The "Ghetto Watercooling Project"

[jonasa97]

Introduction

 

This build log strictly focuses on the watercooling part and does not involve any case or cable management. If you came for beautiful fittings and colorful tubing runs, you are in the wrong thread.

Over the summer I was given the opportunity to work an internship at an undisclosed company at which I learned how to work metal with a file, turning machine, milling machine and more. Throughout the internship we were given various tasks but were also given some time to develop our own little project near the end. I decided to manufacture a CPU waterblock made of 2 plates of aluminium, similar to the one that Linus uses in "Scrapyard Wars 2, the Scrappening". I think I will mostly post pictures and write a little bit of the reasoning that went into the building and assembling of the block. The manufacture of the block took me about 14 hours excluding the planning process and another 2 afternoons were used to complete the assembly and the testing of the CPU waterblock. Unfortunately I was unable to make any pictures of the manufacturing process itself as there was a strict no camera/phone policy at the company. But enough talking, let's jump into it.

 

Planning

 

I will begin by giving you a list of things that were considered in the design of the block. In order to achieve a good cooling performance it is essential to have a large surface area in contact with the water near the part of the block that will later sit on the CPU. However, it is also important that the flow isn't too restricted by the pattern in the CPU block. More mechanical criteria are the dimensions of the block as a too large block would not fit into the socket area of the 1155 test platform that was used. The final block had dimensions of 86x86x13 mm and was fitted onto an MSI MPower Z77. Of course the material used for the block is also an important consideration. It is evident that the material of choice is copper yet such a large piece of copper was unavailable at that time such that I resorted to aluminium. As an alternative, I could have used steel, however steel has a much larger tendency to oxidize (rust) and thus produces gunk in the loop. Finally, the finishing quality and eveness of the surface in contact with the CPU is key to optimal cooling performance and a special miller was used for this task. Below is the final plan for the path that was to be milled into the metal.

 

 

Besides cooling performance standardized dimensions also had to be considered. For example in order to be able to attach the block to an LGA1155/1150/1151 socket, you need to space each adjacent hole's center exactly 75 mm from each other (not the diagonals of course). I also spent a large portion of my time researching about the G1/4 standardized thread (not a forum thread but the helical part of a screw) and found out that contrary to common belief, the G1/4 thread is no where near 1/4" in diameter. Luckily the company had G1/4 screw taps such that I was able to drill those into the block and to make it work with regular watercooling fittings. Pheww. This was necessary in order to guarantee that there would be no leak around the socket area during testing.

 

The finished block

 

Lots of shots of the final block and the self-made thumbscrews can be found below.

 

 

This is a shot of the channel where the water would later flow through. The two square things at each end are the inlet are outlet points of the waterblock.

 

 

A shot of the blocks shows the upper sides of both blocks.

 

 

A close up of the G1/4 thread

 

 

The underside of the CPU waterblock. Although it may not look even due to the reflection of light in the block; running your hand over it and the cooling performance shows otherwise.

 

 

This picture shows the underside of the upper metal plate. Into it, a path was milled in which there would later be poured hot glue in order to seal the block, although initially a more conservative solution using a rubber ring was planned.

 

 

The picture shows 4 M4 thumbscrews that I made myself using a turning machine

 

Assembly

 

The hardest part of the assembly was to achieve a waterproof fit of the two plates that make up the waterblock. Initially the plan was to quickly apply hot glue into the ridge shown in one of the images above and then quickly press the other block onto it. However, the glue would cool almost instantly making it impossible to press the blocks together. It was necessary to find a method to heat the glue once it was applied. So the block was inserted into a pot of water over a propane combustion stove until the glue melted.

 

 

The image shows the heating process of the two plates that make up the aluminium waterblock

 

It was then necessary to properly align the mounting holes as the melting would result in the two plates sliding past each other ever so slightly. In order to be certain that the block was impermeable, fittings and tubing were attached and a significant pressure was applied to the sealing by holding the outlet shut while blowing in air through the inlet. By holding the block into a bucket of water, and by observing the lack of any bubbles it was confirmed that the glue sealing worked.

 

 

The waterblock with the fittings and tubing attached in order to leak test the block

 

 

As previously mentioned the block was tested on two different afternoons. Yet the configuration of the watercooling components was different on both. No temperature testing was performed on the first set of parts, not because the temperatures were too high but because of a drive problem that could not be resolved due to a lack of time. The first configuration is quite unserious but was interesting to build, the second yielded performances that were acceptable for everyday use.

 

Both configurations were attached to/tested on an MPower Z77 with an i5 3570K at various multipliers

 

Watercooling parts 1

 

-H60 pump

-H60 radiator

-500 ml sparkling water bottle

-lots of glue gun sticks

-2 of the cheapest fittings I could find

-aluminium CPU waterblock

 

After all parts were glue gunned together the configuration was tested on its own, without a computer running. It was discovered that the flow rate of the H60 pump was so minimal that it would probably not serve much in this project which was to be expected as it was taken out of service for this project due to high spikes in temperatures.

 

 

 

Finally after all connections were confirmed to be waterproof, the waterblock was finally attached onto the CPU.

 

 

The following is a picture of the entire loop

 

 

As mentioned previously, due to a drive problem, it was not possible to test the system under load such that the only temperature information available was the BIOS temperature.

 

 

 

Watercooling parts 2

 

-H60 radiator

-Alphacool cheapest pump/reservoir EU, with 220V AC

-4 of the cheapest fittings I could find

-cheapest tubing EU 10/13 mm

-aluminium CPU waterblock

 

 

The Eheim 600 Station II 220V AC

 

 

The picture above shows the H60 radiator with the tubing removed. It turned out that the 10/13 mm tubing is a waterproof fit without any barbs.

 

 

As visible on the pictures, very long runs were used for CPU in and outlet in order to put a large distance in between the motherboard and a potential leak at the radiator or pump

 

 

 

Temperatures

 

Before we look at the temperatures I would like to ask all of you for a minute of silence for our beloved friend, H60!

 

 

Surprisingly the temperatures were not at all as bad as I had expected. The thermal paste I used was Arctic MX4. Ambient temperature was about 23 degrees Celsius during testing.

 

 

i5 3570K @ 4.0 GHz with 1.00 Vcore under prime 95 load (above)

 

 

i5 3570K @ 4.5 GHz with 1.25 Vcore under prime 95 load (above)

 

And no, the task manager is lying, 5.94 GHz were unfortunately not achieved.

 

 

Conclusion

 

The temperatures achieved were very surprising as there were so many factors against the cooling performance such as a lack of surface area on the waterblock, the wrong material, a rather thick layer of metal in between water and CPU and a very small radiator. Certainly, this configuration would have benefitted from a 240 mm radiator but unfortunately this would have meant additional costs for this project. In total about 50 EUR were spent on additional materials making this a worthy replacement of the H60.

 

Feel free comment and discuss!
 

[Guest Johnny5g]

scrapyard wars

[LNW]

Woah. This is SO COOL.

[iHardware Shelden]

I saw Ghetto and clicked follow.

[ShiftHitTheFan]

Since when does ghetto involve custom CNC milling?

[0ld_Chicken]

Since when does ghetto involve custom CNC milling?

maybe ill give one a go with a dremel

 

edit: not talking crap though. that is really cool jonasa97

[jonasa97]

Since when does ghetto involve custom CNC milling?

What may look like CNC milling is in fact me guiding the milling machine myself. Just because there is a complex pattern in the block that doesn't appear to have any defects, doesn't mean mean it was done by a computer.

 

EDIT: I don't know if you read the whole thread attentively but the "Ghetto" part is the water bottle as reservoir and the H60 pump and radiator attached with a glue gun and duct tape.

[ShiftHitTheFan]

What may look like CNC milling is in fact me guiding the milling machine myself. Just because there is a complex pattern in the block that doesn't appear to have any defects, doesn't mean mean it was done by a computer.

 

EDIT: I don't know if you read the whole thread attentively but the "Ghetto" part is the water bottle as reservoir and the H60 pump and radiator attached with a glue gun and duct tape.

 

Im huge into diy, and kinda a cheapskate, so I would have done the same thing. Now, if you're going to keep the water bottle as the reservoir, then that's kinda sketchy (ghetto if you want). Im not hating, I wish I had a mill to work with.

[Nizo_GTO]

For the first 3 pictures is was like "How is that ghetto?" and then i saw the water bottle and it got more ghetto from there. Good job amazing.