Multiple PC Water Cooling suggestions

[chazz469]

OK...between my roommate and myself, we have about 6 pcs floating around in various states of disrepair. In the immediate, we have 4 that are up-and-running serving different purposes on our home network. I was wondering if any one had any suggestions as to the components that would be needed to water cool all four computers. One of them (my main one) is currently water cooled (see sig). The rest are air cooled. The current setup is as follows:

PC1 (his):
AMD XP2100+ (at stock for now)-->MSI KT3 Ultra-ARU --> 768MB PC2100

PC2 (linux storage):
AMD K6-2 500MHz @ 605MHz --> Abit KT7-RAID --> 128MB PC100

PC3 (VPN):
PIII 700MHz --> Iwill mobo (don't know the model)--> 512MB PC133

PC4 (mine):
(see sig)

I will be posting a site shortly that will have different configurations that I am thinking of. Basically I was thinking of getting a pump like hmale's 1000GPH pump, and having heatercore's (with fans) inbetween each system, then returning to a large resevoir (or bong). Maybe adding a few rads in the return. Would 1000GPH be enough? It seems overkill for one, but for four? And how many rads would be too many?

[#Rotor]

Ya... it's as easy as pie.... get a nice little reservoir with a fairly beefy pump, a couple of T or Y connectors, absolutely have to use 1/2" id tubes for the infrastructure. I would recommend medium flow high velocity blocks though...

connect all the pc's up with the blocks in parallel.

you will have to run a Heater core, properly ventilated.


in this picture you can see the core and the open reservoir with the 210GPH Beckit. note the temperature of the liquid, even with 4 overclocked DURONS and a chipset in the loop.


here is some pictures of two of the nodes, the other pc is my Dualy.

[Brad]

use a maze2 for that second athlon, and maze1c's for the other two cpu's. You could use a big 2x 120mm heatercore to cool it all down, or several smaller cores. Danner make fairly cheap 700 and 1200gph pumps, that'd be about the size I'd go for.

Also, I'd use 1/2" tubing to the two athlons and 3/8" to the others, simply because the athlons produce much more heat.

Lastly, try to use some big tubing everywhere else to keep the restriction low.

[chazz469]

This is the setup that I was envisioning.......


The hosing would be all 1/2 in. braided vinyl, until I could afford enough tygon (braided vinyl =$.45/ft., tygon=$2.35/ft). The rads would most likely be heater cores (maybe BIXs, but HC are cheaper and better), except for PC4 which would have the DD Cooler Cube that's already in it. The rad/fan units would NOT lay on the PCs just for airflow purposes. PC4 also already has the MAZE3, why use earlier MAZE blocks (honest question Brad)?
Feedback???

[bigben2k]

Consider bigger tubing.

[Cova]

Ok - I'm assuming that all of these computers are quite close together. If they aren't this isn't going to work (and any other way you do it will likely have flow problems with long tubing too).

I would run 2 PVC pipes of about 2" diameter along the back wall behind all of the computers, one for cool water to computers, one for hot water return. Spaced out every foot or two along each pipe have a quick-disconnect fitting.

Each computer would have just the block and some tubing coming out the back which would have the other end of the quick-disconnects on it. This way computers could be attached/detached from the system to work on them and such.

At one end of the 2 pipes they would be capped off. At the other end would be a single large (1000gph+) pump, a large rad with 2 or 4 120mm fans, and an old AT PSU to power the fans (unless you can get 120V fans for the rad). Plug the PSU (or 120V fans) and the pump all into a power-bar together and you can turn the entire cooling system on/off with the power-bar switch if you need to.

You should probably also have some type of pressure blow-off valve connecting the 2 pipes together at the capped end. So that if there are no computers plugged into the system that the pump can still circulate a little bit of coolant around.

[chazz469]

Interesting idea Cova. I like the simplicity. Something like this?

[jtroutma]

That sounds like a great idea. Just make sure you get a big enough pump to keep the pressure up in the main tube otherwise you will get a low flowrate in all the blocks.

(That setup looks almost like a hydrolic pressure system)

[chazz469]

Originally posted by jtroutma
(That setup looks almost like a hydrolic pressure system)

What's that?

I think that the pressure/flowrate could be assisted by maybe using a smaller "backbone". Having just read Bigben's thread on the importance of flowrate, I am very conscious of that. In that respect, I wonder if it might be better to use a smaller "system pump" (like around 300gph) and have supplemental pumps at each PC (like 4x100gph pumps) and then just use the "backbone" as the res.
Like this:


(Yes, I like drawing pictures. I'm visually oriented, if I can see it, I can understand it.)

[Cova]

I would take the bucket out of the picture and run it as a single, low-volume sealed system. But if built right it should be very modular (perhaps support up to 10 computers if you can afford the quick-disconnect fittings) and the cooling system itself should be independant of any computers connected to it.

Actually..., I just had 3 good ideas for a setup like this.

1. Go find that thread again that linked to the other site talking about vortex-coolers. The person building those was using washing-machine pumps that had like 2" fittings right on the pump. Then you wouldn't have a giant flow restriction coming out of the pump and going into the pipe. Those pumps also moved about 1700 GPH - enough for lots of comps.

2. I wonder about the possibilities of putting a pelt (or a few) between the coolant pipes - it would chill the coolant in the cold pipe and use the hot-pipe to cool the pelt. This would make the rad more efficient by giving it hotter water (bigger temp delta to ambient) and also give cooler water to the CPU's.

3. With the large volume of water in 2 lengths of 2" pipe that are each likely quite long, the pipe itself can function as a res. But I wonder if it might not be possible to use that large volume somehow as a large passive rad. Perhaps make the hot return pipe into 5 or 6 lengths (to keep large combined flow-rate/low resistance to flow) of 1/2" copper tubing. Especially if this is in a basement - you could lay all those pipes right on the cement floor and dissipate the heat into the ground.

[chazz469]

I think you must have posted this at the same time that I was drawing my picture. I did take the bucket out. Anyhow, I don't have the luxury of a basement, and as a matter of fact, my comps are in one of the warmest spaces in my house (maybe because of all the comps??? ). So, while I like the integrated RAD idea, I don't think that it would serve my purpose. I might need something a little more aggressive to cool that volume of water. Maybe a bong, or a car radiator. Most of this is brainstorming at the moment. The grand scheme is to custom build a table/workbench that will house/support the water cooling rig and the comps.

Also, where could I get those quick-disconnects. I've seem the before, but is that something I could get at Home Depot? Or only online?

[Cova]

Ya - I made that post based on your first picture...

I would stay away from a bong, but thats just personal preference. I like sealed systems - no evaporation out, no biological growth in.

A large heater-core should have no problem dissipating the heat from all those computers, especially if you can get a pelt between them to make the hot hotter and the cold colder. The hotter the water in the rad, the more effient the rad will be (regardless of it's size) - and the pelt means that even with hot water in the rad, you have cool water to the CPU's. Perhaps a motorcycle rad would work well.

As for the fittings - at least once a week someone on here posts about them - do a search.

[jtroutma]

Chazz:

Your "backbone" idea has good advantages. The reason why that would be a better choice for you is because your CPUs are not all the same. If you had say 4 durons all running at the same speed and temp, then a single large pump would work fine. The individual pumps would work more to your advanage (more flow to the warmer CPUs). I would still stick with a small car radiator to make sure you have PLENTY of surface area for cooling. You can Vmod 4 120mm fans or use a single large house fan for cooling. Either way, if you have the space I would go for the larger radiator.

[chazz469]

jtroutma: Thanks for the input. I do have access to 120mm fans that run on A/C current (or atleast appear to), so I'm planning to go with those. I don't think that a 20in. floor fan would have the same draw/focus that smaller fans would have.

Cova: I get what you're saying about the rad working more efficiently, but I'm curious as to why you think that's better. It's almost as if you're saying that one should heat up the water so that you could say that your rad is operating at 99.99999% effeiciency. I think that if your temps are good, or even outstanding, what difference does it make if your rad is only operating at 50% efficiency? Like I said in BigBen's thread about pressure and stuff, I think that there are two roles for a rad....initial cooling and maintenance cooling. Maintenance cooling may be only 50% efficient, but if it's maintaining a satisfactory temperature, shouldn't that be enough? I understand the need for efficiency, but I think that in this case efficiency should be relative. As a matter of fact, wouldn't the least efficient rad make the best one at the maintenance cooling stage???

[Cova]

I'm just an efficient guy I guess - I like to make everything efficient. But you're mis-interpreting me a bit on the rad thing. I don't want to make the rad perfect (or near perfect) - I want to make it more efficient so that a smaller rad, with fewer (quieter) fans can still dissipate as much heat. My idea to use a pelt between the water-flows to increase the efficiency of the rad is because I think it would be possible to use a fairly small rad and still accomplish the goal - perhaps even a decent sized heater-core with 2 120mm fans could accomplish the task.

In the (theoretically) very best computer cooling system, our water would be at ambient temp, and the CPU core would be significantly hotter. The rad would be 0% efficient. In a really poor cooling system the water would be hot, the CPU would be really hot, but the rad would be at least somewhat efficient. In eg. a car on a hot summer day your coolant may be near 100C (boiling) - now your rad is being efficient. A rad with a fan is capable of dissipating a TON of heat, if we can get it's delta temp compared to ambient up enough - but by doing so we tend to fry the CPU. Phase-change systems get around this by using the different states of the coolant to move heat, and keep the rad very hot (efficient) while the evaporator is very cold. Since I originally started planning my first water-cooling system my goal has been to try and make my rad at least semi-efficient without frying a CPU - one of these days I'll have a brain-fart and figure out a simple/elegant way to do it.

[chazz469]

Aaaahhh.....elegance. It seems that it is always the goal. A noble one at that!

I think that I can see what you're getting at a little better. So a rad is working more efficiently if it's absorbing/dissipating more heat relative to the temperature of the system. I wonder if an "open core" radiator would be more efficient that the tubed ones that most of us use. Something like this (it's not the most idicative drawing, but I think you get the point):

[Brad]

those pics show basically what i was thinking of. I'd go for 2x 200gph pumps for the two athlons and 2x 100gph for the other two, and a big 700gph Danner as the main pump