How-To build waterblocks and bong coolers — Ji Kuang
First, I’ll list out my system config that’s of interest.
- Thunderbird 1000 MHz 100fsb chip AXIA
- Abit KT7A Raid
- Voodoo 5 5500 AGP
- 2 X Twinmos PC133 128 MB RAM
- 1 X 300 watt ATX power supply
- 1 X 230 watt AT power supply
Here’s a few pics of it before I go into the details:
I started watercooling with a copper cross-drilled waterblock and then went onto a DangerDen Maze2. Both were good but I wanted to make a special
waterblock that was different from the standard cross-drilled and maze types.
First, it must have good flow rate. Second, it must have the cold water splash in at the centre core area. Third, I want to maximise the amount of cold water in contact with the copper, as heat transfers increase with the difference in temperatures. In the end, I decided on a 2 inlet 1 outlet block.
Since I wanted 2 inlets, I decided to go 3/8" for both so that there is enough pressure from my pump. Then for a good flowrate, I decided on 5/8" for the outlet. A little restriction there as 2 x 3/8" volume of water would be a little more than a 5/8", but it’s
good because it prevents the problem of water not being refreshed or not moving in certain parts of the block, due to its special design.
In the end, this is the rough sketch of its design. I actually had planned dimensions on a piece of paper which I lost. The center channel is around 1/2" while the outer ones are a little bigger than 3/8". The base plate is a little higher than 1mm.
I wanted the inlet 2 to be a little higher than centre to even the flow as it splits into 2 directions, taking into account gravity.
Next thing was to find a place to make these.
Luckily, one of my friends has a drill press in his shop, and with some milling bits, the block could be made. Also, my friend said he could help me braze the top plate on, so everything’s OK. Since I was going to make a waterblock, I decided to do it with my friend who was also in need of one. The design for his is about the same as mine, only he has only 1 inlet at the center and the fittings are 1/2".
Next was to go shopping for the stuff we needed.
We bought 2 copper blocks of appropriate sizes from a metal shop and a plate for the tops. These cost around $12 US dollars. Then we went to get the hose barbs. As the designs made it so that it would be pretty impossible to screw in the barbs that are so close together, I decided to braze the fittings. So, I bought 2 ended hose barbs and cut them into 2.
Finally, we had to buy some milling bits and so
we bought a few of appropriate sizes. We had bought bits with 2 cutting edges at first, but we found that it was easier to cut with 4 edged bits.
Then, we were ready to make the blocks…
The block making process was a rather tedious one, as we had no
experience and all we had was a drill press. Also we were
working in very cramped quarters with bad ventilation. The number one mistake we made was that we were not using any lubricant at first. Thus, drill bits and milling bits wore out very fast! That’s when we realized our mistake and added some cutting oil. Here are some pics of the process.
The block is drilled with guiding holes before milling.
Here’s me at the drill press controlling the milling with my hands.
Half done blocks – the block on the right has a hole at the center because I tried to make the base plate lower than 1mm.
The blocks before they went on a burning trip.
The one on the left is my friend’s block.
Here, my friend’s block is being brazed with its top plate.
The hose barb goes on…
2 "cooked" blocks. The black stuff is oxidized
copper. They were left to cool on their own – I didn’t pour water because they may crack.
After filing, sanding, polishing and rinsing the insides with
vinegar and water, here’s my completed block:
No leaks at all, plus because it’s brazed, it can
withstand higher PSIs than soldered blocks – not that I was ever going to push such high pressures.
Next, it was time to mount the block on my motherboard.
I used metal screws with plastic washers in the beginning, but switched to nylon ones because the 4 holes on the block were not very aligned for the block to sit properly – some bolts needed to bend a little. I bought a tough spring of a appropriate size and cut it into 4.
As seen here, the two inlets are from a “Y” splitter. I had
thought of using one pump for each inlet to increase turbulence at first but didn’t do so in the end as I felt that a powerful pump would be enough. Its made up of a few parts which was pretty hard to find. The problematic part was the threads which I needed converters to change from male to female and so on. The end product is a “Y” splitter of 5/8" in split into 2 x 3/8" out.
Now let’s go onto the other half of my cooling rig: The pump and the evaporative cooler (AKA Bong). My evaporative cooler is made up of a few parts. The main body is made up of 3 parts:
The base is an 8" PVC part that converts to 6". This is to provide a stable base and to house my submersible pump.
The next part is the top with a 5/8" barb for intake. The barb looks like aluminum, but the guy assured me that it was brass with a different color.
The last part is the white PVC part which is a 6" pipe
with a 4" “Y”. This was a secondhand one which the shopkeeper sold to me cheap. The 4" “Y” is perfect for a 120mm fan. The base cost me around $12 USD, top part and the “Y” for another $12. Total cost around $24 excluding some PVC elbows for the showerhead.
Next, the pump: It’s a submersible 3000l/hr unit Atman product. It comes with 3/4" fittings onto which I was able to squeeze a 5/8" hose. It cost me around $25 USD.
Now, we look at the reservoir I’m using:
I don’t know the actual volume it can hold, but it’s definitely big enough for me to go running for days without refilling if I wanted to. I bought it for around USD $7.
Next is the showerhead. The one I picked was around 3" in
diameter and came with a lot of holes. However, I drilled many more holes so that it had hundreds of holes.
Sounds funny, but I just couldn’t find an appropriate sized
elbow to screw it into. I had to burn the perimeter of the elbow and stuff the shower head in. I also removed the flow restrictor in it and then used silicone on the perimeter to seal it.
Next, putting them together. This shows the position of the pump in the reservoir which had a hose barb attached to it.
Next, in goes the base.
The top part with the showerhead attached.
Next, I went into refining the bong…
Since my bong cooler only has a falling height of around 77
cm, I wanted to increase the surface area of water cooled and the time it stays in the air. Thus, I decided to make add some "egg crate" kinda stuff to break up the water more.
After spending half a day with cardboard, duct tape and aluminum foil, I gave up on this prototype.
So, off I went to hardware shops to look for something that might work. Then I saw this rubber toilet flooring and it was just right. I just needed to cut them to the appropriate size.
Now I had to stack these in the pipe, so I decided to use chopsticks to support this stuff.
I drilled 4 holes for the chopsticks, 2 near the top and 2 near the bottom.
I used silicone and duct tape to seal the holes. Then I stacked 3 pieces of these on each of the 2 sets – they do not affect the airflow too much.
Now that this is done, I went onto refining my reservoir. Since I’m using a submersible pump, I have to keep the water above a certain level for the pump to be fully immersed in water. Therefore I marked out a safety level:
Next, because it takes a long time to chill more water, I decided to reduce the amount of water below the safety line by putting in stuff to take up the space. This will in no way increase the number of times I have to refill, because its all under the safety line.
Here’s what I used: Two plastic bags of stones, a hose, a Tupper Ware of stones and a goggle box. Here’s the other half of the lid which I take out for refilling.
I’ve also added a 120mm fan here – it’s a Nidec 0.2A fan, while the one on the "Y" is 0.48A. A hole which is covered with 2 layers of wire mesh is cut for evaporation, with 2 layers of wire mesh to cover the top.
Finally, a sponge was placed on the water below the pipe to curb the sound of splashing water. There’s still the sound of water running against the walls, but the rubber flooring helped to reduce it. Now it’s just a soft pond waterfall kinda’ sound.
Pics of the completed bong:
The PVC parts were not permanently joined together so that it can be disassembled. The whole thing stands at 1.25 metres. Notice that I didn’t use up the entire top part by placing the showerhead at the very top – I wanted to maintain a high flow rate.
I’m very satisfied with its performance. With the CPU at load,
the water can get 2-3C below ambient if left running for some time. Also, since the wire mesh catches a lot of sprites, it uses up around 1.5 litres of water after running for 10 hours, an OK amount considering the cooling going on. I don’t usually fill my reservoir with too much water so that the water gets cooled more.
Next – overclocking and temperatures. The system running with my homemade cooler, Thunderbird 1000 MHz at 1609 MHz at 2.33v stable, full load of
~40C, 26C ambient.
It idles around ambient at 1576 MHz, 2.2v
Very good temps and overclock. I have tested it with a Maze2 as well.
It has 1/2" fittings, an older version of Maze2 with the exact same
rig except 2 5/8"-1/2" converters gave me 2-3C higher temps at load and was only stable at 1554 MHz. Nice.
Well, I guess it was worth all the effort put into my cooling rig, as it beats commercial watercoolers at lower prices. Plus, it was a fun and a learning process. I think the reason why my block performed so well is because, first, it had a high flow rate.
This is due to the fact that it has hardly any turns at all. Also, it has deep and pretty wide channels. The flow rate through the center channel is very fast – in and out quickly – while the outer channels are slower because of the bigger area it has to travel. This is fine because the center part is most important.
Second, the 2 input chambers certainly increase the turbulence in it, plus the increased amount of cold water. Third, because it was milled free-hand layer by layer, it has slightly uneven walls that increase surface area and turbulence.
Lastly it has a very thin base of around 1mm.