User Password

boakes.org ArticlesForumDownload AboutContact

tags:

Water Cooling Installation

March 30th, 2005, by Rich.

Water Cooling Installation

The water cooling kit arrived at 17:15 yesterday (that’s still officially “next day delivery” because of the four day Easter bank holiday weekend, though it’s really exactly five days since I ordered the kit).

The box arrived and the first thing I noticed was that the packing was a glossy looking corrugated plastic; probably really bad for the environment and entirely pointless given that i bought the goods over the web.

So, into the box and “oh my” what a lot of bits, each with it’s own instruction booklet.

A a water pump, a small reservoir, a radiator, a big fan, a couple of 608 socket blocks, lots of small bolts, washers, springs and thumbscrews, and the extra VGA socket block that I’d ordered, which came completely unassembled.

Water PumpEach piece, as I mentioned, had it’s own instructions; there was no overall “here’s how it alll fits together” book, just lots of fold out instructions, similar in manageability to Ordnance Survey maps.

The bits do look cool though, like something out of the engine-room of the starship Enterprise.

ReservoirSo, I begin by assembling the pump and reservoir. Easy. That just involved cutting a small length of hose, and pressingthe bits together.

Next, joining the fan to the reservoir is easy; they just screw together.

The next thing I decide to look at is the CPU’s. Somehow I have to remove the existing monster fan+heatsinks. The fans unplug, unscrew and decouple easily. Then, very carefully with a screwdriver, I use every newton of force I can muster in order to prize the springy metal clips from their seat, without slipping and gouging thgrough the motherboard’s tracks.

One, of the four clips, pops off. The others won’t budge. I look to the new kit for clues to see if I’m going about this the right way; what they tell me is that I should be looking for 4 screws that go through the motherboard, and that the clips may be a booby trap.

Yes, a booby trap. Right now I feel like I’m defusing a bomb, so taking a leaf out of the Juggernaut school of bomb tinkering, I decide there must be a backdoor to the motherboard and try opening the other side of the case. Much to my chagrin there’s a large immovable metal plate that means I can’t get to the rear of the motherboard, so I’m left no option but to disconnect every internal wire and remove the motherboard from the case. Not something I’ve ever done before, but I’ve come too far to turn back now.

With the motherboard out, I’m able to remove the springy metal clips with ease, and the heatsinks come away without too much of a struggle. What’s left is four fixing brackets that held the springy clips were attached to – and these go through the motherboard and are affixed with plastic rivets.

Using a very, very, sharp knife and a small drill with a wood cutting drill bit, it took my about 30 minutes to remove these 8 rivets.

With the rivets gone it took me no time to screw in the lovely new CPU water blocks. At last i feels like I’m getting somewhere. I reinsert the motherboard and screw it back in, leaving the cables everything disconnected for now.

GPU Water BlockThe next task is to assemble, then fit, the VGA block. Quite why it comes in pieces, when the CPU blcoks come ready made is beyond me, but it does; anyway, it didn’t take long to combine the pieces. The tough it was removing the apparently glued-on heatsink that ATI had thoughtfully added to their board (a Radeon 9700 pro, incidentally). The fan came off without argument, but I thought I might have broken the card because I had to apply so much force in removing the finned heat distributor.

RadiatorThe next task is fitting the fan and radiator to the case. A tough task, but after the CPU’s I believe I can conquer anything. Even though I have an absolutely huge case, the location of the 120mm fan fixings is too close to the edge, this means that I have to file the screw holes before I’m able to fix the fan in place.

So, with the pump, reservoir, CPU blocks, Graphics card block radiator and fan all fitted, I cut and fitted the pipes to length, in a loop from pump, to radiator, to CPU, to CPU, to graphics card, to pump. Then, with all the offcuts and detritus cleared away, I connected the power to the pump, and connected the pump’s electrics to the computer.

In order to test the pump without the risk of destroying the motherboard, a small jumper lead is supplied so that the pump can be activated with the computer’s on switch, whilst the motherboard is still disconnected. This is a very, very, good idea; and it revealed that I had a small leak in the graphics card block that needed to be disconnected, and rebuilt, before it would run without seeping.

Curiously, because I’ve removed a total of 5 fans and 2 huge heatsinks the case actually appears emptier now that the water pump and tubes are installed.

Water Cooled Dual Xeon

Having reconnected all the cables on the motherboard I gave the machine a try; it responded with a long sad beeeeeeeeeeeeeeeep. It’s not happy. I checked the cables a second time and mercifully, I’d missed one. I reboot again and this time I get brisk happy beep. The machine is booting up, and hey, that’s odd, it doesn’t sound like a Harrier Jump Jet is hovering overhead.


For the record, the base macine is a two year old Scan 3XS-HD dual 2.8GHz Xeon machine; it’s spec’d as a server, so it’s powerful, but excessively noisy, with lots of redundant fans. Before the installation it had a 120mm fan and a 90mm fan extracting air from the machine, 2 Sanyo Sancooler 60mm fans (one for each CPU) and a 50mm (approx.) ATI Fan.

Using SuperoDoctor I monitored the CPU temperatures which, when averaged were CPU1:39c and CPU2:33c. Currently whilst 80% through a 2 hour complete virus scan, both CPU temperatures are 38c with only the Power Supply and and radiator fans to make any noise – the radiator fan, indidentally, is set on the 7v option (of 7v/12v).

My initial interpretation is that I can run the computer at the same, or slightly lower temperature, with a lot less noise.

Once I’ve given it a few days run in to ensure the temperatures don’t slowly creep up, I’ll try a fw variables to see if things can be improved.

3 Comments

Leave a Reply