Jets + pyramids?

[JAxen]

Hi there..
I am about to do my first CPU block soon.. I just have som questions first..

I'm lending Djemshut's pix here..


My ide was as simply as

+


Will this work out alright or end in disaster ?


Regards Jonas

[bobo5195]

As far as I know it has not been heavily investigated. Alot of jet impingement methods have looked at using dimples on the surface to increase heat transfer rate and for adding a longer path of channel for the fluid.

There are some studies which have shown that a jet hitting an inclined surface has a higher overall heat transfer coefficant but a lower maximum value. This might be the case here where the pins have the jet hitting at an angle.

Either way its worth a punt. Comparing against a flat surface would be useful.

[JAxen]

Well, I can just do two different second layers where one is with jets and test wich one performs better..
Thnx though

[Seven]

Part of me thinks this would be easier if you simply bought a jet impingent block such as a Swiftech STORM or a G4 and simply machined a new baseplate with pyramids instead of dimples.

[JAxen]

Pherhaps..
But I've changed my mind and thinking 'bout using a cross shaped "jet" instead..
Gotto brainstorm a little on this

[davidzo]

i made a fairly lowflow waterblock with 1.5*1.5mm and 1.0*1.0 Pyramids milled with a 60° and a 36° engraving tool and used some ferules to direct the water onto the pinfield, very much like in your pciture but everything scaled a little down.
It performed really bad, flow was not very good but okay in comparision to other better performing tests i did. With the same pyramid baseplate it showed better results just to drill a plate with holes instead of the jets.
Try a nozzle pattern with multiple slits! I tried it and it worked out best for my pyramid coolers, even better than holes, but be aware that the alignment between pyramids and channels and the slits is perfect, this can affect overall performance extremely. In this case i tried very small slits aligned over the channels, not thy tops of the pyramids, it worked out good. when you make the slits bigger, you can probably set them over the pyramids instead of over the channels.
In this case the water flows also horizintally inside the slits from the sides of the slit to the middle. and the flow in the baseplate ist mostly exact in a right angle to the flow inside of the jet channels. Because of that, the pins are flooded even from every side, not just through the horizintal or the vertical channels like it is with most pincoolers. The crossflow also increases rubulences, like in the a little bit too chunky Innovatek X-flow, XXflow and newer G-flow series.

[JAxen]

Ok, not sure I understand everything
But that is some nice looking blocks you've doned..


This are the AutoCad drawings I've made so far

Mounting Plate


The Top


The Pyramids


There are 110 pyramids on 4.5x4.5mm milled probably with a 60° engraving bit as well..

[bigben2k]

Nice work! I remember exploring the same idea, a long time ago! This should work very well!

[JAxen]

Thnx
Didnt think I'd get this kinda respons =D
Then I just might get to this when I'm aloved to use one of my schools CNCs

[bobo5195]

jet inpingment needs a fast max centerline velocity to work well. The best way to get this is with circular holes. Always use circular holes with jet impingment and always make them as small as you can. Then optimise the number of holes to get the right amount of cooling area.

X shaped holes also are not a good idea because they are complicated to machine, add extra flow path restriction and don't give good performance.

Cross flow increases performance of all jet impingment designs by a measurable amount and is very commonly used in industry.

[Djemshut]

Originally Posted by JAxen

There are 110 pyramids on 4.5x4.5mm milled probably with a 60° engraving bit as well..

I think this is big pyramid, optimal dimension - 3x3x5mm. Area-30x30 mm maximum, bigger-not better. Base plate- 2,5-3 mm for processors with IHS and 4 mm without IHS.
Sorry for my bad english

[davidzo]

diemshut ist just right, make the pyramids smaller and the whole structur too. it should onle be slightly bigger than the die actually is, so you can cut away all the outer pyramids and change them for bigger outlet channels which will result in a better flowrate. as far as i tested, 1,5*1,5mm Pyramids are very nice to mill and give a you a boost in performance over bigger ones. always look at the cross section of the channels in the whole structure around the inlet. it should be less than the corssection of your barb and tube for best velocity and performance. if it is bigger it will result in extremely laminar flow through your pyramids and thats bad

[JAxen]

Oh, OK..
But what kinda tool should I use to mill pyramids that is 3x3x5mm?
Or even 1.5x1.5mm.. What angle do you get when milling 3x3x5mm?
I've got a P4 with IHS, wich I'm planning to remove if I got the guts

Best regards

[JAxen]

Originally Posted by davidzo

always look at the cross section of the channels in the whole structure around the inlet. it should be less than the corssection of your barb and tube for best velocity and performance. if it is bigger it will result in extremely laminar flow through your pyramids and thats bad

Could you please explain a little more?
My english aint to good

[bobo5195]

Davidzo,

Do you mean mass flow conservation? ie Area x Velocity

Also i'm fairly sure that the flow in them there pyrimads is lamina at PC watercooling pumping levels.

The dimples idea has been tired actual I only found one paper and it was actual part authored by one of my docs. He was using dimples to increase the heat transfer inside a intercooler for turbocharging. He got similar results to you that the dimples were very effective provided that they were smaller than the wall jet thickness.

[Djemshut]

My way of jet-pyramyd construction:

[JAxen]

I'm not sure I understand your drawing there Djemshut
Anyway, I've been doing some minor changes in my blueprints
What do yall think about this?





The new pyramids are 3x3mm, they will probably be about 5mm high..