what im talking about.

I'll get to work. hopefully it'll come across better on the internet since i can't exactly explain this in person with yall.

[Twigsoffury]

i used red and blue for opposing flows. The arrows on the side show each tubes hot enter and cold exit. and remember thats a side view of something almost 20 channels deep so it'd be a hell of a lot of individual tubes.


I know that you can't machine something like that directly. But you sure can section it out since its fractally based. basically like building a ship. Build each single loop as a section of boxes and rectangles with one hot enterance and one cold exit, Combine 2 loops together to make a exact copy only on a bigger scale as the single version with 2 hots and 2 colds . combine two of those together to get 4 and so on and so on.

then combine 4 complete rows together for err.. 64 total hots and colds for each device.

then combine four more of those rows together for all the devices together and have 256 total tubes for those 4 devices to get cooled in.



Kinda think of it like a S@#t load of heatpipes inside a heatsink only using extremely..extremely...nerdy math to get the most volume out of the space available without having any of the tubes mixing with each other.

And since its fractional you can step it up as big or small as you want i'd guess.

This is my thinking:

[*]Massive amount of coupling, hopelessly impractical, multiplying leak risk many-fold

I'd use a single machined manifold on each end that has four inlets and four exits.

[*]Massive increase in flow resistance, would need a much higher flow rate. That means a much higher water pressure, and a massive increase in leaks... see point 1

Thats why i would use four single pumps in my design. but if you take into account that im reducing a larger volume several times, like home plumbing, when you force a larger volume of water through a smaller diameter tube large pressures are created in and of itself, and that pressure would be reduced back to the normal pressure by the manifold on the other side and return to 1/2" tubing or what ever you were to use.

So basically while yea it is "high pressure" It's also low pressure before and after the heatsink, and all the high pressures happening are contained inside the heatsink itself. all the hoses in the system would be whatever the pumps pressure is like the system you have now. (Think of a Air conditioning system with the hi-pass and lo-pass sides. Just the act of the Freon passing through a expansion chamber makes it cold, and when its compressed, it gets hot and passes through the radiator to get colder, then expanded again to get ultra cold and take in the heat from the air. then compressed and cycled again)

[*]Enormous blockage risk (though I appreciate that this may be mitigated)

Yup, better use some copper based anit-algae additive or use PPG polypropylene glycol but i'm sure the pressures created internally is going to blow any s@#t away that attempts to make a clog.

[*]Uncleanable (can't visually inspect either)

Can't exactly inspect the inside of the radiator you have now can you?

But if its fractally based with manifolds on each side. if you had a gasket that covered each of the sections and the manifold attached on top of that (think of a F1 cars engines intake manifold) you could break it down and clean each of the sections then assemble it back together. it'd be a bunch of boxes inside a box inside a box inside a box. inside a bigger box with a manifold on each end.

Sort of like that.

[*]Unmachinable (would need layers of laminated material therefore reduces thermal conductivity anyway)

Ultrasonic welding is the s@#t and a leading trend in welding technology should check it out sometime. you could easily assemble and sonic weld two complex shapes together, then weld those two pieces to two pieces that are exactly the same shape only polar opposite. and you've got one section, do that 512 times over and you've got my heatsink design.





Majorbud's job has a huge industrial Sonic welding machine that can assemble a aluminum window frame in less then 1/4 of the time it takes the traditional machine to weld a aluminum frame, they have plans for 5 more units to replace the current welders. Theres also almost no risk of leakage (air,water,argon gas inside the multi pane windows) since well.. a sonic weld is 100% the same edge to edge and doesn't use heat to mesh the two sides together. It uses a Ultra f@#king dank sound to litterely shake the two pieces into one.

If that machine can weld pieces like that together, why can't it weld my two pieces of metal together? then repeat that process hundreds of times over.

[*]Thermal conductivity in copper is so high, warmer spots a few mm apart are removed within moments anyway, mitigating the need for this design (see last point)

exactly, now imagine how much thermal energy you could dissipate in a 1 foot long cubed heatsink with 256 tubes that have opposing cool hot flows 1/32"nd of a inch away from each other for the whole 9 foot trip through the heatsink. with two cools sides always touching the hot side. remember if you add all those 1/32" walls together i'd think they would add upto about 3" of solid material inside the cube. thats a s@#t load of cooling potential in itself.

I'm sure if you had some cool water already looping the opposing lines on the other circuits, you could cool off a 5 gallon container of boiling hot water in less then a few minutes.

[*]...Why not just machine a very large number of straight passages through a thin block?

Because my design has almost 44x the amount of travel and almost (lemme use a calculator) 1,648.374x the surface area in the same space as a straight through pipe.

and i can fit (calculator time again) almost 20x the amount of tubes in the same space as a traditional design with straight through pipes. because i can fit two or more tubes in the volume of space your straight tube occupies in that same heatsink. AND with interposing directions in that same exact space.


its basically just based off the mandelbrot set. in laymans terms. It's like when you try and measure a 1 foot complex shaped cube with a 12" ruler you'll come out with 12" x 12" x 12".

But if you measure that same 12" box only now with a 6" ruler, you'll come out with 24"x24"x24"

Do it again with a 3" ruler and you'll come out with 48x48x48 sized box.



Sort of like that, or

The Mandelbrot set M is defined by a family of complex quadratic polynomials

given by

where c is a complex parameter. For each c, one considers the behavior of the sequence

obtained by iterating Pc(z) starting at critical point z = 0, which either escapes to infinity or stays within a disk of some finite radius. The Mandelbrot set is defined as the set of all points c such that the above sequence does not escape to infinity.


A mathematician's depiction of the Mandelbrot set M. A point c is coloured black if it belongs to the set, and white if not. Re[c] and Im[c] denote the real and imaginary parts of c, respectively.
More formally, if denotes the nth iterate of Pc(z) (i.e. Pc(z) composed with itself n times), the Mandelbrot set is the subset of the complex plane given by

As explained below, it is in fact possible to simplify this definition by taking s = 2.
Mathematically, the Mandelbrot set is just a set of complex numbers. A given complex number c either belongs to M or it does not. A picture of the Mandelbrot set can be made by colouring all the points c which belong to M black, and all other points white. The more colourful pictures usually seen are generated by colouring points not in the set according to how quickly or slowly the sequence diverges to infinity. See the section on computer drawings below for more details.
The Mandelbrot set can also be defined as the connectedness locus of the family of polynomials Pc(z). That is, it is the subset of the complex plane consisting of those parameters c for which the Julia set of Pc is connected

http://en.wikipedia.org/wiki/Mandelbrot_set

[Fuganater]

[/offtopic]

I want that case!!!!!!!

[/offtopic]

[mDust]

Where is the cold fluid coming from? It looks like cold fluid runs into the cpu block which then becomes warm fluid. It runs into the fractal cooler and is cooled by air and cold fluid which is also on its way to the cpu block? I'm fairly certain you won't have hot and cold tubes, but all warm tubes within a degree or two of each other.

Constriction would definitely be a problem. There will be high pressure between the pump and the fractal cooler and lower pressure on the return loop. You will not get any 'refrigeration' effect from this. It will either blow the tube off the pump, leak, or damage the pump. If you increase the size of your fractal cooler the 'larger pipe diameter' will ease the pressure...but not by much. You'll still need at least a couple of pumps with either a huge multi-pass, series loop or multiple dedicated loops.

If you sonically weld the layers together, then it's back to not being cleanable. Fluid additives would be out of the question. Mixed metals and corrosion would be an even larger problem.

Squares might be the first choice when it comes to efficient use of space, but they will play hell on the fluid dynamics. You want a little turbulence in a waterblock to maximize the number of molecules that strike the block, but you don't want as much turbulence as your square fractals would create. This is what would be slowing the fluid and creating all that pressure at the intake. You could at least make use of gravity by putting the intake at the top and outlet at the bottom of a tall stack.

Simple radiators already have an excellent balance of surface area vs restriction. They are also very cheap and easy to manufacture. I could see your design being a cool project for a mod, but it could never be a salable product. That's why they don't exist.

[Twigsoffury]

That's why they don't exist.

don't exist yet

[CorsePerVita]

This thread makes me want watercooling on my PC.

[dr.walrus]

I'd use a single machined manifold on each end that has four inlets and four exits.

There is a fundamental problem here that you're using not using a 'linear' system that goes cold>warm from left>right. The distribution of cold and warm channels means that the incoming water would be heated. Thermal conductivity? Amazing. Thermal performance? You're cutting off your nose to spite your face here.

Thats why i would use four single pumps in my design.

pump failure pump failure pump failure pump failure

but if you take into account that im reducing a larger volume several times, like home plumbing, when you force a larger volume of water through a smaller diameter tube large pressures are created in and of itself, and that pressure would be reduced back to the normal pressure by the manifold on the other side and return to 1/2" tubing or what ever you were to use.

I see what you're getting at, but this would still increase the pressure throughout the system...

Yup, better use some copper based anit-algae additive or use PPG polypropylene glycol but i'm sure the pressures created internally is going to blow any s@#t away that attempts to make a clog.

This would have at least 3 zeros on the price tag, I'm not sure I'd want to take that risk!

Can't exactly inspect the inside of the radiator you have now can you?

Yeah but at a few thousand dollars in cost, I'd be very unhappy if the response to 'it's blocked' was 'oh well'

But if its fractally based with manifolds on each side. if you had a gasket that covered each of the sections and the manifold attached on top of that (think of a F1 cars engines intake manifold) you could break it down and clean each of the sections then assemble it back together. it'd be a bunch of boxes inside a box inside a box inside a box. inside a bigger box with a manifold on each end.

The gaskets would just insulate each layer, making the whole design pointless!

exactly, now imagine how much thermal energy you could dissipate in a 1 foot long cubed heatsink with 256 tubes that have opposing cool hot flows 1/32"nd of a inch away from each other for the whole 9 foot trip through the heatsink. with two cools sides always touching the hot side. remember if you add all those 1/32" walls together i'd think they would add upto about 3" of solid material inside the cube. thats a s@#t load of cooling potential in itself.

I'm sure if you had some cool water already looping the opposing lines on the other circuits, you could cool off a 5 gallon container of boiling hot water in less then a few minutes.



Because my design has almost 44x the amount of travel and almost (lemme use a calculator) 1,648.374x the surface area in the same space as a straight through pipe.

and i can fit (calculator time again) almost 20x the amount of tubes in the same space as a traditional design with straight through pipes. because i can fit two or more tubes in the volume of space your straight tube occupies in that same heatsink. AND with interposing directions in that same exact space.

Bizarrely, this is the exact problem with the heatsink. If it does everyhing you say, it'll be far too good at transmitting heat.

Water would get even 1/10 of the way through the heatsink before normalising with the temperature of the heatsinks. The rest of the way through, it's not doing any cooling, it's just sat there at the same temperature. The only way to deal with this problem, and get the full potential out of this absolutely beautiful idea is with super high flow rates. And therefore silly pressures.

And what do you do with this enormous amount of heat you extract? You're going to need a BIGASS radiator and that would preferably have some auxillary cooling too. Else, you're just pulling all that heat out, and dumping it straight back into the system!