Direct-Die water cooling

[Razorfish]

Hey gizmo, ive just been give the nod to design and build a "Custom water cooling system" for a PC, its for my college final projuect and will be funded by the lab i work at! After reading about you successes (and problems) i am serously concidering making it direct die! Oh and well done my friend, what youve been doing is very impressive, i bow before you SIR!

[Gizmo]

Thanks, RazorFish!

I'm in Atlanta next week, but I hope when I get back I'll be able to take pics of my new WB design made from copper instead of plexi.

[Razorfish]

That would be great, beliveit or not ive got 6months to produce a full project on the subjest so im sure ill be asking you for some pointers!

[Chernobyl]

Hi Gizmo
Excellent project!

A friend and I have been toying with making a direct die cooler too. I found this very useful thread as the first link on Google you will be happy to know
search phrase: "direct die" water cooling

An observation that may improve performance is pressure at the CPU die.
The more pressure, the more heat can transfer in the same time.
Looking at your block, the cross sectional area of the pipe feeding the cpu is way less than that of the container feeding the outlet pipe. This allows the pressure to drop rapidly across the die and reduces the overall max pressure at the die.

An idea we had:
1) Use thinner holes in the plastic block to increase water speed inside the block and raise water pressure at the CPU die.
Feed water directly down as you are by drilling a hole in the dead centre downwards.

2) Shape the die area for the cpu.

3) Drill a hole from left to right at near cpu height intersecting with the CPU hole:
I recommend making holes from the left and right of the cpu, slanted downwards towards the CPU so you dont risk breaking through the bottom of the plexiglass etc.
For best results, it would be nice to have even more water pressure by making the 2 water exit holes at the die even smaller. To do this, drill with a smaller drill first to get the small hole at the die, then use the bigger drill from the 2 edges til you are close to the die hole area. Using this method, the bigger drill can be as large as your feed pipes, helping system flow a little
There is a trade off with pressure and water speed as you will know. Only an experiment would reveal what works best.

4) smooth off any edges and fill gaps.

5) It may be wise to use thin plastic pipe inside the big drilled holes as the rough drilled surface will decrease flow by creating turbulence.

6) Maybe design in a way to allow the height of the inlet jet to be changed. If its mounted in a smooth pipe with O rings, you may be able to perform live height testing! A final model without the mod may perform marginally better, but this could provide good insight.
Make a groove round your pipe to secure the O-ring - thicker pipe may be needed to sink much of the o-ring. Finding parts that fit together well may be a bind.


Have you tried any water wetters? To make sure my Water cooler system was well wetted, I used a very mild solution of water and washing up liquid for testing my system before installing.
This has at least 2 benefits: its a good wetting agent and I havent had any Algae yet in 6 months with clear pipe!
Having drained this off I then put in the recommended amount of Asetek Anti Algae Fluid and distilled water.
I dont know whether wetting additives affect water adhesion to the cpu die, I imagine the effects would be positive though.

Here are my water cooler specs for the curious:
XP-M 2500+ @ 2.7GHz, 2.0V

Custom water setup cooling only CPU:
.Antarctica water block
.1/2" ID pipe through out except at cpu block (1/2" OD required, pipes simply inserted into each other and tightened together with Jubilee clips)
.Eheim 1048 pump
.Ford Sierra Heatercore (Radiator)
.Single 120mm Fan (unknown CFM) mounted with the Radiator in a Stella beer cardboard box, strapped up with duct tape
.No reservoir.

Temps using external sensor:
22C Ambient
35C Idle
42C Load

This was so easy to build out of nearly nothing and performs really well.

Gizmo, can I ask what you are using to seal the the Plexiglas to the CPU please. Before I start my own project, this is one of the things I need to sort out.

I know too that direct die should be able to beat this and be cheaper & easier to make than a water block. I really hope the above ideas are useful as I'd like to see it succeed.


Cheers and good luck


Chernobyl

ps
I accidentally let water flow down my motherboard once. I put it in the bathroom with a Dehumidifer for the night and it was right as rain the next morning.

[Gizmo]

Originally Posted by Chernobyl

Hi Gizmo
Excellent project!

A friend and I have been toying with making a direct die cooler too. I found this very useful thread as the first link on Google you will be happy to know
search phrase: "direct die" water cooling

Thanks for the props, Chernobyl.

And welcome to AOA! Hope you stick around for a while.

Originally Posted by Chernobyl

An observation that may improve performance is pressure at the CPU die.
The more pressure, the more heat can transfer in the same time.
Looking at your block, the cross sectional area of the pipe feeding the cpu is way less than that of the container feeding the outlet pipe. This allows the pressure to drop rapidly across the die and reduces the overall max pressure at the die.

Umm......not quite, I don't think. I don't have a background in fluid dynamics, but I do know electronics, and the two disciplines share some similar concepts. One of which is that to maximize flow, you have to maximize the differential. In this case, I designed the block to operate in precisely this fashion, because I wanted to get the outflowing water across the cpu die and into the reservoir (the block) as quickly as possible. This meant that I had to maximize the pressure differential at the cpu die.

It may be that this was not the optimal approach, but it seemed to make sense at the time.

Originally Posted by Chernobyl

6) Maybe design in a way to allow the height of the inlet jet to be changed. If its mounted in a smooth pipe with O rings, you may be able to perform live height testing! A final model without the mod may perform marginally better, but this could provide good insight.
Make a groove round your pipe to secure the O-ring - thicker pipe may be needed to sink much of the o-ring. Finding parts that fit together well may be a bind.

Go back in the thread and take a look here (item 1), here (third paragraph), and here.

I realize, because of the way the messages are posted, it's not so easy to tie the pieces together. Hopefully, having them organized in this fashion makes it a little easier. Or am I missing your point?

Your other suggestions are all good, but require machining a solid block. If you intend to go this route, I would be very interested in seeing how it turns out for you. Perhaps you could start a thread here and keep us updated with the results? (Hopeful look)

Originally Posted by Chernobyl

Have you tried any water wetters? To make sure my Water cooler system was well wetted, I used a very mild solution of water and washing up liquid for testing my system before installing.
This has at least 2 benefits: its a good wetting agent and I havent had any Algae yet in 6 months with clear pipe!
Having drained this off I then put in the recommended amount of Asetek Anti Algae Fluid and distilled water.
I dont know whether wetting additives affect water adhesion to the cpu die, I imagine the effects would be positive though.

The only 'water wetters' I have messed with are all ethanol based, and didn't work well (they tended to cake onto the cpu; seems our cpus have a heat density that is quite a bit higher than a car engine). I've since switched to using a solution of about 2% alcohol as an algicide, and that seems to work pretty well.

Originally Posted by Chernobyl

Gizmo, can I ask what you are using to seal the the Plexiglas to the CPU please. Before I start my own project, this is one of the things I need to sort out.

I'm using a custom formed gasket now, instead of the closed-cell neoprene foam that I describe here. I need to post some pics and descriptions of the new gasket too.

The main problem with closed-cell neoprene foam is that it deforms too much under pressure and doesn't completely recover. In fact, over a period of time what you find is that you have to periodically tighten your block in order to maintain the seal as the foam collapses.

[XeroHouR]

nice temps chrynobal, those are identical to what my storm block did on my barton at those temps and similar Voltages, I've got a prescott now, so I can't give further temps, but that shows good design on your part

[Chernobyl]

Hi Gizmo
cheers for the welcome
Sorry I missed a few points, its quite a big read. I'm pleasantly surprised to see you have built an adjustable height on the intake, excellent.

Regarding the higher pressure at the die, I'm not so sure that the electrical circuit analogy applies the same way. 2 important things worth considering are the flow rate and pressure at the die 'to maximise heat transfer'.
I am with you on maximising flow rate to get the cooler water in quicker.
What I am wondering (and think is true): Is there a decent tradeoff in terms of loss in water velocity (through restricted flow to give higher die pressure) and increased heat transfer from the die?

An increase in pressure while keeping the flowrate may be achieved with the use of a pump that has a higher head. Perhaps using 2 pumps in series? This needs investigation as the smaller water holes from the die area will certainly have a minimum size before water flow will be severely compromised. I dont plan on getting anywhere near that but do like to consider the extremes.

The water flow pattern to/from the die would need some work too to prevent water 'stagnation' over the die. (quickly shifting the heated water)

As an electronic engineer too I have no doubt you have thought longer and harder than can be written easily and I respect your point of view.
I dont have a choice but to try though

I am on an extemely tight budget so this project may not progress as fast as I would like and lack a few well needed tools (ie flow meter!, precision machining tools etc).
Hopefully this block can be made with an electric drill. My first one will be a proof of concept and will surely look rough!
I am yet to research the best plastic block to use, plexiglas is on there, I am sure there will be a few choices.

Yes, I should start a thread I'm hesitant to do so just yet until I have more of a cobbled together plan and some parts for the block.
The radiator system is underway with a 20" house fan (its MASSIVE) slowed down to make it silent (yet to be done).
Onto this will be strapped a kiddies plastic toybox (already prepared as a airbox) with 2 or more car heater cores as radiators sealed into it. The radiators will run in parallel.

Some postive and negative thoughts on having dual parallel radiators
1) the water will flow at 1/2 the speed meaning it spends more time in each cooling chamber.
2) The pressure drop (at the radiators) associated with the 2 radiators in parallel wont cause the water to boil, but maybe it will release heat a tiny bit better? (drawing a comparison with a Fridge compressor system) I doubt this
3) Double the cooling area (compared to 1 radiator)
4) The water flow restriction caused by the radiator(s) is approximately halved.
5) the pressure drop may impact thermal transfer from water to radiators (opposite to high pressure at the block). With so many other positive aspects to dual parallel radiators, this may not be part of my considerations.

Theres a lot of thoughts buzzing round my head so I should go away and think some more rather than emptying it here in your thread.

Thanks for doing what you have, its spurred me on, and many thanks for posting your results.

I'll be back!

Chernobyl

[Chernobyl]

Hi XeroHour,
I moved the chat to here so Gizmo's thread doesnt go off course
http://www.aoaforums.com/forum/showt...765#post280765

[Gizmo]

Originally Posted by Chernobyl

Regarding the higher pressure at the die, I'm not so sure that the electrical circuit analogy applies the same way. 2 important things worth considering are the flow rate and pressure at the die 'to maximise heat transfer'.
I am with you on maximising flow rate to get the cooler water in quicker.
What I am wondering (and think is true): Is there a decent tradeoff in terms of loss in water velocity (through restricted flow to give higher die pressure) and increased heat transfer from the die?

What little I've seen on this suggests that the tradeoff is highly dependent on the design of the block, although it does appear to be fairly similar within a particular class of block (such as a direct-die jet impingement design).

Originally Posted by Chernobyl

An increase in pressure while keeping the flowrate may be achieved with the use of a pump that has a higher head. Perhaps using 2 pumps in series? This needs investigation as the smaller water holes from the die area will certainly have a minimum size before water flow will be severely compromised. I dont plan on getting anywhere near that but do like to consider the extremes.

My Danner has about 14' of head now, IIRC. A jet impingement design like mine is actually pretty restrictive, and so requires a pump with good head pressure to achieve decent flow rates. I've been meaning to do some tests with varying flow rates for a while now, but just haven't gotten around to it.

Originally Posted by Chernobyl

As an electronic engineer too I have no doubt you have thought longer and harder than can be written easily and I respect your point of view.
I dont have a choice but to try though

I have thought about it a great deal, and have read the work of others (such as our very own SurlyJoe). That doesn't mean that I know all there is to know, or that I have considered every possibility. Since I have not been able to provide a compelling reason why you should not proceed with your experiments, by all means, give it a shot. Even if you fail, you will have learned something. Sometimes our failures are more important than our successes. If you read back through this thread, you will realize that my first few attempts at a design were failures.

Originally Posted by Chernobyl

I am yet to research the best plastic block to use, plexiglas is on there, I am sure there will be a few choices.

If you have a plastics wholesaler in your town, go there and explain what you have in mind to them. You will probably find that they have some scrap they will sell you quite cheaply. I've got some scrap nylon cubes about 2" square here that I picked up for about $2 each a while back because I was considering building something like what you describe. I ended up not building because I realized the way I was trying to construct it was all wrong, and I lacked the machining tools necessary to achieve what I wanted. Your approach is simple enough that you might be able to do it.

Originally Posted by Chernobyl

Some postive and negative thoughts on having dual parallel radiators
1) the water will flow at 1/2 the speed meaning it spends more time in each cooling chamber.
2) The pressure drop (at the radiators) associated with the 2 radiators in parallel wont cause the water to boil, but maybe it will release heat a tiny bit better? (drawing a comparison with a Fridge compressor system) I doubt this
3) Double the cooling area (compared to 1 radiator)
4) The water flow restriction caused by the radiator(s) is approximately halved.
5) the pressure drop may impact thermal transfer from water to radiators (opposite to high pressure at the block). With so many other positive aspects to dual parallel radiators, this may not be part of my considerations.

I am running a dual radiator setup in my system now, although they flow in series and cool in parallel. I had to do so, because a single radiator was insufficient to keep my water at the temp I wanted (the water was getting about 10 °C above room temp, and I wanted better). The current arrangement allows me to keep my water temp at very close to room temp. For optimal performance, you do NOT want to stack the radiators in a front and back arrangement (i.e. air flows over one radiator, then the other).

I encourage to start your own thread, even if you think it may be a while before actually doing something. You can always think out loud and post your thoughts here, even if you aren't able to build anything.

[Chernobyl]

Cheers Gizmo
I will start a thread once the show is on the road.
Thanks for your assistance

Chernobyl.

[Chernobyl]

I spotted this article (in Dutch).
http://www.ocshoot.no/antarctica_xtreeme1.htm

it shows my CPU block being converted for direct die cooling (Asetek Antarctica).
This is a good idea as the central water feed is already present, the 2 outlets will help improve waterflow. This would work with many blocks.

To keep heat down in the loop, 2 Eheim 1048 or higher pumps could be used instead of 1 powerful pump. This would be quieter, cooler and capable of creating a lot of pressure. Join the 2 pump outputs before going into the block and have 1 or 2 circuits running to cool the water. Perhaps have a radiator before the pumps and after if really worried

[Gizmo]

I've thought about that myself. Thing is, I think the jet design of both the Asetek and the Danger Den RBX blocks basically sux. The orifice is too wide and the transition is too abrupt. 'Course, what do I know? I'm just a seat of the pants engineer here, anyway. LOL

[Chernobyl]

I agree, pressure needs to be alleviated everywhere except at the CPU die.

I was thinking of making a glue on holder for a Jet impingement system, it should be pretty easy to bond to acrylic.

Just a thought, have you tried a cross shaped impingement outlet, the width and breadth of the CPU core? That would be a nice way to get the water shifted off the core as fast as possible (4 routes off) and maintain strong pressure over the whole die.

[Gizmo]

Originally Posted by Chernobyl

I agree, pressure needs to be alleviated everywhere except at the CPU die.

I was thinking of making a glue on holder for a Jet impingement system, it should be pretty easy to bond to acrylic.

Just a thought, have you tried a cross shaped impingement outlet, the width and breadth of the CPU core? That would be a nice way to get the water shifted off the core as fast as possible (4 routes off) and maintain strong pressure over the whole die.

I have not tried that. It would be a PITA for me to build, given my limited metal-working equipment.

Thinking about it though, I wonder if the turbulence at the intersection of the two streams wouldn't create too much of a flow disturbance?

I've REALLY gotta get my hands on some simulation software.

Oh, hey! I've got a Linux box now, maybe some of the Unix stuff for doing fluid dynamics symulations will run on it? Hmmmm..............(trots off to do some investigating.......)

[Chernobyl]

Originally Posted by gizmo

I have not tried that. It would be a PITA for me to build, given my limited metal-working equipment.

Thinking about it though, I wonder if the turbulence at the intersection of the two streams wouldn't create too much of a flow disturbance?

I've REALLY gotta get my hands on some simulation software.

Oh, hey! I've got a Linux box now, maybe some of the Unix stuff for doing fluid dynamics symulations will run on it? Hmmmm..............(trots off to do some investigating.......)

lol, GO GIZMO!