Search this site
| Water Cooling Build - June 2008 |
 |
 |
 |
| Written by huddy |
| Monday, 16 June 2008 16:11 |
|
Introduction... It's been quite apparent recently of the restrictions of cooling an overclocked Intel Q6600 on air. Things have got rather toasty inside my PC and the fans have been getting progressively louder. At 3.2GHz running at around 60c on full load, the Artic Freezer 7 is struggling a bit so it’s time to look at an alternative. More fans would be a simpler and more obvious choice but I'm not one for simplicity. Water cooling has been on my mind for awhile but is something I’ve always regarded as a dark area. However, knowing that I can get more from this CPU it would seem to be a natural progression for a PC enthusiasts. So it’s time grab the torch and shine some light on to the subject.
Before I started..
Before I made any decisions, I spent plenty of time browsing forums and the internet for guides to make sure I fully understood fully what was ahead, making sure I know of all the pitfalls and taking notes from those how have achieved successful water cooling implementations. I'm not one who likes surprises and I believe a little time spent on research saves time and heartache later. This was made a lot easier thanks to Custom PC, as they had published a full guide to Water cooling around this time which was really helpful.
As I have already decided that if I’m going to water cool my system then I might as well jump in head first. I’ve decided on a full water cooling loop to cool the CPU, GPU and possibly the chipset. For this I’ll need a decent pump and a triple radiator for optimum cooling. Obviously, all this will need to go somewhere so my first port of call before even considering any water cooling equipment was the case. The kandalf is a full tower with plenty of space inside but is badly designed for internal water cooling.
The components
After plenty of research and choosing a decent water cooling case, the choice of components came to me pretty easy. I've selected a custom choice of components using 1/2" barbs as follows:
Radiator
Slim built and highly recommended, will be ideal in the roof of the Cosmos. Expensive but push a lot of air (68.5cfm @ 24dB(A))
Reservoir/Pump This is a no brainer as far as I was concerned. Large liquid capacity reservoir with built in 12v pump. Saves having to install a separate pump on the loop and is the same price of a normal reservoir. I've not read too much about it so we'll see.
CPU Waterblock Probably the most difficult choice because of mixed reviews and availability. I had original ordered a Swiftech Apogee but it was on back order for ages and available nowhere. The edge was brought to my attention and I was convinced but again, availability was the problem. Very frustrating when all the other components are in the workshop waiting.
The acrylic CPU block of the XSPC edge shows the manufactured copper fins inside, making this a superb choice.
There aren't too many GTX waterblocks on the market, so the juice is limited and this beast came up on OCUK members market at half the cost. Tubing by Masterkleer 7/16" ID tubing is slightly smaller than 1/2" ID. This meant that they will fit very tightly on the 1/2" barbs reducing the risk of any leaks. Learning tips like this is where the research pays off.
Fan Controller With so many fans in the case, I thought it would be good to take some control. The Zalman is a cheap basic solution.
Keeping control, the Zalman is a bargain for under £20..
Case *Manufacturer : CoolerMaster
The Cosmos was chosen for it's well designed interior and it's capability of hosting a fully water cooled system withiout any modification.
The Cosmos "S" looks simply stunning and oozes quality throughout..
With all components now together in my workshop, my attention now starts to turn on the build itself...
The build...
My first job was to remove all the case panels for easier access to the internal cages and for their protection. The roof is taken off revealing to three 120mm fan mounts. With the case empty, It's a good opportunity to inspect and get familiar with fittings and to size up where every thing is going.
The first component to be fitted was the radiators and fans. I started by attaching the three Scythe fans on to the radiator ensuring the air flows over the fins, taking care not to damage any chambers underneath. As an extra precaution, I've used some rubber washers which will stop the screws going in too far and to stop any possible vibrations. The fan cables are located on the same side what will be the back once fitted, so that they can be hidden out of sight.
The ½” barbs where then fitted to the ports. I’ve applied some plumbers tape on the G1/4" thread for a guaranteed seal. Each barb is tightened by hand then using a spanner making sure not to over tighten.
As well as the "O" ring, a thin layer of plumbers tape on the barb guarantees a water sight tight seal.
The radiator is then secured to the roof using the predrilled holes and normal grub screws. I've positioned the radiator so the hot air is extracted out of the roof and the inlet/outlet barbs are towards the rear of the case. The triple radiator with the fans attached occupies half on the top two drive bays so one of these is an ideal place for the Zalman fan controller. I’ve fitted the fan controller in the 2nd drive bay down so I can get to the back and connect the fans.
The bay reservoir takes up two bays which needed a minor alteration to fit.
I've treated myself to new DVD drive since the one in my old machine is beige. I've put this just under the reservoir.
That's if for the time being. The next steps is to start dismantling the old machine and install the components. I've left this to the morning.
Next morning (that was quick), as my half eaten bacon sarnie and cuppa go cold, I start to break down my current PC, boxing up everything I don’t need as I go along. The power supply is removed first and transferred to the new case followed by the hard disk drives. The RC1100 has a drive bay located at the bottom of the case which is behind a 120mm fan (ass seen above). The drive installation is easier than what I had read, it just needed a little force go get the cage out from the rack as there are four large blue anti-vibration rubber bolts keeping the drive secure. You can just make them out in the picture above.
The Artic Cooling Freezer 7 Pro won't be needed so it's removed cleaned up and boxed ready for resale. The removal of the HSF has exposed the CPU and reveals 8 months of heat displacement. Close inspection gives me some idea of the contact being made. Note the lack of contact on the bottom right.. I'm going to lag this later...
The close up of the CPU shows that not all of the surface has been making contact with the HSF.
Time to remove the motherboard. There's no need to remove the RAM and I'm going to leave the CPU in the socket for the time being just for it's own protection. This can be cleaned up when I'm ready to install the CPU waterblock.
Whilst the board is out of the case, I've just given it a quick clean up. The picture below shows the new waterblock and the fitting screws.
With no hanging around, I've installed the CPU waterblock mounting screws and fitted the motherboard into the new case, making sure that that the I/O panel is in the correct place. I've also run the 4pin ATX cable (out of sight) and the fan cables behind the motherboard just for tidiness. This in hindsight was more undoing, which I'll explain later. The picture below shows the mounting screws which protrude from the motherboard. You can appreciate the size of the radiator in the roof and the reservoir to the right.
The motherboard is place, The picture clearly shows the size of the radiator fans.
With the motherboard in place, time to clean up and lag that CPU. My camera ran out of batteries at this point so excuse the lack of pictures. I'm going to cover lagging in a separate section later. I've placed Artic Silver 5 like I normally do and re-seated the CPU back into the socket.
I then gently lowered the the waterblock into place, making sure it sits securely on the CPU. The tension springs and tighten screws are fitted onto on each mounting screw which are tightened in diagonally opposite to each other. Each screws is tightened two whole turns in turn as this ensures the pressure the surface is as even as it can be. Now this is where problems started. What I thought might be an easy exercise turned out to be a bit of a pain. As the screws started tightening, the mounting screws on the motherboard lost their grip. There was no way I could re-tighten them so I had to remove the whole motherboard, remove the waterblock and tighten the mounting screws using a spanner and screw driver. I decided this time to attached the waterblock outside the case so I can place a screw driver on side whilst I tighten the grip screws. Damn annoying but it had to be done.
My next port of call is the Graphics card. I started removing all the screws on the back and gently removed the massive cooler. The thermal pads used on the RAM chips had really bonded so a little force was needed. With the cooler removed, the naked 8800GTX card revealed the mighty G80 chip.
The 8800GTX is striped down showing the massive G80 core, RAM and mosfit chips, next to the waterblock.
To make sure that the GPU, RAM and mosfit chips make contact with the waterblock surface, I placed a small amount of thermal paste on each of the chips, then lowered the waterblock into position gently applying some finger pressure. The squashed paste after removing the block afterwards reveals that contact is being made. Which is good news. I then reapplied the paste on the GPU and RAM chips making sure there is just enough to cover the surface. I've cut up some sticky thermal pads for the mosfit chips as these are too small for the paste. The waterblock is placed back into position and each of the retaining screws is gently tightened in turn. There are no tension springs here so I make sure that the card is secure without over tightening and bending the board.
Once all the screws are in place, I examined the chips from the side using a torch. The Mosfit and RAM chips are clearly visible.
I'm satisfied that the block is correctly fitted and is secure, the the card is fitted into the PCI-e slot.
With both the CPU and GPU installed, it time to start connecting the tubing. I will need four tubes to be fitted in the following order:
Res/pump to Radiator Radiator to CPU Block CPU Block to GPU block GPU block to Res/Pump
I mentioned earlier that the tubing I'm to use is 7/16" ID which is slightly smaller than the 1/2" barbs. This ensures a tighter connection but conversely difficult to fit. Therefore, as each piece is connected, the tube end is submersed into warm water(not boiling) to expand for about a minute. After quickly drying, the tube slips on easily to the barb, which is left to cool before moving on.
Attaching the tubing. Notice the slack given to the reservoir tubing.
There are two things you'll notice from the picture above; firstly, the slack given to the tube from the reservoir to the radiator. The system fill/bleed cap is on the top of the bay reservoir which needs to be pulled out from the front when filling or for bleeding. The slack makes it easier to pull the drive bay reservoir out without the risk of accidently disconnecting any of the tubing. I've also had to connect this to the top port so as not to crunch up the tubing against the graphics card. Instead it sits on top leaving enough room for movement.
The second point is the loop from the radiator to the CPU block. I have to use the bottom port (see above) but the problem is that the port is just slightly lower than the motherboard and a direct connection would create a kink in the tube. So I've created the loop for unrestricted flow. The distance between the two is shown in the picture below.
The radiator port is very close to the CPU block. Notice too the cable ties for extra grip.
The remaining tubes are connected. Time for a quick tea break then we'll start to fill and test.
Filling and testing...
Before I start to fill the water cooling loop, you'll notice I've not connected any of the components. There are two reason for this:
Firstly, as you are well aware, water and electrics don't get on too well. So to protect the system from possible damage and carnage, the water cooling loop will be tested in isolation from the rest of the system. This way I can test for leaks without having to power on the system. Any leaks shouldn't cause any damage providing the components are let to dry afterwards. Let's hope this doesn't happen and I'm not tempting fate. I'm going to use a separate PSU for the test so although I've connected the main 24 and 8 pin ATX connectors, the PSU won't be plugged in.
Secondly, should disaster happen then there is less to undo is I need to strip everything out again. I'm only going to connect everything once I'm 100% happy that the loop is working and there are no leaks.
Ok after a first check, then double check and triple check to ensure all is in it's place and the tubing is secure, I'm happy to start filling the reservoir with liquid coolant.
You should always use distilled water (from Halfords) in your water cooling loop to prevent the system from furring which can lead to all sorts of problems. You know what your kettle looks like after a few brews. However, if you are using a mixture of different metals in water, then you need to be aware of one problem. The battery affect is where two different metals in the same water react to each other. One metal seeps energy from the other causing it to erode. If you think you have two different metals in your loop, then you need to make sure you deionised water to prevent the battery effect. In fact use deionised water anyway. Again available from Halfords. You can mix additives to make sure the water isn't as conductive which will help if the system leaks. However, the main advantage about using an additive is not only does it inhibit the conductivity but it's highly visible and looks cool too (excuse the pun).
That all said, I just bought some ready made coolant called Fesser one as I'm being lazy.
As the water is added gradually and slowly the tubes begin to fill. This is a great feeling. F
The cooling loop gradually fills as the coolant is added
The system is now starting to fill nicely. However, the water needs to be pumped around the system to ensure that no pockets of air are left. This means switching on the 12v pump but as I said earlier, I don't want to switch the PC on so it's time to "Hotwire" the PSU.
Click here on instructions to "Hotwire" a PSU.
I have a spare PSU floating around for this but if you don't have a spare PSU then use the one your are going to use in the system but make sure you disconnect all power to the motherboard and to all device.
Once the PSU was Hotwired, I connected only the pump to the PSU and switched on for just a few seconds. Just enough for the pump to fill the tubes. This is the first time the loop is in action so it's a good first test that everything is ok.
The system is connected to a test "Hotwired" PSU. You can see the reservoir pulled out for filling and bleeding
The tubing quickly fills leaving more room in the reservoir or more liquid. I've filled the reservoir nearly to the top, leaving about 10mm for expansion.
About a litre of Fessor one was needed to fill the system.
I've repeated the "on-off" process a couple of times to make sure the system is full. I'm pretty confident that it is and there are no obvious leaks. However, if you look at the system closely, you can just spot a few air bubbles in the tubing. So before I start the over night test I leave the pump running for about 30 minutes to push the air into the reservoir. I then switch the pump off and tilt the system 45 degrees so that the reservoir is at the top. The bleed key is turned to expel trapped air.
A white blanket is a good indication to spot leaks
I'm now going to leave the pump running for the next 24 hours. This is to check for "not so obvious" leaks. I have placed a white blanket directly under the water loop (as above), making sure it covers any potential leaks. This is an excellent indicator as any leak, no matter how small, will stain the blanket. The advantage of using an additive.
So that's it.. Nothing more to do now. I've now just got to leave the system be without the temptation of popping in every five minutes.
28 hours later the blanket is as clean Bishops reputation. Horra..
It's time now to disconnect the test PSU and start connecting everything up as it should be. The reservoir is pushed back into position and secured into place.
With everything connected, the PC is started up for the first time with water cooling and to my relief, the windows start-up screen is displayed.
All fired up, the system uses water for the first time and Vista starts!
Before I place the case covers, I make sure the system is stable by running Orthas Prime for a few hours and Everest is used to keep an eye on system temperatures both idle and load:
Q6600 @ 3.2GHz (9x356MHz, 1.4v) 8800GTX @ 621 MHz
CPU = 33c Idle and 44c load GPU = 31c idle and 34c load
I'm very pleased with these results. Some good potential for increased overcooking.
The case is tidied and the covers are re-attached. The PC then leaves the workshop and is set-up in our playroom.
The bay reservoir sets the the system off a treat. Notice the liquid is higher on one side as the pump works.
The red lights of the fans and RAM look great against the illuminated blue coolant in the bay.
Conclusion...
Although a very expensive and time consuming project, I think the results are worthwhile. Not Only have I picked up some valuable knowledge along the way but I now have a near silent high machine quite capable of running some high clock speeds which are beyond the scope of air cooling.
My only concern now is future upgrades. With the new graphics cards already available, it isn't going to be to longs before I'm tempted. This won't be a case of just slotting in a new card as the whole system will have to be drained, tubes removed, waterblock fitted, system refilled, tested and so on. I'm sure it won't be a problem but the task will be a major project itself.
My preparation paid dividends and I'm not sure if I would have done anything different. Maybe if I was being really ambitious I could have installed a double radiator at the bottom of the case to cool the fluid between CPU and GPU blocks and maybe the Northbridge too.
Anyway, all that remains now is to get stuck in to some serious overclocking and keep a bucket handy.
|
| Last Updated on Tuesday, 18 November 2008 14:12 |
