Watercooling rack.

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

hagak

Member
Oct 22, 2012
92
4
8
One important thing you want to research when picking a quick disconnect is how much fluid they allow to leak on disconnect. They all leak some amount, the better ones leak an extremely small amount (less than the sweat on your hands amount). Those better ones are really pricey though.
 

vv111y

Member
May 6, 2011
76
4
8
Niagara Falls, Canada
Yesterday I spent a bunch of time researching doubling up power and fluid distribution with poor results. IBM and Intel are currently working on doing this on the chip, so I was hoping.
- power over copper + water = corrosion. Even pure RO water is corrosive.
- non-corrosive fluids have poor thermal conductivity so far that I could find
- hollow copper tube is great for AC power (skin effect), not so great for DC, which is what I was hoping for.

I was hoping for doing this as a way to simplify and reduce cost. I want to take advantage of economies of scale and design according to that.
What I was thinking:
- remove server PSU's , have it rack scale.
- use car battery charger & car batteries for UPS - already been done
- rack scale thermal - Your project here

@hagak I figure if the quick connect is at the back of the box, it is minimal concern. Just use a cloth to grab it and soak up any drips.
 

hagak

Member
Oct 22, 2012
92
4
8
@hagak I figure if the quick connect is at the back of the box, it is minimal concern. Just use a cloth to grab it and soak up any drips.
I think you underestimate the amount of water a cheap quick disconnect will leak when connect/removing. Also you want to make sure they seal well when connected.
 

RimBlock

Active Member
Sep 18, 2011
837
28
28
Singapore
The Coolance items have arrived today and I will take a few photos tonight when I get home.

The CNC machine has an ETA of tomorrow on the FedEx site but I am not convinced as it is currently still in China. I willprobably look at getting a pump over the weekend.

I have had some friends phone around for Alu and Acrylic pricing here as no-one advertises prices over here, you have to call, and the foreigner tax is very real :mad:. Pricing for two 1/2" thick 6" wide by 12" long pieces of alu is around US$15.

RB
 

hagak

Member
Oct 22, 2012
92
4
8
The Coolance items have arrived today and I will take a few photos tonight when I get home.

The CNC machine has an ETA of tomorrow on the FedEx site but I am not convinced as it is currently still in China. I willprobably look at getting a pump over the weekend.

I have had some friends phone around for Alu and Acrylic pricing here as no-one advertises prices over here, you have to call, and the foreigner tax is very real :mad:. Pricing for two 1/2" thick 6" wide by 12" long pieces of alu is around US$15.

RB
$15 US for a 2 1/2"x 6"x12" AL is a freakin steal, not sure what foreigner tax you are talking about. May I ask where you got that price cause I have never seen it that cheap. FYI - 1/2" thick AL is pretty damn thick.
 

RimBlock

Active Member
Sep 18, 2011
837
28
28
Singapore
The Koolance items (Norco drive tray in the background for an idea of size).





The quick disconnects are quite large and that may pose a problem. They are also heavily spring loaded so when you disconnect the plug and socket close very quickly. I suspect there may be a little dribble but a cloth should cover it. Will test to make sure though.
 

RimBlock

Active Member
Sep 18, 2011
837
28
28
Singapore
Some more pics. I have put in a half height, half height PCIe card as I would like to keep that slot available.

Just testing the quick release fittings in the mezzanine cutouts. I have also taken the chipset heatsink off.





As you can see from this pic, I need to make the mezzanine holes a bit bigger.



Top view. Areas I want to aviod are in the red boxes. Usable areas are in the black boxes. The white 'A' is the chipset and the black 'B' shows the narrow gap if I want to keep the ability to use PCIe cards and it also shows where the BIOS reset and other utility jumpers are which is a bit of a pain.



The quick disconnects are spring loaded on the disconnect side rather than on the bulkhead side so when you disconnect it pushes back against the bulkhead. I have some concerns that the mesh area by the mezzanine slot is may be easily deformed by this. The quick disconnect is also quite heavy.

Fall back position is to have the quick disconnect nearer the T-splitter and standard bulkhead barbs on the rear of the node.
 

kendrick

New Member
Dec 23, 2012
17
3
3
The quick disconnects are quite large and that may pose a problem. They are also heavily spring loaded so when you disconnect the plug and socket close very quickly. I suspect there may be a little dribble but a cloth should cover it. Will test to make sure though.
I have a couple racks I plan to do this with as well. there are 3 areas which i still have to work out details from. Since I hate 1u's except for a fattwin/blade usage Im putting my equipment in 2 and 4 u cases. frankenstining some dell 2650's and hp dl580's :D I was hoping to cover memory as well as cpu/chipset. I plan to have a radiator in the crawl space for dumping the heat.

1. swingarms/rail configurations. I have been looking in to making a couple brackets that attach to the swing arm for the hoses. I havent decided if im going to use swivel's or not. what were you thinking on that matter Im not sure i trust the tubing that much and it also makes for pinch points.

2. rack flow regulator presure/flow sensors. Im half way tempted to put a regulator on each branch possibly a fuel reg or some other dynamic flow control. that would allow me to limit pressure and rate by sensor rating. though may be stuck with just good ol pluming style valves. I would have to push the system to 100% set the flow rate to keep the temp to acceptable rates and run it on all servers that are liquid cooled at the same time to balance the flows for the different equipment loads.

3. the auto shutoff also worries me. drip trays for lack of a better word would be about the only thing i can think of. maybe some 29 ga sheat metal with a light rubber coating at the bend to make a proper seal aginst the bottom of the case as it slides in to home position so that it would divert any liquid away from the back of the next pc. most cases are solid bottom and may have their first holes around 1/2 way up the case for the rails. back and front would be where it would leak out of possibly using some filter media for air conditioners to catch dust AND spray if it ever happens may be a way of dealing with it trying to go forwards. put a small rubber ridge just before the front fans so that the liquid doesnt try to go further forwards and a few drain holes in the back to the drip tray. since your setup looks to be a fattwin it wouldent work for you.

hydro electric shutoffs are expensive I may look in to what it would take to make one. I have been thinking of looking in to beer/soda accesories as they may by less expeinsive since they are rated for fluid even though its high pressure. some hydroponics stuff may also work.
 

RimBlock

Active Member
Sep 18, 2011
837
28
28
Singapore
I have a couple racks I plan to do this with as well. there are 3 areas which i still have to work out details from. Since I hate 1u's except for a fattwin/blade usage Im putting my equipment in 2 and 4 u cases. frankenstining some dell 2650's and hp dl580's :D I was hoping to cover memory as well as cpu/chipset. I plan to have a radiator in the crawl space for dumping the heat.
Hi Kendrick,

The main server I have is the Dell C6100, like many others who read here I suspect ;), which is like a Fat Twin. The challenge with these is the lack of available space as the nodes have a 1U height and there is not a lot out there for 1U water cooling that is not 'call us for a quote' expensive :).

I finally bit the bullet and got a DL380 G6 and looking at the insides with the masses of space in its 2U chassis and the dual CPUs being next to each other, it is like a dream compared to trying to do the C6100 nodes but, if I am jumping in to this then may as well go both feet first and get the hard one out of the way so all the others will just be like a walk in the park...

1. swingarms/rail configurations. I have been looking in to making a couple brackets that attach to the swing arm for the hoses. I havent decided if im going to use swivel's or not. what were you thinking on that matter Im not sure i trust the tubing that much and it also makes for pinch points.
Yes, there would need to be enough space on the turn to allow the pipe to bend and not kink. Corrigated piping may help but it would mean the arm channels may need to be quite wide... unless there is a joint where the arm bends. I saw last night that Koolance does a 30Deg movable ball barb. Would not be perfect but may help. I am thinking of using these on the block so I can have an angled entry point. Alternatively, if the arm were made as a lower and upper part (rather then side by side) then a couple of swivel right angels could be used.

2. rack flow regulator presure/flow sensors. Im half way tempted to put a regulator on each branch possibly a fuel reg or some other dynamic flow control. that would allow me to limit pressure and rate by sensor rating. though may be stuck with just good ol pluming style valves. I would have to push the system to 100% set the flow rate to keep the temp to acceptable rates and run it on all servers that are liquid cooled at the same time to balance the flows for the different equipment loads.
Load may be variable so an automatic system would be prefered, but no doubt comes at a cost. I do have a concern when plumbing in more than one server of differing types that the flow though 1 server may be less constricted than another causing a higher flow rate and potentially starving other servers which have more difficult flow paths.

Flow regulation is probably well worth considering but for me this will be for a later date.

3. the auto shutoff also worries me. drip trays for lack of a better word would be about the only thing i can think of. maybe some 29 ga sheat metal with a light rubber coating at the bend to make a proper seal aginst the bottom of the case as it slides in to home position so that it would divert any liquid away from the back of the next pc. most cases are solid bottom and may have their first holes around 1/2 way up the case for the rails. back and front would be where it would leak out of possibly using some filter media for air conditioners to catch dust AND spray if it ever happens may be a way of dealing with it trying to go forwards. put a small rubber ridge just before the front fans so that the liquid doesnt try to go further forwards and a few drain holes in the back to the drip tray. since your setup looks to be a fattwin it wouldent work for you.
Koolance also do a leak detector, but it is fairly simple (wet contacts = red led and buzzer). A bit of hacking could probably wire it to a shutoff and dump valves.

As for the other stuff, it sort of turns it in to a very custom setup for each install. I am trying to make this as generic as possible (give or take custom waterblocks) so if someone else has an interest and gets in contact I can more easily provide a solution for them as well. This is early days though :). Most companies would rather not have their shiny new server drilled with holdes invalidating warranties (although watercooling them also probably has the same effect ;) ).

Just looking at Koolances coolent flow rate meter, it can provide feedback like a fan. It could possibly be rigged to the server fan headers to allow for feedback and appropriate alarms to be set for warnings and shutdown should low pressure occur (ie, a leak). I would imagine the signal would have to be spit to the 4 fan headers for the C6100, in my case, with something worked in to the setup to take in to account the control wire to change fan speed. Maybe that could control a variable speed bpump built in to each branch off the main feed but that could get expensive and calibrating the pump speeds would be tricky.

hydro electric shutoffs are expensive I may look in to what it would take to make one. I have been thinking of looking in to beer/soda accesories as they may by less expeinsive since they are rated for fluid even though its high pressure. some hydroponics stuff may also work.
There is an aquarium shutoff valve here which may be good to take a look at. Thing is that it opens when energised which may be the wrong way round and also wastes power just to keep it open. May be interesting to look at and reverse engineer.

Far too much to do for me to do that at this point though.

RB
 

RimBlock

Active Member
Sep 18, 2011
837
28
28
Singapore
So the CNC should be delivered today if I can get someone at home to receive it.

I have been playing around with Design Spark Mechanical which is free and works with solids. I have no real experience with cad but I have managed to put a simple waterblock design together. I may post up a first draft tonight or over the weekend. I am looking at a square base with a pot for the water chamber. This would allow an angled inlet. Along the top endge next to the ram slots is a rectangular block with a hole through it for the return feed. All of this is in the 1U air heatsinks profile. This is just for a single CPU at the moment. I will expand it to a two CPU waterblock if needed once I have a bit more experience with the design software.

DesignSpark seems to be pretty easy to use but I still need to find out how to duplicate items and change the drawing plane to the inside faces of objects. At the moment I can only draw on the outer edges (i.e. the original square cubes surfaces) and not on the cylinders sides to place the inlet hole or do come more accurate design on the inside of the water block.

I will still have to wait for a week or two until I can get the H60 coolers but can get some practice in beforehand.
 

RimBlock

Active Member
Sep 18, 2011
837
28
28
Singapore
Wow! That looks amazing RB!
The QDR fittings look really nice but I think they will need some more structual support.

It should look nice when complete, hopfully.

Still trying to decide on two piece aluminium or bottom aluminium and top acrylic. Not much price difference here.

I also need to find a supplier of seals to seal the two parts when they are screwed together. Usually it is a compressable rubber ring, although I could cut our a rubber gasket if needed.

If anyone knows where I can get these seals then please let me know.

RB
 
Last edited:

RimBlock

Active Member
Sep 18, 2011
837
28
28
Singapore
Ok, my question answered on the silicone seals.

Here is an instructables on making your own silicone o-rings. I just need to find a small enough tube to use for the seal.
 

RimBlock

Active Member
Sep 18, 2011
837
28
28
Singapore
Initial single waterblock idea. This waterblock is for the CPU closest to the rear of the sled with the square part on the side with the ram slots.

There are no fancy channels and the white cutout is where the copper cooling plate from the H60 will go (hopefully).

I still have to put in the second hole in the main 'pot', some posts to screw the lit to and the threads for the feed holes.



RB
 

kendrick

New Member
Dec 23, 2012
17
3
3
Yes, there would need to be enough space on the turn to allow the pipe to bend and not kink. Corrigated piping may help but it would mean the arm channels may need to be quite wide... unless there is a joint where the arm bends. I saw last night that Koolance does a 30Deg movable ball barb. Would not be perfect but may help. I am thinking of using these on the block so I can have an angled entry point. Alternatively, if the arm were made as a lower and upper part (rather then side by side) then a couple of swivel right angels could be used.
I was looking at this about a year agao and had found some 360 swivel's when both sides were at 0 deg it looked like a u. It would swivel at the bottom of the u shape. they were about an inch or so thick so it would need a 3u ish case so you could put a pair at the top of the case so that you could tie the bottom of the upper and the top of the lower together for the swing arm. and then the same at the other 2 joints. I was thinking of having a bracket tied to the "back" of the swing arm as close to the end of the swivels as possible so that i could bend some sheet metal 3/4 the way around the swivels so they would not rub when closed but would be secured to each arm to keep any strain off the hinging action.


A for the seal is it possible to use rvt/rtv? silicon seal like on a car. its fairly cheep and i know it handles some harsh chemicals.
 

vv111y

Member
May 6, 2011
76
4
8
Niagara Falls, Canada
Hey Rimblock, did you look into Tom's watercooling faq? It looks comprehensive.

***READ FIRST*** Tom's HW Watercooling Sticky v2.0 - Water Cooling - Overclocking.

There are links to other watercooling guides, ie. extremecomputing, overclockers, etc. So works as a directory.
Spreadsheets for calculating needed capacity are supposed to be at those links, if you still need that.

I also found this guy who provides an info site on general industrial fluid pumping
Mc Nally Institute. It's a teaser for his his CD, but there appears to be a lot of info on there.

example, if utilization varies a lot:
Remember that with a centrifugal pump if you change its capacity you change the pressure (head) also. A rotary or positive displacement pump is different. It puts out a constant capacity regardless of the pressure.
pump selection

PS this is O/T so just a quick mention, I looked more into power over copper tube and it might be doable. I'm excited about that, there is a lot of copper there and it should work as a low voltage, high amp DC power bus.
 
Last edited: