Geeks With Blogs
ICaramba Miguel Castro's blog about .NET and its effect on National Security, the Eco-system, and his daughter's sleeping patterns.

Click here for part 1, part 2, part 3, part 4

The Liquid Loop and Powering it up - yikes

So you saw the surgery I performed on the video cards to attach the cooling blocks.  I also talked about the CPU cooling block, the radiator, reservoir, and pump.  Each of these has two barbs on it, used for in and out liquid flow.  The goal here is to make a closed loop where everything gets part of the flow.  Also, do not think that by ordering the loop in a certain way you are providing more cooling in one component than in another.  The entire loop maintains the same temperature throughout.

My loop went as follows:

The pump's input (the part that sucks) pulled liquid directly from the reservoir and the output pumped it to one of the radiator's barbs.  From there I went out of the radiator (through the other barb of course - remember, everything has two) and into the CPU.  I then went out of the CPU and into the first video card, out of that card and into the second, then out of the second card and back into the reservoir.  All my equipment has 1/2" barbs so I can use either 1/2" or 7/16" tubing.  I went with 7/16" because it's a bit easier to bend and still fits relatively easy into each barb.  Some things required more of a twist than others but in any case I sealed it up with some plastic clamps afterward.

Have you noticed I keep using the word "liquid" and not "water"?  There's a reason for that.  When liquid cooling a rig, you need to use either distilled water or an actual coolant fluid.  I went with the later for two reasons.  The first is that I wanted something with color; not for show, because this case does not have any window, but so I can actually see something going on.  The other reason is that most coolants are actually non-conductive and should there ever be a leak, I felt this was safer.  This brings me to testing for leaks.

You don't want to build a rig like this and simply power it up - this thing needs to be tested.  First of all of course, you need to fill the reservoir.  Before I did this, I had to rig the power supply so I can turn it on without actually turning the machine on.  First and foremost of course, this means DO NOT plug any of the power cables to your devices - the most important one being the motherboard.  Not plugging it into the other devices is simply because it isn't necessary yet.  There are some articles out there on how to cross a couple of pins on your 24-pin connector on the PSU so you can power the PSU up but it's seemed like a pain in the butt.  The easy solution will cost you around 30 bucks but worth it because it involves buying a power supply tester, which I feel we should all have anyway.  I picked one up online and most of them have receptacles for every single type of connector, letting you test them all.  Some have lights that tell you the voltage information; mine has an LED screen.  All are about the size of small MP3 player.  In fact, I ordered one online then saw one at MicroCenter.  Since I'm so damn impatient, I bought it, used it to test my liquid loop then returned it - a few days later, the one I ordered (which was much nicer) came in.  Plugging the 24-pin connector into the power supply tester not only tests its voltage, but it also cross-wires it so your power supply can turn on and provide power to the other connectors.  I then connected a standard 6-pin connector to the pump and made sure it worked.

The next step was to start filling the reservoir.  I picked up this cool syringe that's about as thick as a can of Red-Bull.  I used some black coolant that I picked up on the Danger Den web site but later saw that MicroCenter also carries it - the bottle is about 30 bucks.  I poured some coolant into a glass so I can easily suck it into the syringe then squeezed it into the reservoir.  My reservoir like I stated earlier, mounts in a 5 1/4" bay and I left enough length in the tubing connected to it so i could slide it out through the front about 4 inches.  This was enough to unscrew the plastic filling lid on the top-front.  I started squeezing coolant into it and repeated the process a couple of times until the reservoir was filled past the barbs.  That's when the coolant started flowing down into the tubing.  The problem is that of course, gravity is not enough to get the loop filled so it was time to close the reservoir lid and turn on the power.  I heard the pump start right away and almost immediately start to suck coolant and pumping it through the loop.  As soon as the coolant level came down under the reservoir barbs I turned off the power and added more coolant to the reservoir.  I had to repeat this process about 3 or 4 times, effectively using the whole bottle of coolant but I wanted to make sure that the pump can suck coolant properly and I only saw that happening if the coolant level was above the reservoir barbs.  I also noticed that if I left the coolant level too low under the top of the barbs, you hear slushing.  Filling it up to cover the barbs eliminating the slushing sounds.

The idea is to let the coolant circulate for about 24 hours.  This is to ensure that nothing leaks.  I actually set some napkins inside the case to capture a liquid and make it more visible.  My test was successful and I had no leaks (knocking on wood three times now) !

It was now time to finish connecting the power cables and power this puppy up.  I have to tell you, not many thing make me hold my breath in anxiety; but this did.  Remember, this was my first venture into liquid cooling and I just wanted to make sure I did things right.  I checked and double-checked and triple-checked everything and finally decided I had to take the plunge.  I powered it up and...


My SuperRig powered up and I was able to enter the CMOS configuration.  I should tell you that I did have a small problem where the motherboard's internal LED display (it shows you status codes) was showing me an error status code which prevented booting.  This happened every so often and was very annoying, but a call to EVGA support taught me that sometimes they leave the CMOS test-jumper in TEST mode, which they did in mine.  Moving it over fixed the problem and I've been fine since.


So I could not have finished this project without at least one or two gripes.  I do have one small problem with the Cosmos 1000 case.  It's real nice that both sides of the case come off with literally the flip of a lever in the back - I really love that.  It's also nice that both removable sides have sound proofing on them already.  The problem is that the space between the foam on the side panel and the motherboard backboard and case framing is not too much.  This makes it a bit of a pain in the ass when you slip cabling (especially think cabling like the motherboard power cables) through the bottom slot and back into the case through a side slot.  What I ended up doing is just running the 8-pin cable through the back of the case through the bottom slot and back into the front through the top slot (the 8-pin connector on this motherboard is all the way at the top).  The 24-pin cable was simply two thick and would have to squeeze between the case framing bar and the panel with the foam on it; this had the effect of repeatedly popping out that side panel.  My solution was to run it inside the case but I did tie-wrap it nicely so it was not in the way.  As you can see in the picture though, all the other cables like the SATA power cables, the 3-pin connectors, and the 6-pin connectors all come to the back of the case through the bottom.  Then those that need to go back in to the case, like the DVD drives, do so through one of the vertical slots; while the others like the SATA cables stay back there.  Also notice that the three radiator fans use 3-pin power cables and since they're at the top, I just ran them through that top slot to the back of the case and connected them to a power cable (a couple actually) that ran through the bottom slot to the back of the case.

The other gripe I have is really my own fault.  Liquid cooling requires a little maintenance about once a year so I'm told where you need to empty the loop out and clean the cooling blocks.  This means removing them from their components and of course cleaning the chips, and reapplying thermal paste again before rebuilding them; it's the price you pay for the life you choose.  My gripe is that I designed a completely closed loop so emptying the coolant will probably be done by using the syringe to suck it out of the reservoir.  This would of course mean reconnecting the power supply tester so I can run the pump without powering up the machine, running the pump a few seconds so it can move some more coolant into the reservoir and repeating the suck process.  If I would have thought about this, I would have put a T-junction with a stopper in the base of the T, in the part of the loop where the pump pushes liquid into the radiator just after it sucked it out of the reservoir.  This would have let me open the cap on the T-connector, plug a tube into it and put the other end into a jar or something.  Then when I run the pump, it would suck coolant from the loop and push it into the T-connector and into my jar.  Of course, ideally the T-Connector should be able to close itself from pushing coolant back into the loop; I am not sure if something like that exists out there - I'll have to research it.

The Monitors

What's the use of having two powerful video cards when you're running Vista, which at this moment does not support SLI?  Well, the answer is to run multiple monitors of course.  I've been running a triple-head setup for a long time.  My last rig was a P4 and used a Matrox Parhelia card that supported three monitors.  At that time, I had 3 19" KDS LCD monitors.  These only support 1280x1024 resolution which was OK at the time but pretty lame today.  I wanted to keep a triple-head setup (at least) and since my second rig (workstation #2) had a Samsung 21T that always looked great, I picked up three more on eBay.  These were not refurbished; I bought new ones from a reseller that was getting rid of them and I paid about $250 each.  After setting them up on the my primary work area, I decided that it wasn't insane enough so I took one of them and placed it on a wall mount next to the original 21" that was connected to my second rig.  That made that machine dual monitor -- nice (in my best Borat imitation), but it left a gaping hole between my other two monitors that would be attached to the SuperRig.  I decided that this gap needed to be filled with a nice, brand new, 24" Dell wide-screen monitor capable of running 1920 resolution.  The wings were the 21"s which run at 1600 resolution.

Since most of my time on this rig is spent in Visual Studio, I've configured the IDE to take advantage of this.  I've taken every tool window I use, including the Solutions Explorer, Toolbox, Server Explorer, Property Browser, Help, and Output Window and docked them all together - with the Output window across the bottom.  I've then taken that entire set of docked windows and docked it to the right of the code window and stretched Visual Studio out to take up the center and right monitor, with the center one being all code - it is sweeeet !

Now I was happy !


I'm happy to report that daddy and SuperRig are doing well.  The machine is blazing fast and I couldn't be happier - that is, until the 8-cores come out.  I have one available monitor port left in one of the cards and heaven forbid it goes unused, so I purchased a 20" widescreen which will go above the center screen.  I also have an available PCI-E slot so I may actually pick up a cheaper 8600GTS card to connect two more 20" widescreens as top-wings.  Don't say it !  Last time I posted an entry about my old setup, which also has 3 monitors, I got emails from people actually yelling at me with negative comments.  Comments like, "why the hell do you want all that" and "you're an idiot, you could have just gotten one big screen" and "why would anyone need that many screens".  My answer is simply, "mind your own friggin business - because I want them, and believe me, I use them.  And baby, once you play Microsoft Flight Simulator on multiple screens, you're hooked !


Click here for part 1, part 2, part 3, part 4

Posted on Tuesday, October 30, 2007 3:17 AM | Back to top

Copyright © Miguel A. Castro | Powered by: