Water block+Pump=?

[Volenti]

Pump_block

Original concept by fixittt, shamelessly stolen..er..borrowed by me The idea is to have the ouput of the pump as close to the water block as possible so you can take advantage of the extremely turbulent water flow the pump impeller creates.

What could be better than having the impeller "inside" the block it's self? where the ouput from the impeller can be directed over the core area while the water is at it's most turbulent.

I have bought a small 700L/H(186gph) power head that should fit the bill nicely, it leaves just enough room on a P4 water block base to allow for 1/2'' hose barbs.

[Volenti]

pic showing pumps impeller

[Volenti]

Now, if I we're 1337 I'd have a flash exploded 3d model to show you how I'm going to make this thing, but I'm not so my sketches will have to do

this is the original concept sketch, it's a bit confusing since it shows multiple layers (4 in all) it will make more sence hopefully once I chow the individual layers in detail.

[Volenti]

Layer 1, the base, where all the important stuff happens, the hashed area is carved out to a depth of around 3mm, this is the "suction" side of the pump, and allows the block to actually pump water (the vital detail missing in fixittt's concepts). it also has the small retaining hole for one end of the impeller axle as well as shallow grooves directly over the core area for a bit more surface area.

[Volenti]

Layer2 I'm not sure weather this layer will be needed yet or not, it's mainly there (in theory) to help/confirm the sucking action of the impeller, I'll dry assemble the block and run it either way to see if it helps or not.

[Volenti]

Layer 3, This layer acts in the same manner as the detachable part around the impeller on power head pumps, directing the outward force of the water in a single direction (over the core)

[Volenti]

Top layer, handles the hose barbs and the physical mounting and sealing of the pump to the block.

[bigben2k]

I've never seen this before, I like it! I'll have to look back at what Fixittt posted...

[Sanjuro]

Good going! Now finish the thing!

[Skulemate]

Are you going to try to reproduce the impellor housing in the copper block, or have I missed something? It could be tough to make the pump work as well if you are making your own, since they are specifically made to work with their impellor... I'd be interested in seeing before and after H-Q curves for this thing. Also, I think that the discharge needs to be more off centre from the impellor axis (just look at the housing that came with the pump... I think your goal should be to reproduce the inside of the original housing exactly, only in copper). Good luck with your build... seems like an interesting concept.

[airspirit]

Does anybody else think that putting a heat source on your block is a bad thing? What is the W rating of that pump? Also, is the pump pulling or pushing through your block?

[Sanjuro]

Like said, hes basically cooling the CPU with a copper impeller housing, so pushing or pulling doesn't really matter. It does seem like a good idea if he can get the origional impeller housing into CAD and CNC mill out his block. Personally I can't wait to see how this thing prforms

[Volenti]

Originally posted by Skulemate
Are you going to try to reproduce the impellor housing in the copper block, or have I missed something? It could be tough to make the pump work as well if you are making your own, since they are specifically made to work with their impellor... I'd be interested in seeing before and after H-Q curves for this thing. Also, I think that the discharge needs to be more off centre from the impellor axis (just look at the housing that came with the pump... I think your goal should be to reproduce the inside of the original housing exactly, only in copper). Good luck with your build... seems like an interesting concept.

That's that I'm doing basicly, copying the impeller housing as best I can in copper, the factory housing has the outlet in the centre so I'm copying that too.

I don't ecpect it to pump as well as it originally did, I'd be happy if it pumped 100GPH, it's the turbulence I'm after, afterall, I's a simple procedure to hook up an external pump and see if the impeller turbulence has anny effect or not.

[Volenti]

Originally posted by airspirit
Does anybody else think that putting a heat source on your block is a bad thing? What is the W rating of that pump? Also, is the pump pulling or pushing through your block?

The pump is only 12w, It'll be putting less heat into the water/block than my larger pumps currently are, some of the big delta fans on heatsinks must be approaching 10w or so, no one worries about the heat coming off them.

it's both pushing and pulling through the block, the area over the core is on the "push" side.

[myv65]

Originally posted by airspirit
Does anybody else think that putting a heat source on your block is a bad thing?

Nope, just you.

Doesn't matter where the pump is from a power standpoint. Granted, its power may vary a little because he's not matching the original pump volute, but power is power regardless of its "insertion" point. OK, technically you could remove the pump power with a radiator positioned between the pump and block, but at any flow over ~30 gph the delta-T from the pump is far less than 1/2°C.

[Volenti]

WOOT! update

I've used a smaller pump (similar to what those little table top fountains use) and it's almost ready for leak testing.

I'll have to wait a day or so for the neutral cure silicon sealing the base to cure.

[bigben2k]

Pump specs?

[Puzzdre]

the idea is great, but I'm interested in mounting this thing on the wblock, especially because vibrations. I think that would be the hardest point, eliminating vibrations of the pump (plus added weight of copper) directly over the cpu...Somehow don't have the feeling that cpu core need more lapping...

Ofcourse, why not try it!?

Good luck!

[Volenti]

Originally posted by bigben2k
Pump specs?

oh sorry, it's rated at 250L/H, and about 35 or 40cm head,(and a HUGE 6w power draw!) I've streamlined the flow paths though so it does almost 300L/H (like porting an engine intake manifold)

Before I chopped the pump up I ran it through my new U1 heatsink block so I could get some sort of baseline performance figures to compare it to.

The copper base is only 1.7mm thick, I've deliberatly limited the copper to water surface area (about 25-30mm^2) so it really is a test of how the turbulence from the impeller effects the boundary layer, the impeller is around 1.5mm above the surface of the copper.

I have another identical copper plate that I intend to set up a "conventional" water block using the same surface area(same shape even) as the pump block, like a control in an expriment.

Though I may be able to get away with simply leaving the pump-block un-powered and pumping water through the block with another pump throttled back to the same water flow, we'll have to wait and see...

[Jester]

ah so this is the block , not bad indeed .. is their any accual marks or dips in copper base to increase surface area or is it just Flat smooth copper under the impellor ?

dam dude just realised i been browsing these forums for about 6 months no and this is the first post blah

[Volenti]

Originally posted by Jester
ah so this is the block , not bad indeed .. is their any accual marks or dips in copper base to increase surface area or is it just Flat smooth copper under the impellor ?

dam dude just realised i been browsing these forums for about 6 months no and this is the first post blah

no special treatment other than a sand with coarse sand paper, oh and the profile of the whole assembly is small/thin enough to fit a standard 50x80mm socket A block profile (:

[Volenti]

Results!

with a water temp of 24.4 degrees my p4 1.8@2.64, 1.8v idles at 36 degrees, 1 degree above my other blocks,not too bad.

load's a different story, using a cpu burn proggie the temps shoot up to 42 degrees and just as suddenly stop there, quite funny really, I'm like "shit !shit!" as I start the burn proggie and see the temps race up but then see them stablize so fast, like it hit a wall or something.(my other blocks only rise ~3 degrees above idle temp under load)

Now for the part that's really interesting


The pump-block pushes around 200L/H, I hooked up a larger pump in my res in series with the pump block (a 700L/H 1m head pump) brining the flow rate up to ~550L/H

I fired up the cpu burn proggie to see if the added flow made any difference, 0, nada, zilch, still 42 degrees, hmm I thought. I then turned off the pump block (with every thing still going)overall flow was barely affected, however the temps rose 2 degrees without the pump-block active, I checked this several times.

Conclusion;

The pump-block, running by it's self, achieves load temps 2 degrees lower than when it's off and passing over twice the water flow of when it's on.

Like that bigben2k?

Keep in mind that this block was built solely to test a theory, and it's surface area is deliberatly small to make the results more pronounced.

[bigben2k]

Ok, so basically your experiment showed that you got 2 deg C improvement, by actively turbulating the water over the coldplate.

Interesting indeed!

You're getting very low water temps, but there's a 12 to 18 deg diff between the water and the CPU, and that's an indication that there's a problem with your baseplate. I'd guess that it's too thin, but like you said, it was just an experiment, to test a theory.

Would you consider doing this again and reducing the gap between the impeler and baseplate even further, down from 1.5 mm? Alternatively, you could add brushed to the impeler arms.

[Volenti]

Originally posted by bigben2k
Ok, so basically your experiment showed that you got 2 deg C improvement, by actively turbulating the water over the coldplate.

Interesting indeed!

You're getting very low water temps, but there's a 12 to 18 deg diff between the water and the CPU, and that's an indication that there's a problem with your baseplate. I'd guess that it's too thin, but like you said, it was just an experiment, to test a theory.

Would you consider doing this again and reducing the gap between the impeler and baseplate even further, down from 1.5 mm? Alternatively, you could add brushed to the impeler arms.

The water temps are simply a result of the ambient temps I have here, nothing special about that, my best blocks get about 10-13 degrees above water temps (idle-load) that's a fairly typical delta with the other water cooled P4's that I've seen.

it wouldn't help any even if the BP was thicker since the only copper that comes in contact with the water is directly above the heat spreader, more or less, that's the main reason why i used a very thin BP.

The 1.5mm gap was an estimate since the impeller can move slightly up or down (it's held in place by the magnetic field ) the pump makes a sound similar to what I'd expect if the impeller was touching the BP slightly (slightly scratchy/rattly sound) that isn't very surprising given the suction the impeller creates would tend to try and pull the impeller closer to the BP.

I'm not sure what I'll do with version 2 yet, may try to modify/make an impeller to do more mixing over a larger area and use a booster pump to assist flow rate. will have to sleep on it.

[bigben2k]

You could try sticking a little pin on the BP, to keep the impeller from rubbing it.

I guess you could try putting divots in the baseplate (like Swiftech). Just use a 1/4 in drill bit and make a nice circular pattern, or something.