Cooling Water Block Design

Diamond block could be grown to shape. There was a method for diamond deposit used to form diamond speakers cones. It'd still be expensive.

 

No offense, but adding random features doesnt make a good block. The key to a good waterblock isnt the surface area. Water is a poor heat conductor. A good block design should minimize the flow boundry layer where the fluid is moving slowly. Yeah, I know it's confusing. To fully understand everything, you would probably have to take 4 or 5 college courses. Even after that you would need to do a lot of modelling and testing to make sure your block works well.

Anyways, I don't recommend you making your own unless you are cooling a tec or something.

 

I never said I was going to mill them, I just wanted to know what makes a good water block

 

If you do a PhD in the subject you might be able to answer that. I've only got a Master's degree in Aerodynamics/Fluid Mechanics and I can't answer that question in any but the most general terms!

 

Both Cathar and DangerDen have produced ranges of blocks with silver baseplates... you just have to use the right silver and the right machining processes.

 

Aquacomputer used 0.925 Sterling Silver on their silver Cuplex blocks.

That all I know.

 

I suggest removing most of that. If you want a good waterblock, you want to minimize restriction, reduce laminar flow, and focus mainly on the center, where the die is. Try Jets or flow nozzles, and a standard pin grid array.

 

thanks

 

Personally i think watercooling had reached its full potential. Any further improvements would require resources that most companies don't have. I do have few ideas in the back of my head. Maybe you can look into them if you are serious about this.

First idea, coolant! Like I said before, water is not good at conducting heat. Using a better coolant can reduce the THERMAL boundry layer and therefore improve your waterblock performace. Yes, mecury is probably good for this, but you will poison yourself for sure.

Second idea, break the boundry layer mechanically! Yes, that means a waterblock with moving parts!

 

What about adding divits line on a golf ball to the larger flat areas, this would increase surface area on those areas where you don't want a lot of pins.

 

Murcury isn't that bad. As long as it's contained within the system you're safe. If it vapourises or combines with an organic element, then you're in trouble.

 

If your flow rate is above zero, thermal conductivity is irrelevant to performance - the water isn't in the block long enough, and the turbulence is so high as to flatten out any thermal differences in the working fluid almost instantaneously.

What exactly, pray tell, is the water doing as it moves through the block???

 

Watch out for it forming intermetallic compounds - IIRC Mercury is bad for that, which means it might start eating your blocks if you use the wrong material.

 

If you're thinking about milling, try to think about how the mill works. It uses a cylindrical cutting tool (or ball ended), so imagine drawing a cutting path with a round marker. That's how it'll cut, so certain sharp corners on the inside are not possible. You can get sort of sharp corners, but you'll have to use a tiny tiny bit (fragile and flexes), and it'll be excruciatingly slow process.

There's a number of ways to get effective cooling from a waterblock. Maximizing surface area is one method. Imagine the waterblock as hollow where the metal is. and then imagine heat as like pancake syrup slowwwwly filling into the hollow. The larger the cross section of the fin/pin the less impedance the heat has to travel, but the thicker your fins/pins the less surface area you have to work with. So it's a balancing game. Someone touched on jet impingement, which is another method. Combining the two works too (i.e. a plate with tiny holes drilled into it, set above a patch of pins). You can play with materials, coolants, etc. but like I said, in the end it all comes down to balance.

I had a design that I dumped a while back that had high surface area and used jets not for impingement, but for a nozzle effect to create turbulence. I could find the drawings and show them to you, but it'll take me a bit to find them (forgot where I put them). It was on the heavy side at a pound of aluminum.

Mercury is more trouble than it's worth. It's toxic, it's very heavy, and it will destroy most pumps. You'll need a special pump to pump mercury (read expensive). Pure water is actually the ideal coolant (good thermal capacitance), though it won't stay pure when in contact with metals, so you'll have to settle for something slightly less than ideal.

 

I am a mechanical engineer myself. But fluid/heat transfer isn't exactly my concentration, so correct me if i am wrong.

I think you are talking about the AVERAGE flow rate. Just because there is water flowing out of your block, it doesn't mean there is flow everywhere. If you look at the flow profile, the fluid right on top of the surface has little/zero flow. In the case of water, it's like have an insulating coating! Now don't get me wrong. I am not recommending mercury. I was just using it as an example.

umm... look up boundary layer.


Again, surface area isn't everything. Think about storm and apogee. Apogee has much more surface area than storm, but storm is still a better waterblock than apogee.

 

No, I was talking about the specific flow rate at the hot points within a waterblock. Even a boundary layer is moving (albeit slowly) right down to within a few hundred molecular thicknesses of the metal of the block. Furthermore, the stagnant layer is so thin that you are unlikely to get measurable performance gains by changing the conductivity of this layer.

I've spent enough time doing boundary layers that I could probably give a lecture on them (in fact for a potted lecture, see the "so you want to watercool your PC" sticky at the top of this forum). The point is that any mechanical device capable of achieving a thinner boundary layer than modern impingement blocks will have to have clearances smaller than that boundary layer between the mechanical agitator and the block surface. That means that not only are you getting into precision engineering (at best very expensive machining, more likely grinding or even match grinding - the typical 50 micron tolerances you get from CNC machining aren't nearly good enough), but the working fluid will be kept out of contact with the block surface for a substantial period of time - thus making the situation worse than before.
I would also argue that impingement blocks ultimately are using a "mechanical device" to break up the boundary layer - in the form of the combination of pump and nozzles.

I would also question whether storm (in either the Cathar original or Swiftech Rev 2 variants) is in fact a better waterblock than the Apogee. Apogee appears to operate at a lower pressure drop with very similar performance per litre/sec, implying that it will probably perform better for most pumps.

 

Hehe, I know a few processes that can handle these tolerances no sweat, but they don't involve any machining It has been a while since I learned about nanofabrication so I don't know if there is either a wet (standard chemicals) or dry (plasma process) etch for copper at this point, but there are some other exotic methods such as ion beam milling if we want to explore the realm of the 'possible but not likely'.

nanofab+fluids=micro fluidics

This is a fascinating topic but I'm pretty sure someone has already looked into this

 

hey, a lot of this stuff is over my head, but couldnt you make a design where the water comes in from the sides of the block and passes directly over the Core and out the other side, wouldnt that let you make a design that is more "tube-like" and eliminate a lot of the turblulence issues? so it owuld just go through a slightly widened tube and pass over some fins and then funnelled out the other side?

 

Yes, that was the design I mentioned. I found a pic too:





This one is a bit old, there were several design revisions aftewards to the internal components, but the exterior maintained it's simple high-pressure-viewport look.

Problem with the design was that it worked with very few motherboards, and clearance was always an issue.

 

Problem is, are they affordable? In reality if you want to sell reasonable numbers of these blocks and still make a profit then you have to be able to make the block base plate (including any moving parts) for under $30 maximum, and do so with order sizes that in industrial terms are tiny. That alone limits you to manufacturing processes costing about $10 or less in true mass production. I'm not well up on etch processes (everything I do at work involves CNC machining or grinding - but we do work down to sub-micron tolerances on some parts), but I'd be very surprised if they can be done that cheaply.
If they can't, then we're back to CNC machining - where typical tolerances are 50 microns but a good supplier might manage 20 microns.

Thing is, turbulence is a good thing rather than a bad thing - in and around the hot bits of the waterblock you want as much of it as possible, as it aids cooling (take a look at the "so you want to watercool your PC" sticky if you want to learn more). That's why channel flow blocks have gone out of favour in recent years for CPU blocks.