DWPG.Com - Project Stealth Cooling
By: Sverre Sjøthun

Introduction

stealth
In this article I’ll outline my quest for PC silence. I’m not 100% there yet but many people are requesting details now so I quickly put this "report" together. I’ll try to keep it interesting & informative as I can so expect many pictures because I’m not really a writer. A way to have PC silence has been floating around in my head for many years mainly since I caught the overclocking bug.

Fans really do get on my nerves so I wanted a way to get rid of the fans, or at least the noisy ones, to start with. I considered watercooling with an aircooled radiator, but this just swaps the noisy heatsink fans to almost equally noisy radiator fans, not really helping the noise situation much. I figured that a CPU gives out a relatively small amount of heat in the scheme of things so surely if you have a very large amount of water and remove the rad & fans, it will not be able to affect the temp of it to any great degree. I really wanted cooling that was consistent, lower and unaffected by the ambient air temperature changes. My goal also changed into having Zero fans in the complete system. This is when it occurred to me underground water temps are around 8°C to 10°C where I live all year round. Below ground about 6 to 8 foot down the temp remains pretty stable at the yearly average temp.

To set something up quickly I placed a stainless 120-litre tank under the mobile home, (I’m building a house), coolant to be circulated by an Eheim 1250 pump. This is of course still subject to the ambient temp but it did prove the CPU couldn’t heat up a large volume of water to any noticeable degree. Using this tank I watercooled the CPU with an OverClockers Hideout copper water block and hold down device. This is a very well Cnc’d solid copper block with a copper lid that is attached with screws and sealed with some hard black “stuff”.

Fig. 1 Copper Waterblock

Fig. 1 Copper Waterblock


After taking it apart, (something I can never resist), I found in measuring with digital callipers the base part was 4.6mm thick. Not happy with this much copper separating the CPU from the coolant I had the base machined down carefully so it was just under 1mm thin. It then required some minor lapping due to it denting in the middle a bit during the machining. This fairly simple mod has given around a 6°C lower max CPU temp over all.


Planning and designing

Next item to watercool was the Video card, a fairly new Asus v8200 Geforce3 Deluxe. I previously had an Asus V7700 GTS that has served me well and I love the Asus hardware-monitoring App that gives GPU and Ram temps on the GF3, AGP & card voltage, fan RPM check and failure alert, and a type of “Rain” idle software cooling for the GPU. I also chose the Asus Geforce 3 card because it was the only one to offer easily removable Pinned Ram heatsinks. Asus has broken away from the Reference Geforce3 PCB layout set down by nVidia, and in doing so moved the ram closer to the GPU.

Maybe Asus thought it would overclock better and more stable if the Ram were nearer the GPU. You've got to admire that kind of logic, but unfortunately I think Ram & GPU batch quality makes more difference. You only need one ram chip on the card that's below par and it will limit the o/c potential. The ram is only nearer by a small amount (a few millimetres), but more importantly for me, the sink pins require holes in the card which is perfect for attaching a RAM waterblock. I thought long and hard about ways to do it and decided it would be best if it were a one-piece unit covering the entire ram so using all the four retaining holes. Firstly I removed the Ram and GPU sinks to take accurate measurements.

Fig. 2 Original heatsink/fan

Fig. 2 Original heatsink/fan


Fig. 3 Original heatsinks disassembled

Fig. 3 Original heatsinks disassembled


Fig. 4 Ramsink design

Fig. 4 Ramsink design


This was then converted into a proposed design done in MS paint. I know there are quicker and better graphic apps but I like the simplicity of paint myself, plus it’s part of windows.


Waterblocks for the videocard

I did not have any sophisticated equipment to make this item with, just a 40 year old mini lathe, (with no motor), as an X – Y table, and cheap and nasty Wickes pillar drill with the worst chuck I’ve ever come across. At first I attempted to try slot cutting but the equipment was just not up to it so it had to be drilled down, adjoining in a line.

Fig. 5 Drill and milling tool

Fig. 5 Drill and milling tool


Fig. 6 Metals

Fig. 6 Metals


Fig. 7 Copper base

Fig. 7 Copper base

Fig. 8 Final cut

Fig. 8 Final cut


Fig. 9 Finished block, top view

Fig. 9 Finished block, top view


Fig. 10 Finished block, bottom view

Fig. 10 Finished block, bottom view


Fig. 11 Finished block, disassembled

Fig. 11 Finished block, disassembled


All final finishing was done with basic tools, time, and elbow grease:-)

I started by cutting out the block that leaves you with a rough edge.

1. I used a basic flat metal file to get the hacksawed edge even.

2. I then used Wet & Dry (wet), starting at about 250 grade and going down to 1200 Grade with various grades in between as required. I do this with the Wet & Dry sheet on a flat surface (glass is good), and move the block on the Wet & Dry, similar to lapping.

3. When it is smooth with the final wet & dry grade, I use T-Cut colour restorer,
(designed for use on car paint that has gone dull or tarnished). When this is done I finish with car polish, (Auto Glym Silicon Resin polish), for that deep shine. Both of these are also done with a cloth flat on the glass and moving the block back and fourth.

4. Once I'm happy with the finish I use white spirit or brake cleaner to remove the polish residue so it can be sprayed with clear cellulose lacquer to prevent the shiny metal tarnishing, You need to build it up in thin coats or it will "ball", "fish eye", or "run" on shiny metal.


Waterblocks for the videocard continued

While doing this I had ordered both VGA GPU & Mobo Chipset blocks from www.dangerden.com.

The only difference is in pipe inlet/outlet placement:

Fig. 12 GPU block

Fig. 12 GPU block


Fig. 13 Chipset block

Fig. 13 Chipset block


Due to card components & the RAM block being in the way, it was easier to make the mobo chipset block fit the card. This I did by using some 90° elbows bought locally and making my own screw in extensions from brass welding bottle barbs. This was necessary to be able to screw the elbows past each other.

Fig. 14 Fittings

Fig. 14 Fittings


These blocks come with no form of mounting so I decided to make something simple and secure using the four holes around the GPU on GF3 cards. I firstly inset the base a little in the corners where the feet were to go. I realised It would be hard to solder them on individually as I had to heat the whole block up for the solder to flow. I got around this by soldering the feet on as one piece and then cutting out the sections in-between and shaping to finish. I then drilled the holes and tapped threads into them with an M2 tap. Finally it was retained by steal screws using nylon washers to avoid any PCB damage. This has worked fine and the block is firmly and neatly mounted without any nasty clamps or brackets. It would have been nice to just drill and tap the corners of the block, but it was a few mm to small to be able to.

Here are some pics that explain how I did it:

Fig. 15 Cutting the base plate

Fig. 15 Cutting the base plate


Fig. 16 Filing the base plate

Fig. 16 Filing the base plate


Fig. 17 Soldering

Fig. 17 Soldering


Fig. 18 GPU tabs

Fig. 18 GPU tabs


Fig. 19 Finished GPU block

Fig. 19 Finished GPU block


Waterblocks mounted

These are the finished waterblocks installed on the card and also showing the fixing method. RAM block is attached with Nylon screws with nylon washers, the GPU the same but with steel screws simply because they don’t make nylon types small enough. Artic Silver thermal paste used on the GPU block, normal non electro-conductive thermal paste used on the ram chips. The card is as you would expect heavier than standard but not excessively so. It isn’t a problem in the slot and the “L” shaped Ram block helps “stiffens” the card a bit also.

Fig. 20 Watercooled Asus v8200

Fig. 20 Watercooled Asus v8200


Fig. 21 Retaining the waterblocks

Fig. 21 Retaining the waterblocks



Results so far on coolant @ 14°C 260/560 without the voltage mod and at these temps taken after a few hours running the Aquarium D3D screen saver.

Fig. 22 Air cooling @ 230/530

Fig. 22 Air cooling @ 230/530


Fig. 23 Watercooling @ 260/560

Fig. 23 Watercooling @ 260/560


The PulseGen

Now onto the Pulse Gen. This is just a name I made up because it is basically what it does. It isn't a part of the watercooling as such, but required so that I can use the Asus Smart Doctor hardware monitoring that comes with the Asus GF3 card. It monitors fan rotation, temps, voltages and has GPU software cooling. Because I'm after a zero fan PC the card fan had to go which makes the Smart Doctor unusable due to it popping up continuously saying the fan is dead. I made this simple circuit, (with much help from others in this forum), that gives the card the same information via an artificially generated "pulse" as a fan does running at about 4500RPM.

Fig. 24 Disassembled pulse generator

Fig. 24 Disassembled pulse generator

Fig. 25 No fan connected, using pulse generator

Fig. 25 No fan connected, using pulse generator


Fig. 26 No fan or pulse generator connected

Fig. 26 No fan or pulse generator connected


The plan is to later run the signal or power for the pulse gen through a flow switch so it will pop up on pump death. (Thanks go to the members @ Hardware Central forums that helped me with building the pulse gen circuit).


Chipsetcooling

Next was the mobo chipset. Here I had to use the waterblock bought for the VGA and decided I wanted it removable without having to take the motherboard out once it was fixed. It did require the removal of the board to fix “nut screws” to the board (for want of a better description). These are similar to the type of thing you screw into the motherboard tray to space the mobo away from the tray. They have a screw thread and a nut on top also with a threaded hole to take the screws to hold the mobo in place. I fixed two of these in place again with nylon washers on the PCB to avoid circuit damage. I then made some “Tabs” which were soldered onto the side of the copper block, shaped and then drilled to accept screws which would screw into the “nut screws” Again pictures show this better:

Fig. 27 Chipset fastening tabs

Fig. 27 Chipset fastening tabs


Fig. 28 Parts for chipset block

Fig. 28 Parts for chipset block


Fig. 29 Soldered chipset block

Fig. 29 Soldered chipset block


Fig. 30 Finished chipset block

Fig. 30 Finished chipset block


Fig. 31 Fastening mechanism

Fig. 31 Fastening mechanism


Other peripherals

HDD and PSU watercooling is underway, I decided to have the HDD waterblocks CNC’d as there are three of them, (for a total of up to 6 HDD’s), and I just couldn’t face all that tedious drilling, (with the possibility of messing it up).

Here’s the plan and the channel design it is being CNC’d too.

Fig. 32 Harddisk waterblock layout

Fig. 32 Harddisk waterblock layout


Fig. 33 Waterblock coutouts

Fig. 33 Waterblock coutouts


Here’s a pic of the current PC set-up using some temporary manifolds I knocked up. Eventually it will all be Pneumatic fittings.

Fig. 34 All water-cooled!

Fig. 34 All water-cooled!


The bomb

Now on to the underground "bomb" tank, I already had a large quantity of 10mm Polyurethane/ Nylon tubing so I really wanted to be able to use it up. As I thought one pipe run might be too restrictive due to everything I'm water-cooling I decided to just triple up the flow pipes making three connections for input & output. One for the CPU block, one for VGA GPU, Mobo Chipset, PSU, & one for HDD's, VGA Ram block, and maybe system ram later, all supplied from the same pump. All fittings & pipe are pneumatic rated at 10bar. Here are some construction pics. Just tapped some washer spacers with 1/4 " BSP thread, welded them on to the tank where I'd pre-drilled holes, Sealed over all the welds with JB weld, painted the tank with black tar paint & then the pneumatic fittings were screwed in with Araldite on the threads to ensure leak proof seal, hopefully it will remain leak proof for a good while.

Fig. 36 The Bomb

Fig. 36 The Bomb


Fig. 37 Threading tank fittings

Fig. 37 Threading tank fittings


Fig. 38 Tank fittings welded on

Fig. 38 Tank fittings welded on


Fig. 39 Tank inlets

Fig. 39 Tank inlets


Fig. 40 Tank outlets

Fig. 40 Tank outlets


Fig. 41 Finished tank

Fig. 41 Finished tank


I had this “bomb” at the end of my drive all day just for fun and it was getting some very weird looks, but the bomb squad never turned up fortunately!

Next part was even more fun… Digging the hole!

I have an ace up my sleeve here though as I’m building a house. Enter a tatty 1972 JCB

Fig. 42 Starting to dig

Fig. 42 Starting to dig


Fig. 43 The abyss

Fig. 43 The abyss


Fig. 44 JCB arm fully extended

Fig. 44 JCB arm fully extended


Fig. 45 JCB in action

Fig. 45 JCB in action


It's hard to gauge hole depth from those 2D pics, but you can see it was full JCB arm depth. It is about 8½ feet deep. Once the tank was seated in the hole, (tank is 4 foot high), if I stood on the top of it my head would just stick out the top. I'm about 5'8"

Fig. 46 The tank is in place

Fig. 46 The tank is in place


Fig. 47 Down there with the tank

Fig. 47 Down there with the tank


Future plans and expectations

That’s about all for now as the tank has only just been filled. I'm now running my coolant out of the ground tank, (just had to get it bodged up quick to see some results for all the effort). Temps are not low yet, only 18°C, as the coolant, 100 + litres, has only just gone in so it will take a while to cool to the ground temp. I theory it should go to about 9°C eventually as that is the yearly average here. It might take a week or two though for the complete area I dug up; the hole, the tank & coolant to settle back down to the underground temps.......... or of course I could be wrong…

The Eheim 1250 is giving reasonable flow so far, but I will be using a domestic UK central heating pump on the finished build. Last pics are of the flow rate with the Ehiem, which is ok but could do with being a bit faster. Remember it is going through the manifolds and all waterblocks, plus 16 meters of nylon tubing and a few meters of plastic tubing. Outlet shown is ¼ BSP.

Fig. 48 Water outlet

Fig. 48 Water outlet


Fig. 49 More water flowing

Fig. 49 More water flowing


Fig. 50 Submerged Eheim 1250

Fig. 50 Submerged Eheim 1250


A final note. I now have only one fan in the system. Just for interest I set Windows to turn off the single HDD I have in the PC at the moment after 1 min. This it did and I then stopped the PSU fan with a pen for a few seconds. WOW!! Total silence running at 1596Mhz, 145fsb with an o/c'd GF3

And the story goes on:

Jonathan "BladeRunner" Spencer who wrote this tutorial has started his own site where he restlessly keep up his quest for a completely silent computer. I insist that you all head over to his site and have a look: ZeroFanZone. You can also direct any questions related to this tutorial to Jonathan.

Sverre Sjøthun

Copyright © 1999 - 2019, DWPG.Com. All rights reserved.