Choosing the parts, their orientation and design, and connectivity

[derelict]

New to the forum and watercooling; I've yet to install water. After gleaning through the forum for waterblock, reservoir, and radiator design ideas, I'm still unclear on whats best/ideal . So I'll ask a few general and specific questions in hopes that someone can help me out.

-Radiadtor: A 14x3x1.5 (approx) heatercore for Renault

I was thinking I could put 4 80mm fans inline across this radiator. However I'm not sure if the physical orientation is important. Does it matter if its standing tall like the picture? Can it be upright but on its side? Can it be laying down flat? How would this effect its performance? Any ideas? Is the radiator choice good or bad for any other reasons, like the barb size?

-Reservoir: I would prefer making my own, and the physical shape would depend on the radiator dimensions, orientation, and whether it was internal or external. I'm sure metal based reservoirs are better for dissipating additional heat, but by how much? Lots or just marginal? Also With tube style reservoirs, are there any gains to be had for having the inlet up high and outlet near the bottom (assuming reservoir is upright)?

-Waterblock: I again want to make my own. I've read conflicting things and I can't seem to find a definitive answer; is high-flow, laminar wbs or high-turbulence wbs more effective? Some make wbs with straight through, low resistance paths while others bore a grid of deep dimples then cut paths between them - I can't tell which I ought to be doing.

---Direct water-to-processor: Has anyone tried making a shim for direct water-to-hs for 775/939/AM2 processors? Is it more effective? Again should water splash on for high turbulence or glide past in a smooth laminar flow?

-Pump choice: SWIFTECH MCP655-B (I don't remember who else has the same pump under a diferent name) is what I'm likely going to buy and use, it seems to be an all around good performer and I can power it with my PSU if everything fits internally. I don't have multiple GPU/CPU, my setup will consist of only CPU GPU and NB WBs. Any additional comments on this?

-Tubing: I've seen varying sizes of tubing and was wondering if there is a "can't go wrong" size for CPU, GPU, NB. Regarding the CPU wb does one want the highest amount of volume they can push? Of course there is a loss of speed with an increase in tube diameter, so is there a sweet spot? Is that wb design dependent?

Sorry for repeating a few common thread ideas here, but I just wasn't able to find what I was looking for. I'm not in a hurry to buy-fab-assemble all this, so I'm asking questions so I can learn a lot and do a good job the first time around (I can be a bit of a perfectionist). All comments, ideas, knowledge, and help gladly welcomed

[jaydee]

Shouldn't matter what way the radiator is. As long as air flow is going through it. i would suggest however you get one designed for PC water cooling as they are pretty cheap these days unless you just happen to have that heater core laying around. Also make sure it is copper brass not aluminum.

As for a rez however you want to go. They have cheap plastic ones that fit in the bay for like $12. http://www.dangerden.com/store/produ...1&cat=6&page=1

If you make your own make sure it is copper and/or brass. Personally I just drill a hole in my radiator and solder on a fill/bleed tube and use that as the rez. Swiftech has a new radiator/rez out also. http://www.heatsinkfactory.com/swift...reservoir.html

Water block... If you want to do it yourself go ahead. Drill holes and connect the dots. http://www.procooling.com/index.php?...icles&disp=128 Will not be much more just to buy a good commercial one though. Even on e-bay for dirt cheap.

Pump. DD12V-D5 Pump Variable Speed is good. Can adjust the RPM for noise/performance.

Tubing I prefer 3/8"ID min and 1/2"ID max. Anything over 1/2" tends to kink to easy. Anything under 3/8 maybe reduce flow to much however 1/4" will work but most blocks are setup for 1/2".

[jaydee]

Originally Posted by derelict

---Direct water-to-processor: Has anyone tried making a shim for direct water-to-hs for 775/939/AM2 processors? Is it more effective? Again should water splash on for high turbulence or glide past in a smooth laminar flow?

Forgot to add. Forget this idea. The IHS needs center pressure to work right. Also been there done that and direct die cooling does not work as well as a good water block.

[pHaestus]

As far as overall system design goes, be CONSISTENT.

Think about what you want to accomplish first: Top performance or least possible noise? You CANT have both.

If your main goal is low noise, then you'll be undervolting 120mm fans probably and a thin radiator will be your best bet. If you're undervolting fans and using a thin radiator then you'll probably also be undervolting a 12vdc pump. If the above is all true, then you won't lose too much performance by going with 3/8" tubing, and it'll result in a more compact and clean system to be honest.

Flip it around though and say you're after max cooling. You'll be using higher pressure 120mm fans (or 160mms) and thicker larger radiators (heatercore more than likely? Thermochill perhaps). You'll be using a pump that can generate a high flowrate with a lot of resistance in the loop, and to get the max flowrate possible to your blocks you'll be using 1/2" or even 5/8" tubing.

It looks like you're kinda mixing and matching ideas and trying to do things "on the cheap". There's nothing at all wrong with that. If saving money is the biggest thing, then I'd say go with a heatercore, but a larger width so you can make a shroud and fit a 120mm fan on it. Then pick your pump based on either availability where you live or price. I really like the 12vdc pumps that are sold under various names (MCP655 I think is the swiftech one). Like Jaydee said I'd just check ebay for a newer waterblock (Swiftech Apogee, Dangerden, Dtek, etc) and buy whatever is on sale. Swiftech has clearance waterblocks too.

Hope that helps

[derelict]

Hey guys thanks for the valuable input. PH is right by guessing I'm leaning toward a budget watercooling setup; I'm married now and still a student = 0$ one benefit to the student scenario tho is I have access to a machine shop (with CNC) and free materials (like copper blocks), this is partly why I want to do the DIY route (plus I really like designing and fab'n stuff and want more experience doing it).

If I had to choose between noise and performance I'll take less noise. Currently I have an older 939 Opt 165 @ 2.75ghz, 1.35v on air. I believe that switching to water will enable me to go past this point, hopefully to 2.9+ (however getting more of an OC is not as important to me right now as just getting more experience with water cooling and design and fab). But before I put water in this system I'm going to experiment with an older 478 P4 3ghz system that I have (and don't care if I ruin). Eventually I'd like to try and do watercooling for my HTPC to make it real quiet, but that will only happen after I have some experience doing watercooling.

If there are no gains to be had from DIY parts, like saving $ or better performance, then I'm fine buying commercial products, I appreciate the links too.

@Jaydee: Thanks for comments on direct-die watercooling, I saw that you had done it once before with an older socket A processor (with a naked die), but haven't seen it done with modern processors with IHS. I didn't think about pressure on the CPU as a necessary component of efficient cooling, but that makes perfect sense.

[derelict]

As far as access to parts is concerned I really don't know. I live in Denver suburbs so I'm sure I can track down most of what I need, but was actually expecting to get everything online. I believe I could get a suitable heatercore from checker or some other car parts store. I don't want to go to a junk yard to pull one out tho, I'd rather have a nice clean one. However I suspect that alot of car heatercores are not copper/brass? I chose that renault heater core because I envision having it attached to case door upright, and right next to it an acrylic see-through reservoir. On second thought though I guess having that equipment on the door of the case probably makes getting into the computer a real pain unless you make your tubes to them extra long.

Again I'm just shooting out ideas and looking for help from those who have the experience. Thanks again.

[derelict]

Here is a sketch of what I thought I could do if I got the Renault or similar heatercore:



The fans would be blowing out of course, and I've considered making a slit in between the res and rad and making a scoop for the fans, so they're not just getting hotter air from inside the case but that might be a bit over the top

[derelict]

Here is an external concept:
If its not very clear, there is a computer shelf on my desk that has space behind the computer, a place I thought I could put a res and rad if the internal route didn't work.

[pHaestus]

Here are a couple articles that might help you a little bit: http://www.procooling.com/index.php?...ticles&disp=73
http://www.procooling.com/index.php?...ticles&disp=37

Second one is kinda nerdy but whatever... I can't off the top of my head recall the fedco number of the heatercore I always used. I'll see if I have it in my notes somewhere at home...

[billbartuska]

No pickies............

[jaydee]

Originally Posted by derelict

@Jaydee: Thanks for comments on direct-die watercooling, I saw that you had done it once before with an older socket A processor (with a naked die), but haven't seen it done with modern processors with IHS. I didn't think about pressure on the CPU as a necessary component of efficient cooling, but that makes perfect sense.

Sorry I was short on my answers but had little time. To further the direct die the main point it fails is lack of surface area. With die becoming smaller the heat density gets worst. The same amount of heat in a smaller package makes cooling it harder. This is one reason heat spreaders were added and for the same reason a good water block is better than direct die. The more surface area you have to cool the better. Spread the heat through a water block or heatsink and you can get more coolant contact.

Them main reason I dropped direct cooling the IHS is there is the heat spreader is not bonded to the cores. So if there is not pressure on the heat spreader then the heat spreader is not contacting the die(s) well and thermal transfer is fubar.

Another reason to avoid it is damaging parts. to many failure points. If you are on a budget that is not something you want to work on.

Now back to radiators. I still see no reason to get a PC specific one. They are just as cheap as a heater core if not cheaper and better designed for lower pressure fans that we use. Black Ice Pro $27.95. http://www.dangerden.com/store/produ...3&cat=5&page=1

[derelict]

ok yeah if the DD res is cheaper (most heater cores I've looked at are $30+) and performs as good or better I'll go with that.

-I had a question about res location in the loop. Does it have to be in between pump and wb's? I assumed the loop would go pump -> wb -> rad -> res -> back to pump? Are there benefits of having the res right after pump? Air trapping done best after pump? I assumed res would be good right before pump to help the intake. I read this from one of the threads that pH linked.

[jman1310]

Originally Posted by derelict

ok yeah if the DD res is cheaper (most heater cores I've looked at are $30+) and performs as good or better I'll go with that.

-I had a question about res location in the loop. Does it have to be in between pump and wb's? I assumed the loop would go pump -> wb -> rad -> res -> back to pump? Are there benefits of having the res right after pump? Air trapping done best after pump? I assumed res would be good right before pump to help the intake. I read this from one of the threads that pH linked.

put the res on the suction (intake) side of the pump
if possible use a pump res combo

[pHaestus]

Res before pump is good, but make sure you're not putting a lot of restrictions (via barbs and shit) right before the pump's intake.

[jaydee]

Originally Posted by jaydee

Forgot to add. Forget this idea. The IHS needs center pressure to work right. Also been there done that and direct die cooling does not work as well as a good water block.

I got thinking about my statment here and decided to prove or disprove it.

http://forums.procooling.com/vbb/sho...022#post169022

[Brians256]

I'll make a comment about dangerden's reservoir. I don't like it, personally.

1) It's horizontal, meaning that if it is even partially empty, it makes noise.
2) It's not big enough, so you need to monitor the fluid level more often. Remember, tygon/PVC tubing is actually not water tight. Over time, water WILL evaporate out of the system.
3) The fill mechanism is awkward. Since it is horizontal and shallow, you have to work at it to get it to full without spilling the coolant all over the computer. Too many iterations of fill res then start pump for 5 seconds to move coolant around the loop.
4) The fill point cap is air tight and a bit awkward to open/close. Airtight is BAD because over time as water leaks out the tubing, the tubing will tend to collapse. You can get around this with watertight material but it's awkward to use. The cap is opened with a quarter but it's not as convenient as a finger-accessible cap.

My preference is a tall/narrow reservoir with a big honkin cap on it. Make it about a quart in capacity. I used a 24-oz pepsi bottle at one time and the only problem I had was keeping it well sealed. Even better was a 1 qt tupperware liquid container (it looked like a quart container for milk). It was big and awkward to put inside a normal case, but it was superb at doing the job. It was VERY quiet, held a lot of reserve capacity and was extremely simple to fill because of the wide-mouth opening. That system ran for about 3 years with NO changes.