OK, it has been a while since I posted. I keep meaning to post about my watercooling work, but can never find the time because of classes, work, etc. I’ve finished the watercooling setup, so here’s the full story.
The first thing I did was to measure and cut tubing. Since the computer was together and air cooled, I could fairly accurately figure out how long to make each piece of tube. The plan was to go pump -> CPU -> GPU -> rad -> pump. I decided to put the T on the outside of the case, just before the tube went back into the computer. I measured each run and cut tubing to that length. Here are the tubes:
Next, I attached barbs to the radiator and the EK GPU water block. The CPU block already had barbs. I also took the opportunity to swap out the stock nozzle in the CPU block for the quad core nozzle.
I happen to have a power supply tester in my bag of computer fixing tools, so I figured that was a good way to run the pump when I was ready for that. I also have an old 200W power supply, so I hooked that up and made sure it worked, so that I could run the water cooling setup outside of my case for a while first.
Figuring out how to mount the radiator turned into a bit of a challenge, as I didn’t know all of the properties of the radiator. For example, I couldn’t get a for sure answer whether I could mount it upside-down, so I decided to mount it right side up anyway (barbs at the top). That way, at least air bubbles wouldn’t get stuck at the far end of the radiator. With the Sonata II, I had to mount the radiator so that the middle of the radiator was connected to the radbox, so the top of the radiator sticks up past the top of the case. The only way I could find to route the tubes was through what used to be the air intake for the CPU air duct. I cut out some of the honeycomb using a wire cutter. I’ve got a couple of pics of this later. In order to keep the tubing as straight as possible, I mounted the radiator with the barbs at the top, facing away from the case. That way, the tubes could come out of the air duct hole, around the radiator, and up to the barbs.
Getting the tubing attached to each of the components was fairly tough. The D-Tek barbs weren’t too bad, but the Swiftech pump barbs were tough. Someone recommended boiling some water and dipping the tubing into the water before trying to attach it to the barb. This worked well. I actually forgot to take pictures before I put the fluid in, so here are some pictures after draining the fluid. Pretend there isn’t any fluid in the tubes in these pics!
I then filled the system with fluid to start my leak check. Bleeding the air bubbles even when I could manipulate every part of the system independently was troublesome, as the EK block really didn’t want to bleed its bubbles, so I ended up having to run the pump and let the T line catch the bubbles as they went by.
I then prepared the loop to be installed in my case. First, I attached the video card to its water block, as I figured it would be easier to get it attached now. I left the tubing all connected, so that the chance of it developing a leak would be lower.
Next, I mounted the water cooling into the computer. Unfortunately, as some parts are inside and some are outside, I had to detach some of the tubing from its barbs. This proved to be more difficult than getting the tubing onto the barbs in the first place, and I ended up mangling the tubing badly enough that I had to cut off the part that was pushed over the barb. When I did the initial cuts, I made sure to allow a bit extra for that sort of thing, so I still had enough length to reach the barbs once I cut off the inch or so of mangled tubing. Here are some pictures of the watercooling system in the case, before I powered it on.
I then ran the watercooling inside of the case for a couple of hours before powering on to make sure that there were still no leaks after moving everything around a whole bunch. To do that, I used another computer I had spare to power the pump only. I could have used the power supply tester, but it turns out that the tester likes to beep continually to let me know that “the power is on and working!” even though I was planning to leave it running overnight. So, setting up another box accomplished the same purpose. This also served as a way to bleed the bubbles from the system. By the time I woke up, there were no significant bubbles left in the system.
Then, it was time to turn the system on with the watercooling installed for the first time. After a bit of hesitation, I hit the switch and fired it up. It booted perfectly, and was really quiet! Once I got into Windows, I ran a whole bunch of tests to find the optimal fan speed and pump speed. Right now I’ve only tested fan push configuration... I haven’t gotten around to switching the fans around to do a pull configuration. As the best compromise between noise and temperatures, I chose running the pump at speed 2 and the fans on speed low (with the ULNA attached). At full load (running both 4 threads of Prime95 and 3DMark 2006’s vertex shader run on loop, I had load temps of 47 for the CPU, and 45 for the CPU. That’s down from 60 and 75 respectively, which is a huge improvement. Plus, it’s way quieter. Idle temperatures are 33 for the CPU and 38 for the GPU.
Here are some more eye candy pictures of the finished system! I haven’t done thorough cable management yet, so just concentrate on the watercooling part.