Best pattern when making waterblocks ?

[Snubben]

Yep, the crazy redneck from the north wilderness of Sweden is back!

This time I've got a waterblock problem.
I've built several waterblocks before, but noone with the demands of effectivity and size as this one.
It has to be able to transfer 300W (not a typo) of heat and can only be 4x4x1cm. The in and outlet pipes must fit on the same side (the 4x1cm side)

The ASCII sketch looks like this:

4cm
_____
| |
| | 4cm
|_____|
|| ||
in out



I've used zig-zag patterns in different variations in my previous waterblocks, but I don't think that's such a good idea in this case.

The thing that worries me is the duckfoot-effect.



For those of you who slept through the biology lessons, here's a summary:

How is it that warmblooded ducks can stand on icy lakes and don't freeze stuck ?
Well, nature has got a smart solution: The (blood)veins in their legs are wrapped around eachother and by so, creating a little heatexchanger. The warm blood coming from the heart flows through veins that goes parallell to the "return-veins" and the warmth is therefor transferred to the returning blood. (the blood in their feet keeps the same temperature as the ice they're standing on)



By creating a small waterblock where the coolant flows in a zig-zag pattern the duckfoot-effect might appear. Therefor I need a block where the heated coolant doesn't flow near the unheated water. At the same time I cannot have thick walls inside the block since I need as much water as possible to get rid of the 300W of heat.


The equation follows:


Need high water volume

Must keep the surface effect low
(that's the W/cm², in other words, must have a large surface of the waterblock exposed to the coolant to transfer the heat) I've reached 3260mm² as best in my calculations (that makes a volume of 22ml.)

Cannot be larger than 4x4x1cm



If anyone can give me some advise on how to solve this, please let me know.


It's not life-or-death-critical, the weather up here is quite good for overclocking right now, -25 to -30 degrees C this week. You just put you box outside in a pile of snow. ;)
No, seriously, I've been working on this cooling setup for almost 2½ years now. It's soon ready for testing.
I reached 2880W@24V TEC power yesterday, that makes a effective cooling of 1720W. (compare that to the CPU dissapation of about 100W)

If anyone is interested I can post some info on the project here (with some pictures perhaps).


//Snubben

[RainMaQer]

I have no idea... but I am interested in hearing some more about it:)

[Twinkle]

I have an idea. I'm not sure how good it is so I really dont know.

[Sonny]

Twinkle - Here you go:beer:

[Diggrr]

Hey Snubben,
Saw your problem. The waterblock I made has a pretty good solution. I whipped up a drawing for you.

The center section is not channels per say, they are 'v' shaved grooves. I had a dremmel not a mill, this is something I could make in the basement. Keeping the side channels the same area in cross section as the tubing, and the v grooves an 3.5mm deep, gives high flow rates and even water distribution. Also being cut with a dremmel left tiny "chatter" marks along the v grooves so there's more surface area than apears, and water cannot "adhere" slowing down the flow.

I used 1 cm copper stock for the bottom, and used lexan for the top, glued on with some marine goop. A copper sheet could easily replace it. The fittings are glued to the top on mine, I can't use side fittings.

Let me know if this helps....ciao

[Snubben]

Thanx for the ideas, keep posting more!! (I like to have a big smorgasbord to choose from :) )
Here's some info on my project:

It's designed to keep a low noise level, preferably below 20dBA.
My first thoughts was that I'd use a compressor from a freezer to chill a coolant pumped through a waterblock cooling a couple of TECs mounted on the CPU, but that idea was kinda crappy, it would've made a noise lika a bulldozer and the compressor was expensive as hell.
As you might guess, the project has transformed a bit on the way, but now I've gone along with the same setup for 6 months and it seems to be a good one.
It's based mostly on TECs, 172W each, mainly because I got a good price and that they're quiet, small and effective.

The setup is a little complicated, but I'll take it step by step.
It's a two-circuit system where I use two coolants, water and a 50/50 mixture of water and methanol. (the water/methanol coolant is chilled to zub zero temperatures)


The flow-scheme looks like this:

50/50 coolant flows from insulated tank -> pump -> BIG alu-waterblocks with TECs -> 4x4cm waterblocks -> back to tank

The big alu-waterblocks are actually air-heatsinks with a aluminium cover. At the present they're 2 blocks of them, each measuring 20x10x4cm with a 1cm thick base. There's 3 172W TECs mounted in each block (1 block = 2 sinks & 3 TECs in between)

Cold side
[|||||||||] <--- Alu heatsink
--- --- --- <--- 3 TECs
[TTTTTTT] <--- Alu heatsink
Hot side


The TECs are cooled by the standard water circuit.
By this setup I can increase the cooling power by just adding another big alu-waterblock and take the 50/50 coolant down even colder.

On the CPU (seems like it's going to be a P4 1700 s478) sits a 4x4x4cm massive copper cube wich has one 172W TEC on each side (not top and bottom, ofcourse :) ) and some holes drilled to measure the temp.
It's here the 4x4cm waterblocks fits in. One on each TEC.

My 172W TECs have a Delta T of 69°C and needs to be chilled to sub zero on the hot side to be able to preform what I need.

All the parts that will keep a temperature below room temp. will have proper ESD-safe insulation.


Now you're probably wondering how I'm going to keep the water-only from boiling.. -Well, I haven't been able to test it out in the real wolrd but my theory goes like this:
The water is going to be pretty hot when it comes out of the alu-block so I was thinking that a bong cooler would steam it off a bit, when it's passed the bong, I pump it through a big radiator (with 6 silent 120mm fans mounted) and back to a reservoir.

Hope you understand, I'm terrible at explaining things.


For those of you who counted it adds up to 2880W@24V (and maybe more if I add more alu-blocks) and that's nothing a standard psu will be able to handle, so I've built my own powersupply from a used 3-phase MIG welding transformer and a couple of power capacitors.

It maxes out @ 315A, 24V (the voltage is adjustable) But it has some bad sides too, it draws 32A @ 400V (thats 12,8 KW)
I'm designing a soft-start circuit for it now wich will make it run on standard 16A fuses.
Another bad thing is that it's not a switched psu, but an analog standard transformer, and you probably know that a transformer needs a soft iron core, and so does this. A 300 kg core to be exact!

The whole pakage consists of 3 parts, a selfmade chassie measuring 100x100x70cm, the 325kg psu and the insulated 50/50 coolant tank. The bong, radiator, alu-blocks and water-only reservoir fits in the chassie along with the cooling-mointoring-system and the standard computer-stuff (HDDs, CDR, DVD, psu's...)

When this project passed the 350kg-barrier last year I realised that I wouldn't be able to move it by car, so I bought an old 1960s swedish lorry with a hydraulic crane on. (cost less than the computer parts). With this I'm able to move the whole system, now totally weighing ½ ton.
Since this computer (or what to call it) is mostly going to stand as a server and not be moved around, it works good anyway. (concrete floor recomended)
The only time it's going to get moved is when I travel to LAN partys. (2-4 times/year).


Well, that's probably it, I'm sure I've left out some parts and that this description has a lot of holes in it, please feel free to ask me if there's anything you want to know.

I'll post some pics later on, anything special you want to see?


//Snubben

[hunter1895]

huh ?!

[Snubben]

I've looked at your suggestions, mixed them with my own experience and come up with a prototype.

It's a block made of an 8mm thick copper plate with 4mm wide and 6mm deep coolant channels in the pattern you see in the picture.
Each corner will be rounded (not sharp like in the crappy picture) due to the milling process.

I will then solder a copper lid onto the plate, sealing it tight.
But in the testing phase I'll use a plexi cover
so I can monitor the flow.

I'm a bit worried about the place where the coolant passes the middle of the block (top of the pic), it's not symmetric there, but perhaps it doesn't matter much.


The calculated specs:

Total surface exposed to coolant: 33,6cm²
Coolant volume inside block: 5.472 cm³ (ml)
Coolant must absorbe: 8.93W/cm² or 54.82W/cm³


I'd appreciate if you could give me some feedback on this, I'm also looking for some formulas on how to calculate wich flowrate is best for a particular case.

[DarkInferno_IV]

OMG!

So my system finally is toppled as "Most Ludicrously Cooled".

Mine=
Water flowing from mains through small freezer with pipes through, full of solid block of ice. Through a dual 220W chiller cooled by copper blocks feeding straight off the taps, onto a custom silverbase block with 3 220w TECs on the AthlonXP. Then to the drain.

Current speed:

15*193 @2.7V

A grand total of
2.895 GIGAHERTZ!

Seriously, (with good DDR) you will hit LUDICROUS speeds (3GHZ +).

However, I am making a QUAD 220 chiller, using the old dual chiller parts, and getting a 4th 220 TEC for the block. Voltage will soon hit 3.1V....

Y'all think I have a chance of 3 GIG off an 1800+?

[RangerJoe]

HOLY $H!T!!!!!....thats incredible, what kind of 3dmark 2001 scores do you get with that, or better yet, what are your sandra scores for that beast. what are you temps also...thats is just incredible.

anywho, on the front page, under the watercooling category, this PhD in fluid mechanics i think made up this block....id say its a winner....

[DarkInferno_IV]

2C under load, -14C at idle.

You must be amazed how much heat a 2.7V CPU can put out.
That's why I am planning the quad 220 chillers and 880W waterblock (!). Especially runing a 1.85V max CPU at 3.1V.

I'm getting bad benchmarks since I have to ****** all the settings, even disable the L2 cache to get these speeds. Benches like a standard 2000+ right now. I'm just aiming for the 3000 (15*200) POST screen......