WC Advice needed for Very Extensive Custom Case Project

[zackbass]

Hi everybody! This is my first post here in about 3.5 years so I might as well be a new member. Please don't get scared off by the big post, there's a lot of important info and it really is an interesting problem for those that like WC system design.

I'm in need of some water cooling advice in preparation for my new computer. Of course if this was normal system I wouldn't bother you all like this, but this one is a little different. I'm about halfway done with the case, it's a complete from-scratch build with stainless steel tubing, but the special thing about it is that the frame is part of the water cooling system (the second one I've done this way, here's the first:
http://i5.photobucket.com/albums/y156/zackbass/IMG_0023compressedSmall.jpg ) and it's being built like a safe. That means innocent things like unguarded airholes are a no-no. I've got a detailed worklog over at bit-tech:http://forums.bit-tech.net/showthread.php?t=89645 that I'd suggest skimming to get a feel for the design but I'll lay out a summary here.

Here's a render that shows off the basic structure:


Its footprint is about 19.5"x14.5". The large bottom section and the two uprights are connected internally and form the reservoir while the connecting section up top is internally separated and is the distribution manifold. The two side doors lift up on hinges at the top. A transmission cooler will be mounted on small standoffs to the slanted section of the outside of each door.

The system is going to be entirely fanless, and to add to that it's being built like a safe so internal airflow is going to be almost nonexistant. I won't be overclocking, but the load on the system is going to be pretty substantial.

Here's a partial list of major heat contributing hardware inside the case:

Dual Core Pentium 4
Undetermined mobo
Antec Neopower
Two 7200RPM HDDs
High end video card
Eheim 1048 (I hope it's beefy enough)

Now, I wouldn't try something like this if I didn't know what I was doing, so there's no need telling me that it's too hard or that it'll never work, so let's get down to business and make it work.

My two transmission coolers just came in today. They are Fluidyne units, but they're unlike anything I've ever seen before. Here are some pics:




In the center of the bottom picture you can see the internal structure of the tubes. It's a very interesting design, my gut feeling is that they're extemely effective but pretty restrictive (confirmed by the "put my mouth over the inlet and blow" test). I have yet to decide whether to run them in series or parallel. Some air holes will be drilled through the doors directly underneath the rads to get some air circulation (the rads guard againt a breach through the holes). I've yet to find a place where I can put the inlet holes and not compromise security. I may just have to pick a spot out in the open and use some steel mesh to cover it.

The power supply modification is almost complete. I decided to go with a copper riser connected to a U-tube. It's all welded with near 100% penetration, so it's basically solid copper (no nasty solder interfaces to worry about). This is it in its most finished state:



So to recap we have the following to work with:
A custom case with a very large reservoir that slowly bleeds off heat
An Eheim 1048
A water cooled PSU
Two restrictive but highly effective rads
Nearly zero airflow
Keeping it in the spirit of a secure safe

What I need is suggestions on how to pull this off effectively and details on potential trouble spots. What components need to be water cooled since they don't have any airflow? Should I be worried about the northbridge or mobo power circuitry? How about the hard disks? I can whip up waterblocks for anything no problem, I have a Bridgeport milling machine in my shop and know how to use it.

I also need to pick waterblocks for the CPU and GPU. I'm not afraid to lay out some cash for the good stuff, I've come this far already. I would love to have someone's custom work but I'm not sure of what's out there and how to buy it.

Congratulations on making it to the end of my question(s), any and all help besides the assumed "you're crazy and it'll never work" is very much appreciated.

 

[sunrunner20]

I can already see one trouble spot. The PSU is still going to need some amount of air flow, granted you have water-cooled the greatest sources of heat inside the power supply, the coils also need some amount of cooling. Not to mention, PSUs dump an insane amount of heat into a cooling loop. (often they can dump as much, if not more heat than the CPU)
Also, I would check to see if you are getting any charge on the copper plates/tubing inside the PSU before doing anything with it.

Also, reservoirs do not really make good radiators. The concept of more water= more cooling capacity has been an issue for a long time, more water simply gives the appearance of better cooling. In reality the extra water simply takes longer to reach its equilibrium.

 

[SewerBeing]

Ok the eheim will probably not work I recommend a Swiftech MCP350 because it has high head and does not dump a lot of heat into the loop. Also that radiator is probably a very poor choice for this. It is true that more surface area does equal better cooling. However there will be almost no airflow across them and if you look at the numbers the BIP outperforms a BIX with less airflow because it has less restriction. Given the density of that rad it is probably a poor choice unless you get 100+cfm across it. Now the heat also has to leave the case so these rads would need to be mounted on the outside of the case otherwise said heat stays in the case and it overheats rather quickly. Also watercool the northbridge since thats a noisy and annoying fan

 

[zackbass]

I looked up the efficiency numbers on the Neopower to get a better idea of what it'll dump into the system, it looks like about 70w. That's quite a bit to deal with. I did a bit of testing on the PSU before I ripped it apart and the coils didn't seem to get hot enough to concern me, but that wasn't under much load. The only way I can think of getting some airflow to the PSU is to make it an air inlet and try to use convection to get it moving. I'm already on top of electrical isolation of the PSU WB.

I'm very familiar with the whole small res/big res argument. My personal feeling is that the longer it takes to reach equilibrium the better since it smooths out the temperature transitions. Typically a system is only under load for short bursts so the lag of the system coming up from idle can be used to an advantage since it may be back to idle before the equilibrium is reached. Every bit of delta T across the interface helps.

As for reservoirs making bad radiators, that has never been my intention but in my experience with my current system it helps out quite bit even though stainless is a horrible conductor. I haven't quantified how much it does, but the frame tubes are always feel warm enough to contributing to cooling. It's hard to compare the system to a normal res since it has such a large surface area in relation to volume and its metal construction.

Convection is -the- tool I have to work with. Air is going to be picked up from the bottom of the case then slowly heated as it rises, eventually to the holes in the doors right below the radiators and then pass through them. Does anyone have data on how radiators perform without fans? It seems to me that it would be an entirely different game from what is normally dealt with. In other words, less airflow isn't nearly the same thing as almost no airflow. I would think that SewerBeing's concerns about the density of the rad wouldn't really apply as there is so little air passing over them and that I should be more concerned with getting as much heat into the small amount of air as possible and letting convection do its work.

Opps, forgot about the pump concerns. How does the MCP350 compare to the Eheim noise and vibration wise? How much more heat will an Eheim 1250 dump into the loop than the Swifty? Remember, this is being built to run silently, I'm more than happy to sacrifice degrees Celcius all the way up to the safe operating limits of the equipment in exchange for noise reduction.

 

[SewerBeing]

Ok now if I remember all of this correctly the heatflow is the greatest if there is a large difference in temps between the two substances. So you should be concerned with something that has low air restriction so you get more air over it and maintain the greatest rate of heat transfer. So my concerns are indeed valid when applied to your situation. You do not want the air sticking around since the rate of heat transfer decreases exponentially as the two substances equal out in temperature. Also I do not think convection is the most efficient way to get rid of heat at those temperatures (just a hunch I have). So I stand by my concerns however the only way to find out if for you to test it. Also more volume is probably a good idea for you especially if its held in copper.