Krakatoa - GPL DIY Waterblock Design

[Captain Slug]

I'm working on a waterblock design with the intention of it being relatively easy to make (I.E. not requiring any milling), but still having competitive performance.

The design is sort of a crossbreed of the RBX and WhiteWater designs because this staggered drilling pattern offers the best fin density and turbulent surface.

The base is 3"x2"x.188" with .125" (3mm) channel depth.
The top is a sandwhich of two .188" thickness polycarbonate or polyester (PETG) plastic sheets. This will make the top extremely easy to machine because they can be cut out to square dimensions on a tablesaw, drilled, and then tapped before being solvent-glued together.
After finding sheets of plastic in the right thickness I realized that I can make an o-ring groove without having to mill it out of either piece. The outer lip of the top is then glued on and the inner lip bolts into place from underneath using countersunk screws. This will allow for different nozzle designs to be added by simply swapping out that piece. Or alternatively it could be glued in place instead.

Thoughts? Once I've made enough revisions and made a prototype I will make a complete set of cutting templates that you can print out on label paper as well as some instructions.

[bobo5195]

that looks like it would take ages to make.

I think that more could be done to make it better.

I need to post an artical on turbulence. Turbulence is best avoided (unless you have good reason) as the performance gain will almost never be worth the head loss.

[Captain Slug]

Ages? The top pieces are easy to make with a scrollsaw. The base would take time, but not any more than the #rotor would take to make.

Base revision

[bobo5195]

I think you could make a storm fairly easily if you wished using ickle power tools.

[Captain Slug]

Yeah, if I had an easy way to make nozzles. The only "easy" method I've seen is steel capillary tube.

I'm contemplating just switching to a micro-channel design by cutting the channels out of the base using a bandsaw or hacksaw instead of requiring a heap of drilling and dremeling. This design is $5 cheaper per block in terms of materials and will be easier to machine.

Again, the center inlet is optional so the design could be changed to a 2-barb design by simply not machining the center barb hole out of those two plates. The top is made from layers of .118" thick 2-inch wide polycarbonate sheets.

[pH(x)]

But that wouldn't be your original vision, which I think in all fairness would be much better given a high-flow system

My WhiteWater block from Cathar definitely gets me some excellent temps on my OC'ed P4 -- lower than I could have ever dreamed of. You see with my high-flow Danner Mag-drive system it allows me idle temps of 22C on a 700MHz overclock

[Captain Slug]

Well, the ability to not have the center inlet is there if you don't have a high heat-load setup or a powerful pump. But it will be there on the prototype because it's definitely a more effective setup.
This block should be so easy to make that experimenting with different nozzles and top configurations would be a piece of cake. The material I've put on the part list for this block comes in 2" so you only need to cut the different pieces to 3" length and then drill/cut the pattern into it.
Here's how you could make it into a white-water clone by simply adding another piece.


This construction method is proving very interesting and will allow for very complicated top without the need for milling. If I can figure out an easy and cheap way of adding nozzles I could make a dirt-cheap storm-esque block too.

[pH(x)]

So you're planning to do a bit of a mass-production with your block? If so, then I can acknowledge your notion. Good luck with your project!

[Captain Slug]

Well either I can make them in batches (no big deal considering how stupid-simple it's going to be to make) or just make a detailed set of plans to allow others to make them.

[Captain Slug]

Here's the final revision of the CPU waterblock. Same configuration as a white water.

I decided that it would make perfect sense to make a similar GPU/chipset design that make use of the leftover materials from the CPU block design. It uses the same channel cutting method as well and the top plate has all the mounting holes needed for CPU or northbridge mounting.

[Captain Slug]

2 other CPU block configurations that, although requiring twice as many cuts with the bandsaw/hacksaw, will result in a block that costs only $5.50 in materials.
Apogee look-alike

Nexxos XP clone

Not sure which one would be better. I'll just have to make all three to find out.

[pH(x)]

Have you tried a lateral entry design? It may be a bit harder to mount, but you will see a substantial decrease in resistance pressure over convnetional designs.

[BGP Spook]

I have mused over just such a concept.

A design I have spent some time considering would involve a nozzel which would widen and flatten to feed the water uniformly over a field of pins.

But, in my mind, I doubt it would:
1. perform very well,
2. reduce cost,
3. ease construction,
4. reduce flow resistance,
5. be possible to mount(ok, possible but it might be comparable to pulling teeth)

A single round inlet would be easier to make, and might perform a little better but that would depend on the design of the rest of the block.

But here again, in my mind, I doubt it would:
1. perform very well,
2. reduce cost,
3. ease construction,
4. reduce flow resistance,
5. be possible to mount(see caveat above)

To me, the easiest and most practical way to facilitate lateral flow would be to use an L-fitting, but to do so would likely mean to give up more ground(with respect to resistance) than would be gain from lateral flow.

Another possibility, I have considered is 1/4" tubing which would allow all sorts of ways to snake around most anything that would interfere with mounting; but again to do so would likely mean giving up more ground than would be gained from lateral flow.

In my mind, we are then left with either fitting a custom block to each individual model motherboard/system or reducing the physical size of the waterblock.

Again, this is all in my mind so take all of the above as you may.

Is my analysis flawed or does my reasoning sound plausible?

[Captain Slug]

Lateral entry is not at all practical space-wise for every socket except 939 and 940. I have seen one DIY block made with side inlet and outlet (it was a pass-through design with 45-degree elbows) but he wasn't aiming for top performance.
It has no benefits aside from lower restriction. It would also complicate machining, which is not something I'm too keen on doing.

[Captain Slug]

I have the first base finished and I'm halfway through making the first set of plates for the top.
Prototype images to be posted tomorrow when I can get access to a decent camera. Milled copper is impossible to take pictures of with my cheap camera.

[Captain Slug]

[Dryadsoul]

Wow, looks longer than it did in sketchup.

[Captain Slug]

Well, it took me long enough to get around to it. The top is now glued together and the block is finished. I just need to cut an O-Ring for it and then somebody can test it.


I decided to forego the interchangable nozzle plate. It's an option that will be included in the plans but I didn't see a real need to include it in the prototype.
And there you have it. The $8 whitewater clone.

[modster]

nice work. did you use a cnc machine?

[Captain Slug]

Nope, I used a manual mill to flatten the outer perimeter of the base, then I cut the channels using a bandsaw. I simply made a wood bracket to hold a cross-slide vice in a vertical position.

I then put the copper in the vice and cut a channel to depth with the bandsaw, moved the vice over the correct spacing, then repeated the process until I had all the channels cut.
It worked out really well and I managed to make the base after only 2 hours or total work.

[Captain Slug]

Here's the cutting template file.
http://www.captainslug.com/modding/krakatoa.doc (144kb)

And here's the materials list for making the waterblocks

8574K24 - Polycarbonate Sheet 1/16" Thick, 12" X 12", Clear = $3.51
8574K26 - Polycarbonate Sheet 1/8" Thick, 12" X 12", Clear = $5.50
8964K415 - Alloy 110 Copper Rectangle 3/16" Thick, 2" Width, 1' Length = $20.83
9407K11 - Buna-N O-Ring Cord Stock = $1.00 for 5 feet

To make the two O-Rings you simply need to cut two 7.5" lengths from the stock and glue the ends together using super glue. The pieces go together like so.

Use IPS Weld-On #3 or #4 to glue the 1/16th" thick pieces to each other (excluding the nozzle plate if you want to be able to remove it). And similarly glue the 1/8th" (or thicker) together. Add the O-rings and then bolt the unit together.

The design is meant to be adapted to what tools you have available.

[DryFire]

1' of the same copper bar is $15.34 at online metals, dunno how shipping changes the cost though.

[Captain Slug]

0.1875" thickness x 2" width x 1 foot is $23.80 at online metals.

If you buy just 3 inches from them it's $8.31.

With copper I'd rather have some leftovers since it costs a bit to get it shipped.

[DryFire]

ahh wrong thickness on my part.

[bigben2k]

Welcome back, Captain Slug! Hope all is well with you now!