Krazy's WC project worklog

[krazy]

I started out by doing a lot of lurking and research on these boards, and once I had an idea of what I wanted to build and how it could be done, I started shopping around for parts.

My basic goals for the project are as follows:
- I want it to be very quiet
- I want it to perform quite a bit better than air cooling can
- I want to cool everything that needs active cooling with one water loop
- the system should have only two fans (on the radiator) unless this is decided to be a bad idea
- it should look nice when complete
etc...

I picked up a Danner Mag3 from reefgeek.com, a WhiteWater from d-tek, and pretty much everything else from cooltechnica. My local hardware hank has proven to have very little useful stuff save for some nice brass hose barbs.

I got the whitewater and a shipment from cooltechnica yesterday afternoon, and I was running an ice-cooled computer by suppertime. My heatercore needs some attention before it can be incorporated into a case, so I started out by running a dishpan full of cool tapwater through just the CPU block. It cooled very well with the chilly tapwater.



After 20 or 30 minutes, the CPU temperature was starting to rise. I put a bunch of ice in the dishpan, and was able to bring the water down to a few degrees above freezing. Needless to say, the CPU cooled very well like this, and I was even beginning to get small amounts of condensation and fog on the tubing and the block's base.



After a few hours of ice cooling, I had used up all of the ice in my freezer trying to see how low I could get the CPU temp. I shut the machine down and fabricated a wooden stand to hold my heatercore over the dishpan. I don't have proper barbs on the core yet (I'm just shoving the 3/4"OD clearflex inside the 3/4"OD tubes) so I wanted to have the connections over the dishpan to contain any dripping. Once I had the heatercore mounted and hooked up, I made a cardboard shroud to hold my two 120mm Roton fans above the core in a suction setup.




Running the fans at 5v makes them undetectable over the noise of my hard drive and power supply fan (I plan to WC and/or noiseproof these eventually). The ghetto shoved-in connections to the heatercore leak maybe a drop every hour or two (much better than I expected), so I can now justify putting off adapting the barbs for a while. The water temp (with CPU idling or at low loads) leveled out around 80 degrees F.



cooltechnica botched my order and only sent me two of the twelve hose barbs I ordered, so I'm holding off on adding in the GPU and NB blocks until the rest of my hose barbs arrive in a few days. (I explained the situation and they said they'd have the rest of them off to me ASAP. good guys.)

I'm in the process of building a homebrew double 5.75" bay reservoir out of acrylic. I have some good (I think) ideas for making blocks for my HDDs and FETs, and I'm still brainstorming how I'll cool my PSU.

I plan to update here when I make new developments, get more goodies to drool on, or break stuff. Lemme know what you think and speak up if you have any suggestions or advice.

[MMZ_TimeLord]

Good start... I like the temporary stand to test the radiator. Ingenious.

[Meethoss]

Looks cool. Slightly nuts with the ghetto Also, is it wise running tap water through it? Especially untreated!? I imagine if you ain't doing it for long then it should be ok, but you'll risk corrosion if you do it a long time.

[krazy]

I hope to have a more permentant setup going within a week or so. I just need to get this reservoir done, have the barbs show up, and get the new tower. I'm sort of waiting for another paycheck before I can order the tower

My tapwater isn't too bad for minerals and stuff, but I imagine I'll still have to clean a bit of crud out of the system when I switch over to the more permenant setup.

[krazy]

PUMP MODIFICATIONS:
I made some modifications to the Danner Mag3 pump, many of which are detailed in this thread. I learned that the danner (and maybe other pumps like it) need the flop joint between the magnet and the impeller to work reliably.

CHIPSET & VGA BLOCKS:
My chipset and northbridge blocks are identically modified oldschool Innovatek GF4 blocks, which will run off of the dual outlets of the WhiteWater and then reconverge. I am replacing the stock 8mm hose clamp things with standard 1/2" hose barbs (not the 3/8" one as shown in that picture). I was happy to discover that the weird metric thread on Innovatek's fittings was so close to 1/4" NPT thread that I could simply thread in standard fittings in place of the 8mm ones. The tpi is slightly off, but I only need two or three threads to grip to get a secure physical connection and be able to seal the threads with teflon pipe dope compound.

I was dissatisfied with the cross sectional area between the block's solid base and it's cap. I decided that a modification here was in order as well. Initially, I wanted to use a dremel to grind grooves or inverse fins into the base to provide an easier path for the coolant to travel through and increase surface area at the same time. To accomplish this, I built a jig to turn my dremel into a sort of chop saw/milling tool. The dremel would lower down on arms to grind a series of carefully-controlled notches into the copper baseplate. I made it halfway through two of these notches and gave up. The block was smoking hot, the dremel was bogging down and chattering, and the copper was not cutting nicely at all! I figured that rather than using a half dozen discs and spending a weekend trying to finish all these grooves, I should switch to plan B.

I took the parts over to my former high school and talked to my former shop teacher. He was amused by the concept of what I was working on as a whole, and agreed to let me use the shop's milling machine after classes let out. The two-fluted TIN coated end mills cut reasonably well (a heck of a lot nicer than the lousy job the dremel was doing) and in 30 minutes, I had finished modifying both baseplates. My finished product should reduce the flow restriction these blocks caused in their original form. It would have been nice if I could have found a way to make the grooves happen, but that would have taken forever the way I was doing it. These blocks only need to cool a GeForce4ti4200 and an nForce2 chipset.

[Meethoss]

Heh, looks wicked. I should get my first water cooled rig up and running in the next few weeks. Have been waiting for my mums boyfriend to make me an aluminium case for my external stuff (rad and fans, res, pump and accompanying PSU with relay attached to PC PSU) and then I'll try it out with my blocks (they could do with a bit more lapping).

Hopefully in the future we're gonna try making some of our own blocks and if that works then maybe try making some for sale. Will be fun either way

[krazy]

Well, I'm dragging this thread kicking and screaming back from the dead.

I had some cash inflow and ordered my new case. When it arrived, I promptly began measuring, photographing, and laying out holes and cuts. To see pictures of the case stripped of plastic parts and on the operating table, look here:
http://stoerzinger.dyndns.org/~karl/...1_case_before/

-----------

Front fans:
First to go were the drive cages and their mounting shelves. I laid out a rectangular cut on the front face of the tower and cut the whole thing out. Partway through, my dremel kicked and sheared the collet screw. Fortunately, I made it over to Home Depot and picked up a new one 2 minutes before they closed and was able to continue working.

Before cutting out front panel
And again afterwards
the new panel held in place with kleekos
cutting the fan holes with a fly cutter on the drill press
finished product ready to be toughed-up and riveted in

The new heater core:
The ugly heater core I got used off of ebay has three problems: for starters, it's making white buildup in my loop, and it's too big to fit nicely in the back panel of the new case, and did I mention that it's uuuuugly?

I stopped at several car parts stores around town and paged through catalogs. Eventually I found one that was almost perfect for what I wanted to do, and the place could have it sent over from their warehouse by that evening. I stopped by after work and picked up my shiny new single-pass core for $34. Perfect.

I modified the pipes coming out of it and epoxied on thread adapters. I would have soldered them, but I'm certain I would have ended up melting the joints to the tanks and made a royal mess out of the whole deal. The epoxy I used is the 8-hour cure time waterproof rated stuff. I'm pretty sure it'll do the job for a low-temp, non-pressurized system like I'll be dealing with.

face view
side view
I cut the pipes off with a dremel and a hack saw blade
wire-brushed copper ready for epoxy
epoxied-on thread adapters

Installing the heatercore
My plan was to do what Ben did with a similar setup and use the PSU mount in the back of the case to contain the heatercore. I intended originally to omit the two crossbar stabilizer pieces to get a little more room for the heatercore. I planned to use hex coupling nuts epoxied onto the corners of the core to secure it to the case.

These pictures don't show it, but I ended up changing plans slightly near the end when my epoxied mounts proved to be too weak. Two of them snapped off as I was tightening another one, and the core almost made it's way towards the floor (a quick grab saved it). Now the core is physically sitting on top of the (now slightly notched) rails and secured by the coupling nuts only at the top. When I re-epoxied the nuts, I did a much better job cleaning and roughening the metal up so I'm fairly confident they'll stay put this time. If they don't, the core can't fall down into the rest of the computer at least.

Cutting the hole was easy. The bottom was already in the perfect place, so I just had to trim the square ears out of the corners. The top was enlarved upwards by slightly over an inch, and the hole was already the perfect width from the factory.

the basic idea (with the case on it's side)
the fit at the bottom (can it get any better?)
and the fit at the top (perfect)
the factory hole
the same hole after I was finished with it

Getting power into the case
Since I'll obviously be putting the PSU on the inside of the case, I needed a way to get power in for both the PSU and the pump. My solution was the most obvious elegant solution: add a power plug to the back of the case somewhere else.

I scavenged the plug from an old power supply and cut a hole for it in the lower rear corner of the back panel. I'll run power up to the PSU and pump relay from here with a normal-type power cable wire and put shrink wrap over the soldered ends to keep from electrocuting someone if they go digging around inside my case.

marking the cuts
roughed-out with a dremel; time for hand tools
the finished product

Inital airflow improvement: eliminating that pesky shelf thing

I didn't like the tight path left for air to get from my front fans up to the heater core, so I decided to remove the shelf in the way, leaving the front portion intact to support the drive bays and whatnot. I changed my mind at the last minute (after drilling out the rivets already) and decided there was probably a reason for this part, or de designers would have omitted it. I figured all it could be doing was diagonally bracing the middle of the case, so I changed my design and left an outline of it as an angle brace. This opened up the middle for good airflow and also left the structural support in place.

shelf intact with freshly-notched side support trusses to hold the heater core
beginning to cut with the dremel
the finished product. Look better?

---------------

That's all for now. My reservoir and new cabling will arrive on the 13th, and then I will have almost everything to put the whole shebang into the new case.

Comments? Advice? Questions?

edit: fixed bad tag

[sandman]

looks nice, you got some mad dremel skills bro.