GPU/NB Water block (development thread)

[Socko]

Originally posted by jaydee116
My first Asus board was the original A7V with the KT133 chipset and NOT the KT133A chipset, is that the one your refering to? If so I agree. Mine was a damn nice board. Still regret selling it.... I got the Abit KT7A right after I sold it thought and that was a good board to. I dropped a hard drive onthe South bridge and killed the hard drive controller. I loved overclocking Durons on those boards.

Kinda ironic that a hard drive killed a hard drive controller. I've been sticking with abit from the beginning, and they kick the crap out of almost every other board when it comes to overclocking.

[Socko]

How about something like this:

[tex707]

Originally posted by Blackeagle
What I thinking at this point Tex707 is to cut the Lexan 1/2" top (one side) at an angle. Combined with a thin based block with pins raised up out of it I think this can work.

Pins would be approx. .2" (5mm) and the top would be .5" Lexan,
with a thinner 1/8-3/16" base.

I'll rout a square area for the pins into the Lexan top. Although I may also use a copper 1/2"=>3/8" bell reducer as the inlet to get a bit more water velocity to impact the base with, & a bit of 1/2" copper tube for the open low restriction outlet.

A question I still have.

Is the .2" pin hieght a good length? I'm open to advice on this as I picked this .2 height as it would be handy to work with along with a 1/2" top of Lexan.

Is this at least close to your idea..??..I'll wotk on the copper base and a pocket in the top later...

[tex707]

Originally posted by Socko
How about something like this:

Well, this looks a lot like jaydee's project...except for a few details. I think that the pattern of the fins could be at 45 degrees angle to the sides.

Two bolts for mounting a top on the copper base could be sufficient due to a high rigidity of the top, but I think that four bolts would make a more secure fit...don't forget that the O-ring is to be compressed as evenly as possible.

About the Allen bolts....I think that counterbored holes should be used wherever possible.

[tex707]

...and here's what I mean...

[jaydee]

Originally posted by tex707
...and here's what I mean...

Fins/pins should be rotated 90 degreed so the longer sides are facing the barbs IMO. That will help flow a little.

[pHaestus]

tex those angled barbs look good to me.

[tex707]

Originally posted by jaydee116
Fins/pins should be rotated 90 degreed so the longer sides are facing the barbs IMO. That will help flow a little.

That will definitely help flow, but how about turbulence?

[tex707]

Originally posted by pHaestus
tex those angled barbs look good to me.

Do you mean the barbs only, or the Plexi top with barbs mounted at an angle?

[jaydee]

Originally posted by tex707
That will definitely help flow, but how about turbulence?

Should be good. The higher flow rate should make up for less turbulance not only in water volocity but less heat added to the water from the pump not having to work so hard.

[pHaestus]

The whole package tex. It looks like something I would want in my loop. Any concerns that it would make it less sturdy though?

[Blackeagle]

Tex707,

That's about it, if you could angle the barbs a bit lower as JD suggests it'll be spot on.

You're a magick man at CAD! What a sweet pic!

We really do need a wider selection of smilies, this is the second time in about a week I've seen something deserving of more, much more.

I'm gonna print off some pics to keep of this.

Thanks VERY much Tex707!

Jaydee,

What do you think of this idea? Advice a further suggestions would be great.

[jaydee]

Originally posted by Blackeagle


Jaydee,

What do you think of this idea? Advice a further suggestions would be great.

The problems I see is you will be restricted to having the hose barbs on the same side of the block. Which will pretty much destroy my theory on flow through my pin desing. This can easily be worked around with a few changes though. Namely a wall down the middle of the block. No big deal.

Next problem, and more importantly, I see is the connectors maybe to close together. I always figure about 7/16" for gap for 1/2" barbs. 1/8" for each hose so thats 1/4" plus 1/16" for meatal hose clamps x2 = 3/8" plus another 1/16" to be able to move the clamps around.....

If those things can be worked around it should be good. Might even be doable with my mill.

[tex707]

Originally posted by pHaestus
The whole package tex. It looks like something I would want in my loop. Any concerns that it would make it less sturdy though?

I will try to make some structural analysis, but don't expect to find a stress get "over the boundary" anyplace over the Plexi top.

[tex707]

Originally posted by Blackeagle
Tex707,

That's about it, if you could angle the barbs a bit lower as JD suggests it'll be spot on.

You're a magick man at CAD! What a sweet pic!

We really do need a wider selection of smilies, this is the second time in about a week I've seen something deserving of more, much more.

I'm gonna print off some pics to keep of this.

Thanks VERY much Tex707!

Jaydee,

What do you think of this idea? Advice a further suggestions would be great.

First of all, thank you...

About the angle of the barbs...this is about maximal possible with given acrylic top thickness of .5"....if I go to 5/8 or more, the angle could, probably, drop to 45 degrees or even less.

Here is a complete 3D model and one cross-section...please note that the copper base is pretty thin with pins protruding above the upper base plane into the acrylic top pocket.

[tex707]

Originally posted by jaydee116
The problems I see is you will be restricted to having the hose barbs on the same side of the block. Which will pretty much destroy my theory on flow through my pin desing. This can easily be worked around with a few changes though. Namely a wall down the middle of the block. No big deal.

Next problem, and more importantly, I see is the connectors maybe to close together. I always figure about 7/16" for gap for 1/2" barbs. 1/8" for each hose so thats 1/4" plus 1/16" for meatal hose clamps x2 = 3/8" plus another 1/16" to be able to move the clamps around.....

If those things can be worked around it should be good. Might even be doable with my mill.

You're right about this...I was thinking about separating the barbs as much as possible...maybe we should consider redesigning the acrylic top from the scratch...

[tex707]

Originally posted by jaydee116
Should be good. The higher flow rate should make up for less turbulance not only in water volocity but less heat added to the water from the pump not having to work so hard.

Is there any data around on the amount of heat that some pumps add to the water...I have three pumps (one in the outer and two in the inner circle of the water tower) and the water is still very cold (cooler that the surrounding air). My EHEIM 1048 in the inner circle feels just a bit warm...

[jaydee]

Originally posted by tex707
Is there any data on the amount of heat that some pumps add to the water...I have three pumps (one in the outer and two in the inner circle of the water tower) and the water is still very cold (cooler that the surrounding air). My EHEIM 1048 in the inner circle feels just a bit warm...

Non that I am aware of. I have seen up 3C added in my testing with various pumps. Submerged pumps show the gratest differences.

[tex707]

Originally posted by jaydee116
Non that I am aware of. I have seen up 3C added in my testing with various pumps. Submerged pumps show the gratest differences.

3C sounds a lot to me, JD....pretty disappointing...

BTW, it seems that I've missed one of the most important facts...do you intend to use the NB and/or GPU W/B in line or parallel with the CPU W/B?

[tex707]

Originally posted by pHaestus
The whole package tex. It looks like something I would want in my loop. Any concerns that it would make it less sturdy though?

OK, the structural analysis has shown what we have already expected...the stress didn't even reach 2e+006 N_m2 with 50N load at each spring...

[jaydee]

Originally posted by tex707
3C sounds a lot to me, JD....pretty disappointing...

BTW, it seems that I've missed one of the most important facts...do you intend to use the NB and/or GPU W/B in line or parallel with the CPU W/B?

I had originally planned to have 2 NB and one GPU block in inline on their own loop seperate from the CPU loop. But this has all changed. I scraped the 3 comps on one water cooling loop idea. Instead I am resetting up my test bench.

So now it would be CPU to NB to GPU inline. Also took out the GF4 MX and the mounting holes are totally different from anything I have seen so far.

[Blackeagle]

Originally posted by jaydee116
I had originally planned to have 2 NB and one GPU block in inline on their own loop seperate from the CPU loop. But this has all changed. I scraped the 3 comps on one water cooling loop idea. Instead I am resetting up my test bench.

So now it would be CPU to NB to GPU inline. Also took out the GF4 MX and the mounting holes are totally different from anything I have seen so far.

Very glad to hear you are setting up your test bench Jaydee. Good for you.

I have few more question for you on my ideas. How large should the pin fin area be for full a proper cooling of a ATI 9700-9800 die?

I'm going to recut/shape a Akasa one piece copper 1U pin fin sink to try to make a block along the ideas discussed here. While the pins are smaller (1/16"), round and more densly placed, I hope to be able to make a block with a top along the lines Tex707 last showed. This air sinks base is 3/16" thick with pins being 3/8" tall. So I need to both remove many of them and shorten them by about 1/2 the height they are now.

As this base is thicker than has been shown to be ideal for cooling I could cut some grooves between the pins with a reinforced dremel cutting wheel, it's just the right size to fit through between the rows of pins. But would the roughness and narrow width of these grooves be self defeating (stagnant flow in them)?

Last I don't have anything beyond a dremel at the moment to try to cut the O-ring groove. Thinking I'll have to use some Marine goop instead.

Tex707,

You're right of course about the angle being at it's limits. Your cut away of the design made that plain. I wish I had thicker Lexan to work with, but the thickest I have is .5" at this time, and not much chance to get any that's thicker for a bit.

While I found the stress work interesting, you used acrylic, but Lexan (polycarbonate) is FAR stronger (approx 30X I think) for a given thickness of material than acrylic. Would not Lexan pretty well remove stress risks for the design?


Thanks to all for your continued help and advice.

BE

[pHaestus]

Blackeagle:

Instead of the Akasa, what about the possibility of using a drill press to make a #rotor style base? You could get the same pin effect and you could get a much thinner baseplate by just drilling deeper.

[Blackeagle]

pH,

You're right, I could make it that way. And I may try it that way as well. I would want to get a X-Y or cross vise first though I think. Would be much harder to do it really well free hand, and right now I have no drill press vise at all.

Along with some hand tools heres a list of power tools I have at the moment.

Radial arm saw 10"

drill press

small band saw, I have both metal and wood blades for it.

hand held belt sander

2 hand drills

Dremel with a router attachment for it along with a pretty extensive selection of bits for it.

So for now I'm going to have a go with the Akassa sink.

Can you tell me the size the pin fin area needs to be to cover the the GPU of a 9700-9800?

Thanks,

BE

[tex707]

Originally posted by Blackeagle


<snip>

Tex707,

You're right of course about the angle being at it's limits. Your cut away of the design made that plain. I wish I had thicker Lexan to work with, but the thickest I have is .5" at this time, and not much chance to get any that's thicker for a bit.

While I found the stress work interesting, you used acrylic, but Lexan (polycarbonate) is FAR stronger (approx 30X I think) for a given thickness of material than acrylic. Would not Lexan pretty well remove stress risks for the design?


Thanks to all for your continued help and advice.

BE

I'm working on a double top right now...will see the results soon......however, I'm afraid that the assembly is going to be too thick...

About the stress...Plexi is good enough and I am doing the analysis just to prove it and not because I doubt it. However, it would be interesting to check Lexan out but. Since I have no data about it at, I would have to input the data myself. I have never had a chance to see the material (except there are some Lexan parts in Boeing planes that I don't know of ). If you have the basic data, please let us know.