Production Aquanaut Basic/Extreme - Ultra Low Profile CPU Block & Pump Mount Combo

[Allhopeforhumanity]

Looking at this picture specifically and thinking about the Steck and other sandwich style chassis, it would seem to me like having the inlet/outlet ports rotated 90 degrees would be more optimal. In this direction I'm having trouble imagining a tubing layout where the hoses don't interfere with each other, or have to cross "vertically" (out of the boards plane) in order to reach a radiator.

Maybe you've got something up your sleeve; anyway I'm excited to see this project continue.

 

[Nouvolo]

Looking at this picture specifically and thinking about the Steck and other sandwich style chassis, it would seem to me like having the inlet/outlet ports rotated 90 degrees would be more optimal. In this direction I'm having trouble imagining a tubing layout where the hoses don't interfere with each other, or have to cross "vertically" (out of the boards plane) in order to reach a radiator.

Maybe you've got something up your sleeve; anyway I'm excited to see this project continue.

The block can be rotated 90deg and should work. I just need to find some other screws as the thumbscrew head is a bit too big in that orientation.

 

[lehman]

How about the impeller? Diameter of the cavity? And how much does it stick out from the body of the pump? Well will be best if I can find the dimensions of the top.

The impeller is 39.91mm and is 3.81mm tall.

Pump body is 17.75mm in height and 61.50 mm x 61.50 mm.

Pump top cavity is 7.19mm deep. It grows from 43.04mm to 45.20 mm in diameter. There is a ring around the pump inlet that is 1.77mm tall.

Hopefully this helps.

Lehman

 

[Nouvolo]

The impeller is 39.91mm and is 3.81mm tall.

Pump body is 17.75mm in height and 61.50 mm x 61.50 mm.

Pump top cavity is 7.19mm deep. It grows from 43.04mm to 45.20 mm in diameter. There is a ring around the pump inlet that is 1.77mm tall.

Hopefully this helps.

Lehman

I will probably get a SPC myself, too risky prototyping without actually installing one on to the block, and then claim it is compatible?. Thanks anyway for the great effort, really appreciated, at least from the numbers it seems it is only going to be smaller than the DDC, which makes changes easier.

 

[lehman]

I will probably get a SPC myself, too risky prototyping without actually installing one on to the block, and then claim it is compatible?. Thanks anyway for the great effort, really appreciated, at least from the numbers it seems it is only going to be smaller than the DDC, which makes changes easier.

I definitely understand. I am building my block for this specific pump and I am running too tight of tolerance for production variances of the pumps.

The big plus for using the SPC is the smaller height added to your block and the one connection to a pump header on the MB.

One of my previous designs is a lot like your design. But I am needing sub 39mm total height since I went a little overboard on my case I designed.

Lehman

 

[Nouvolo]

I definitely understand. I am building my block for this specific pump and I am running too tight of tolerance for production variances of the pumps.

The big plus for using the SPC is the smaller height added to your block and the one connection to a pump header on the MB.

One of my previous designs is a lot like your design. But I am needing sub 39mm total height since I went a little overboard on my case I designed.

Lehman

Yea, there is a lot of limitation with height, mounting position, motherboard component height clearance, and the design will end up converge to one very similar.

One thing that I have no idea and requires real testing is that the flow direction is reversed comparing with a standard water block. i.e. reverse micro channel jet flow. May need to test if performance will be better without the jet plate. My hunch is that for SPC (lower head pressure), jet plate could be a bigger restriction.

 

[Nouvolo]

@all

I am looking for some review or paper that talk about water block (jet plate type) performance with reverse flow direction. I can't seem to find any on the web. EK has tested GPU block and they get a difference of about 1.5° C at worst, which is nothing really. [Link]. But many also suggest that difference will be greater for CPU block (jet plate type).

Would appreciate if anybody can post some links of such, share your experience.

There seems to be a consensus that reversing flow direction affects performance, but there isn't much quantitative result, as to by how many degrees?

 

[Shatrod]

I recall Linus did a video comparing cheap Chinese waterblocks and they were definitely worse but not by a whole lot. Some of them didn't even have any fins at all, just a flat copper plate and they were "fine". i took apart an old corsair h50 and found that it did not have a jet plate. Water just entered on one side, flowed across the fins, and then put the other side. (I 3d printed a top for it to use as a gpu block, still have it). Until fairly recently most gpu blocks were flow direction agnostic as well.

I think just the properties of copper and water are good enough that you can neglect this? Slightly higher flow rate should make up for it.

Edit: I found this for you. Not exactly a scientific paper but something nonetheless: https://linustechtips.com/main/topic/821495-cpu-water-block-flow-direction-test-does-it-matter/

 

[flar]

See also the Fr33Flow: https://smallformfactor.net/forum/threads/fr33flow-48mm-cpu-block-pump-res-maybe.9837/

 

[lehman]

@all

I am looking for some review or paper that talk about water block (jet plate type) performance with reverse flow direction. I can't seem to find any on the web. EK has tested GPU block and they get a difference of about 1.5° C at worst, which is nothing really. [Link]. But many also suggest that difference will be greater for CPU block (jet plate type).

Would appreciate if anybody can post some links of such, share your experience.

There seems to be a consensus that reversing flow direction affects performance, but there isn't much quantitative result, as to by how many degrees?

My design is reverse flow. I searched high and low for any testing. The best I found was people running their EK cpu blocks with the flow reversed due to wanting a cleaner layout for tubing. Anecdotal evidence at best. But they all seem to say 1 degree c at most.

I will be running an Optimus waterblock coldplate. I asked their engineer and he said it would minimally effect the temperature.

Lehman

 

[Nouvolo]

My design is reverse flow. I searched high and low for any testing. The best I found was people running their EK cpu blocks with the flow reversed due to wanting a cleaner layout for tubing. Anecdotal evidence at best. But they all seem to say 1 degree c at most.

I will be running an Optimus waterblock coldplate. I asked their engineer and he said it would minimally effect the temperature.

Lehman

Yea, even if reverse flow makes very little difference, the water block vendors (EK and others), who have jet plate products offering, might not want to publicize it, as the jet flow is their product selling point. Reviewers who get free samples and/or sponsorship would play along as well (e.g. just avoid doing this particular test).

Just imagine how easy it is to "extra" test a water block with reverse flow direction (just swap 2 tubes at pump or water block end), if they are already doing a test on it. But somehow it was hardly ever done by any of the tech reviewers. Not even done accidentally by mixing up the inlet/outlet ports? One of life's mysteries...

 

[dKenGuru]

淘宝网 - 淘！我喜欢

淘宝网(Taobao.com)作为专业的购物网站拥有全球时尚前沿的消费者购物集市,100%认证网上商城及超值二手商品区，同时购物安全，产品丰富，应有尽有,任你选购,让你尽享网上在线购物乐趣！

item.taobao.com

 

[Nouvolo]

淘宝网 - 淘！我喜欢

淘宝网(Taobao.com)作为专业的购物网站拥有全球时尚前沿的消费者购物集市,100%认证网上商城及超值二手商品区，同时购物安全，产品丰富，应有尽有,任你选购,让你尽享网上在线购物乐趣！

item.taobao.com

Nice! Guess I can charge higher price than that guy if I make it better. I think I can make a better one than that ?

 

[Shatrod]

Have you decided on what you will do about the copper cold plate? Where to source them?

 

[Allhopeforhumanity]

@all

I am looking for some review or paper that talk about water block (jet plate type) performance with reverse flow direction. I can't seem to find any on the web. EK has tested GPU block and they get a difference of about 1.5° C at worst, which is nothing really. [Link]. But many also suggest that difference will be greater for CPU block (jet plate type).

Would appreciate if anybody can post some links of such, share your experience.

There seems to be a consensus that reversing flow direction affects performance, but there isn't much quantitative result, as to by how many degrees?

If you have a CAD model of a jet plate you're interested in, I could try to run a few thermal-flow finite element models of the block and housing to give you an idea of whether there is appreciable pressure/flow difference between the two directions. It's not the same as experimental results, but if it helps I recently ran a set of studies for work on some industrial cooling jackets, and they validated within ~10% of experimental results with matching trends.

Edit: Looks like these guys have some SEM images of their microfin design that I could use to make a model

Edit2 : Looks like common wisdom looking at several designs and in several forum posts is for CPU blocks to have maximum pressure impinging on the top center of the jetplate, distributing to the edge and then collecting from the rim. For GPU blocks, it looks like the flow is typically long the whole length of the fin array, but the jetplates look to be less dense allowing them to be pretty comfortably omni-directional. This is probably an optimization problem trading convection coefficient at the heat source for how much pressure you're willing to drop (or can drop before stagnation) and how much noise you're willing to live with. The latter 2 details are highly coupled to pump performance, so you'll likely see quite a bit of variability between DDC/D5/SPC as well as pumps from different vendors.

Edit3: Found a paper (warning PDF) on Si micro-channel design that has similar design considerations. And a second on local nusselt number optimization.

 

[Nouvolo]

Have you decided on what you will do about the copper cold plate? Where to source them?

I have sourced the cold plate, mounting brackets, and thumbscrews. Waiting to get some samples. Will post some photos later, they are similar to EK's

 

[lehman]

Something to look at is the apogee drive ii. It is a reverse flow pump block. Though it has a pin based cold plate design vs the normal fins.

I am not sure if it should really cause any real difference though. You will have the same amount of water flowing through the coldplate either direction. And most are designed around low restriction. I am probably missing something.

It could be water coming together vs splitting at the coldplate causes more turbulence and more resistance?

Lehman

 

[Allhopeforhumanity]

Something to look at is the apogee drive ii. It is a reverse flow pump block. Though it has a pin based cold plate design vs the normal fins.

I am not sure if it should really cause any real difference though. You will have the same amount of water flowing through the coldplate either direction. And most are designed around low restriction. I am probably missing something.

It could be water coming together vs splitting at the coldplate causes more turbulence and more resistance?

Lehman

There's definitely some subtlety in the flow behavior vs pressure magnitude at even simple geometries. It blew my mind that an entire PhD thesis could be dedicated to something seemingly simple as "Turbulent flow over a reverse step" as recently as 2002. There can be pretty significant vorticity induced convection differences when you consider path-traversed and at what pressure when flow channels are this small. I think the biggest consideration is between the fin pitch, the length before the flow returns to near laminar behavior, and the pressure head requirements of the pump.

 

[lehman]

I tried a simple test with the first prototype. Stock pump top vs the pump block reverse flow through an ek supremacy coldplate. Both attached to a 15ft length of tubing.

The stock pump top pushed water to just over 9.5 feet. The reverse flow through the coldplate just barely got 9 ft of lift.

Both were fed from a mini res at the same level as the pump.

Not the most scientific. But it did the job.

Lehman

 

[Nouvolo]

Some photos of the cold plate and mount bracket that I have managed to source, decent quality, and similar to EK Supremacy. Well at least I am happy with it? . Let me what you think.