block designs

[Volenti]

dual block in action;

[Jessfm]

Originally posted by gone_fishin


Hi, Jessfm I've seen your block a while back. It is hard to imagine even flow throughout that design. What is the length and width of it to get a size perspective and your thoughts on the apparent slow zone behind the exit wall? (Don't take it the wrong way, you must have come across this question and if not you soon will)

Size is 60x50mm

I will get some pics of the latest version, it tries to help 1: even the flow 2: make use of more surface area of the wall. 3: creat more turbulance on WB floor.

I would have liked to have made the block slightly different - but we are limit4ed by the endmill size - too small means slower times & higher expense. This will be a mass production unit so realy does not belong in this thread - just wanted to show where my line of development is taking me - may help give others some ideas

[Sanjuro]

Hmm.. i have to say i like brad's design the best

[Brad]

Volanti, are the two inlets at the parrallel port end, then two outlets at the ddr end? or two inlets at the chipset side, then two outlets at the top of the motherboard side?

My guess would be that two inlets at the parrallel port end and two at the ddr end would be the most effective.


BTW, thanks sanjuro

[Volenti]

Volenti, are the two inlets at the parrallel port end, then two outlets at the ddr end?

That is correct. When I was testing the blocks flowrate I noticed that the water exiting the block was... "rough" for want of a better term, presumably I'm getting a fair bit of turbulence inside, with a flow of 1500L/H (400GPH) it had better be turbulent

this will be my "testbed" block, since I can easily un-solder it and expriment with various geometric shapes inside.

Though I want to refine the entry and exit to be more flow friendly than the current setup, see if I can eek some more flow rate.

[bigben2k]

Originally posted by morphling1
Ok, not 5 but 4, 3 axis for x,y,z then one so that you can turn milling bit to be perpendicular to the surface of the cone.

And here's innovatek :

How deep are those grooves? In other words, how big is the central post?

[maskedgeek]

heres one that i designed up and if the person who is making it wants to chime in on this go for it, i wont mention the name....

but its got 2 inlets and 1 outlet, itll be a super fast spinning tornado block... the cone in the center rises up to the top of the block and the top has a cone too so i dont loose flow

[morphling1]

Bigben2k, I don't have exact dimension, but if it can fit on socketA and not using 4 holes for mounting, I would say black cube is 50*50mm square, so you can rough dimensions from the pic.

Maskedgeek, you want lose flow, but you will lose velocity of the flow, because cross section area is getting bigger when moving from center of the cone out.

[bigben2k]

Thanks morphling, but I was asking about the copper part!

As for masked geek's design, considering that it flows upwards (not downwards, morphling), I think it could be interesting, but I believe that it would perform better in reverse. Here's why:

I don't believe that you are likely to achieve a fast tornado in this design, but if you do, it would greatly increase the flow restriction: a circular motion (tornado) will have a tendancy to throw the coolant on the outside, due to centripedal force. Also, if there is a fast tornado, then you are inducing a further amount of friction, which will also increase flow restriction. I think that what you are more likely to achieve, is a whole bunch of small eddies, around a very slow tornado: the water may not even achieve a full rotation before exiting the block, it's going to depend on the flow rate.

If you want to force a tornado, then you're looking at a spiral design. Around a cone like yours, that's what morphling was thinking, but then he realized that it would required a fancier CNC...

[bigben2k]

Originally posted by morphling1
Bigben2k, that base thickness is for waterblock with straight chanell design, and it doesn't aply for every design out, I can only say one thing did you ever look at innovatek wb, check it out and you'll see that you've got things wrong, btw it outperform prabably most blocks out there (comercial) except maybe some diy blocks.
#Rotor, there is no gap between the block and the cone, I was also thinking about making chanells a little spiral, but how on earth could I do that, for something like that I would need like 5 axis cnc mill
Jessfm, that block looks awesome too, great job m8

Morphling, on second thought, that block may work extremely well.

Going over some reading, it's clear that your pyramid is approximating the heat dispersion pattern of a solid block.

If anything, I'd consider a half sphere as an alternative, with fins and all... (Is that a good machining challenge? )

[Fixittt]

I too like morphs thinking on that cone.

Only I would have went a bit different on the fins. More of a curve. but still I like the way the block will target the heat. What would be interesting. If the copper cone is a tight press fit to the aluminum hoising. Is take a very small heatload. centered under the copper slug. And take tem readings on the copper core and the aluminum as close to the core as possable. Take the temps at small intervals and see if the copper heats up faster then the aluminum, or see if the aluminum draws the heat quickly from the copper. im still not sold on the copper is better then aluminum stuff, but hey thats just me. Then when temps stablize. heat sateration. See which material cools off faster.

hey Morph if U cant get the CNC code for different fins, maybe I can pass it on to a buddy of mine and see if he can get the code setup for ya. just an offer.

[morphling1]

Bigben2k, I know you we're thinking about copper slug, but if the black cube is 50*50mm, I would say that the valve is aprox 30mm in diameter and each fin 2mm thick and 2mm deep.
Fixittt, all I can say I'll try.
Btw. I thinking of buying used mill, and then put stepper motors on hand wheels, to make me a cnc mill, could you tell me what kind of steppers do you have on your desktop, so I'll know what's the minimum I need. As for controler, I saw one for ~140$ for controling 3 motors up to 2.5 amps, and for software I was thinking turbo cnc, shareware program, I already work with it, and it's pretty decent with lots of users getting feedback for making it better and better. What kind of software do you use?

[bigben2k]

Fixitt: why curved fins, versus straight down? If you're thinking about turbulent flow, then you can achieve that by simply applying a higher flow rate. A curve would only add to the flow restriction :shrug: .

Morphling: What I wanted to know, about that copper slug, is how deep those grooves are, or, how big that center post is. You really can't tell from the picture, the center post could be only 5mm thick, or it could be 25mm.

[myv65]

Fixitt,

You say you don't completely buy into the copper better than aluminum thing. So long as you have access to solid copper and solid aluminum, I've got a couple experiments for you that may help.

First, let me state the common misconception and why it arises. Many people believe that copper is more effective at absorbing heat while aluminum is better at getting rid of it. The simple fact is that materials simply move thermal energy from point A to point B based on temperature differentials. They don't care whether point A or B is warmer, they just move heat from high temperature to low temperature.

The biggest reason people seem to be confused by this is that these materials behave differently under transient conditions. Aluminum actually has a lot higher specific heat than copper. This means that mass for mass, aluminum holds thermal energy a lot more effectively than copper, to the tune of 903 Joules/kg-K vs 385 Joules/kg-K. This means that 903 joules of energy (.856 BTU for us Americans) would raise a one kilogram block of aluminum by 1°C. The same amount of energy would raise a one kilogram block of copper by 2.35°C. Hmmm, sounds like aluminum might actually absorb heat better, right? Well, this is offset by aluminum's lower density. Aluminum's density is 2702 kg/m^3 vs 8933 kg/m^3 for copper. This means that if you take two pieces that are the same size, the aluminum one will weigh only 30% what the copper one weighs. Stuffing that same 903 joules into two equal size blocks would raise the temperature of the copper one by only 71% as much as the aluminum because of copper's higher density.

OK, what's this mean in English? It means that on a volume for volume basis, copper will store heat more compactly than aluminum. This heat storage means that copper will heat more slowly when you turn on your computer and will stay warm longer when you turn off your computer. The real question is what happens in between when the computer runs steady-state?

This situation is perhaps the most confusing of all. Copper has a higher conductivity, 401 W/m-K versus 237. On a purely volumetric basis, copper requires less temperature differential to transmit "X" watts of heat from point A to point B. This is why you'll find it in all high end heat sinks where the sink meets the core. Now, why do many high end heat sinks then use aluminum fins or pins? The density once more plays a role. Getting rid of heat requires convection to air. Convection requires surface area. Aluminum's lower density means that you can make a lot more surface area from a given mass of material. So where convection matters, aluminum has an advantage.

OK, I mentioned some experiments that you can do that may help. What you'll need are two blocks of material of equal dimensions, one of copper and one of aluminum. It's also handy to have another block of identical dimensions made from wood. Place all three blocks in the fridge for a few hours until you're sure they are the same temperature as the fridge interior. Pull all three out and quickly place them on your forearm. Pay attention to hold cold each one feels and for how long. The copper will feel coldest for the longest time. The aluminum will feel cold, but not as long. The wood will scarcely feel cold at all. The sensation of cold comes from how quickly the blocks are transferring heat from your forearm. Copper does the job best of the group while wood does the job poorly.

Take these same blocks and put them in a kettle of water on the stove using the lowest setting. Let them sit there for a couple of hours. Test the water to make sure it isn't too hot to touch. So long as it isn't, take the blocks out and put them on your forearm again. The copper will feel hottest the longest, then the aluminum, then the wood.

Each of these two tests shows the transient properties of the materials and represents what happens when you first turn on (cold test) and finally turn off (hot test) your computer.

Now to know what happens while your computer is running, put a kettle of water on the stove with a lid on it. Set the temperature on the stove fairly low and place the blocks on top of the lid. Wait a while and touch the top of each block. The copper should feel the warmest, the aluminum almost as warm, and the wood probably cool to the touch. This means that copper is getting heat out of the lid most effectively and would make the best material for getting heat off a core. If you think the aluminum would be better because it doesn't feel as warm, then why not use the wood to make a heat sink?

In the end what you'll find is that for identical shape blocks, conductivity rules and copper beats aluminum. For identical material masses, the decreased density of aluminum allows so much more material. This excess of material means a lower overall thermal resistance in conduction and copious surface area for convection. Air heat sinks and water blocks are constrained around the core such that you can't use aluminum's lower density to put a bigger hunk in there. This is why copper rules for inserts and water blocks. In air heat sinks you tend to have all sorts of room for fins/pins, so aluminum does a better here for a given weight.

Honestly, the differences in water blocks are pretty small and you really shouldn't see more than 1-2°C difference between aluminum and copper. Rest assured, however, that for two water blocks of identical geometry the copper one will always win.

[bigben2k]

Well put!

I think morphling's got it right, with a copper cone, and the rest being aluminium, since the heat source would come through the copper. The Alu is actually inconsequential, and this block might do well with any other material than Alu: we've all seen that in those fancy german blocks.

[bigben2k]

BTW, for those interested, silver's properties are:

-heat capacity: 235 J/kg.K
(energy, in Joules, required to raise a 1 kg block of silver by 1deg C)
-density:10490 kg/m^3
-heat conductivity: 429 W/mK

So to recap:
Aluminium
-heat capacity: 385 J/kg.K
-density: 2702 kg/m^3
-heat conductivity: 237 W/mK

Copper
-heat capacity: 903 J/kg.K
-density: 8933 kg/m^3
-heat conductivity: 401 W/mK

[myv65]

Originally posted by bigben2k
Well put!

I think morphling's got it right, with a copper cone, and the rest being aluminium, since the heat source would come through the copper. The Alu is actually inconsequential, and this block might do well with any other material than Alu: we've all seen that in those fancy german blocks.

Probably a no-brainer in this crowd, but won't hurt to state the obvious. You generally don't want to put aluminum and copper in close proximity. Sure you can use anti-corrosive additives or anodize the aluminum, but it's nice if you don't have to bother. Also worth noting that anti-corrosive additives do no good when you have metal-to-metal contact. Here a gasket or anodizing is a must.

[Sanjuro]

Originally posted by morphling1
Btw. I thinking of buying used mill, and then put stepper motors on hand wheels, to make me a cnc mill, could you tell me what kind of steppers do you have on your desktop, so I'll know what's the minimum I need. As for controler, I saw one for ~140$ for controling 3 motors up to 2.5 amps, and for software I was thinking turbo cnc, shareware program, I already work with it, and it's pretty decent with lots of users getting feedback for making it better and better. What kind of software do you use?

I can answer a bit of that since Fixittt already has so many questions to answer Anways really depands on the size of the mill and the desired rapid speed you wanna achieve. I think all Maxnc mills (fixittt's) come with Pacific Scientific Powermax II steppers with shaft encoders on the CL versions. Running bipolar you get about 200oz/in from these, and they are pretty strong for a little mill. Normally just look around for 100oz ones with double sided shaft (so you can get your handwheels on there)

Bipolar is stronger but you'd need drivers that support it, and its more complicated thus usually more expensive. Unipolar on the other hand is very easy to drive but weaker. I think all drivers these days can do both.

[Fixittt]

Ok, first off... Morph. check this link

www.hobbycnc.com all sorts of stuff there.

also
www.desktopcnc.com


also.....

I have taken 2 identical waterblocks. Both Original Spir@ls

one of copper, and one of aluminum.

And with the same setup...... I got the same temps. using a temp probe touching the die.

Now this is prolly due to the fact that the system may not be tuned exactly right. Example (I have a small rio pump in a res)

so whjy would I trash endmills, and waste 3 more hours machinine the copper when I can machine the aluminum in 1/3rd of the time?

[myv65]

Fixitt,

Believe me, I understand your preference for aluminum. I agree with you 100% that it makes a lot more sense to mill water blocks from aluminum. I was only explaining the differences between the two materials from a performance perspective with no consideration to cost or machinability.

To me, the slight gain that you will get from copper is not worth the hassle, but there is a slight gain.

Anyway, I am curious what you think of the explanation and homeboy testing that I suggested.

[Sanjuro]

would a bigger load say a 226w pelt show the difference between al and cu?

[bigben2k]

Originally posted by Sanjuro
would a bigger load say a 226w pelt show the difference between al and cu?

It's really going to depend on the cooling solution that you implement. More specifically, the flow rate is going to be critical.

If you're asking which would perform better, everyone will agree that copper does.

If you're asking how much better, then that depends on how you set it up. If you use a rio 180, then yes, there will be a great difference. If you use an Eheim 1250, you might see a few degrees. If you use an Iwaki, then you might see 1 deg C.

[invariant]

I've been thinking about a simple block design that might work well. I wanted to maximize tubulence in the velocity normal to the base, and thought perhaps mechanical forcing of the fluid might be the way to go. Analogous to wind flowing over a steep mountain.

The base and walls are just a simple box, but the top has what is essentially a huge "speed bump" ( or series of obstacles) across the middle of the top, over the core. The inlet and outlet are on either side of this mechanical obstacle. It wouldn't really matter if the top and walls are constructed from plastic or metal. It would be dirt cheap to produce

.

[Cova]

Ok, IMHO this block should perform pretty damn well, if it can be built. The idea being to have really high flow, with enough surface area coming off the CPU core to transmit heat away from the core fast, and yet as little copper as possible between the core and the water, so that the temp delta is as low as possible. Think along the lines of Volenti's dual-block and you should get the idea of what I'm thinking. Except with just one inlet/outlet (the larger diameter the better, as much flow as possible) to keep it simple.

Then again, I have no idea how you'd ever machine something like this out of the copper base.

[Cova]

This might help a bit... I want one of these, on a smaller scale, with water flowing through it horizontally. Of course, it would be positioned right over the CPU core. (yes, image blatantly stolen from that site you all hate, as the logo in the corner suggests - there, I've given credit where due).