PDA

View Full Version : DIY CPU water cooler



wiatroda
17-03-2011, 07:57 PM
My recent project DIY CPU water cooler. I'm not desperate to overclock my pc, but want to make it to work quiet. First I'm going to replace CPU heatsink/fan with heat exchanger , if this is going to work, a graphic card will be next. I've managed today to cut out 2 parts of heat exchanger from 12mm aluminium plate.

3846 3847



3848 3844

job for tomorrow: to cut out cooler mount ( I don't have any alu plate apart of 12mm so I have no other choice than to use 3 mm steel plate)

3849

http://www.mycncuk.com/forums/images/misc/pencil.png

Jonathan
18-03-2011, 12:33 AM
Looks nice, will be interesting to see how well it performs. I think it would be better to have thinner channels and more of them, but I'm sure that whatever you do will be far better than most air cooling systems.

What are you doing about getting the area where the block contacts the CPU smooth? It needs to be a very flat surface to work well.

I'm thinking of making one at some point. You can simulate the heat flow very nicely in solid works without too much effort. What socket is yours?

Gary
18-03-2011, 09:26 AM
brings back memories, i used to overclock my PC's when i was younger, and had full water cooling to keep the noise down.
May be better to make it out of copper and also lap the underside to get it as flat and smooth as possible.

wiatroda
18-03-2011, 09:50 AM
I have Intel Core2 6400, 2.13 GHz LGA775 Socket. Heatsink is original Intel supplied together with CPU. It's wholly aluminium. When running with few aps on average CPU temp is about 50C. After light 400 grit wet sanding a contact surface I would risk to say is more smooth than original heatsink.


3850

wiatroda
18-03-2011, 04:02 PM
Pics below shows a heat exchanger mount cut out from 3mm stell.

3854 3853
There is some more pictures & details of this project at my website:
http://malamuterun.co.uk/CPU%20water%20cooler/cpu_water_cooler.htm

wiatroda
22-03-2011, 05:46 PM
Finally a clear pipe arrived today. I decided to try gravity circulating cooling system before I start to mess with any pump. At the moment of writing this a computer is on for more than an hour. Temperature of CPU is 33C. CPU usage is about the same like last time picture. So for now full success :dance:.

3866

Mad Professor
22-03-2011, 06:13 PM
When you get time can you run IntelBurnTest v2.50 and report your CPU temps?
http://www.ultimate-filez.com/files/IntelBurnTest.zip

This will really test how good your cooling is.

wiatroda
22-03-2011, 06:51 PM
This will really test how good your cooling is.
I'll, but to have precise cooling capability comparison I would have to use this program with old heatsink as well. For temp check I use ASUS AI Suite- program supplied together with mainboard. 2h 15min of running a comp and temp is 37C, but I believe as a result of water heating up in a water storage tank(3l plastic bucket).

3867

wiatroda
22-03-2011, 08:16 PM
I've attached two DS1820 temp sensors as pics below(below black tape) to measure temp of in and out coming water from cooler.

3868

That's a water temp readout shown in 0,1C.
Top line- Outlet
Lower line-Inlet
While pics below taken CPU temp - 42C

3869

wiatroda
05-04-2011, 08:53 PM
When you get time can you run IntelBurnTest v2.50 and report your CPU temps?

http://www.ultimate-filez.com/files/IntelBurnTest.zip

This will really test how good your cooling is.

I've run this test @ standard stress level,5 times to run. Max CPU temp reached 59 centigrade in a peak (when started this test the comp was running for nearly 10h non stop since morning). Is my cooling any good then?
Few issues appears:
1st- Top surface of the CPU is not perfectly flat
2nd- Average temp of the chipset is higher than 55 centigrade with standard heatsink.

3896

Mad Professor
06-04-2011, 06:59 AM
Just to compare, I have an Intel QuadCore Q6600 running at stock 2.40Ghz, and I am using air cooling via a Akasa Venom (http://www.akasa.com.tw/update.php?tpl=product/cpu.product.tpl&no=181&type=CPU coolers&type_sub=Retail cooler&model=AK-CCX-4002HPV2) cpu cooler, and 3x 120mm Akasa Viper (http://www.akasa.com.tw/update.php?tpl=product/product.detail.tpl&no=181&type=Fans&type_sub=Ultra Quiet Fan&model=AK-FN059) fans.

I have lapped both my CPU and Cooler to get them as flat as possible by hand.

Using Core Temp v0.99.7 it shows my idle cpu core temps are around 30-32c.
Running IntelBurnTest with a standard test level and 5 test cycles one of my cores peeks at 51c.
After 10 test cycles two cores peeked at 51c.

wiatroda
06-04-2011, 08:02 AM
My duo is oveclocked from 2.13 to 2.68Ghz. Temp of cooling water was 33degr. Core temp is 1.2V. As I said before flat surface of the CPU touching heatsink is not perfectly flat( doesn't make flat impression on cooler and piece of glass eider) and this can cause poor heat transfer. I don't want to mess with it any more, but I'm tempted to test CPU with stock heatsink. Did you remove CPU from socket to lap it ??

Mad Professor
06-04-2011, 08:18 AM
I did indeed remove the cpu from the socket to lap it.

I just used a small A4 sized sheet of glass, and taped down some Wet and Dry sand paper to the sheet, starting at a 600grade and slowly working my way upto a 2000grade, took about 30-45mins in total, I also did end up sanding down the tips of my fingers as well.

There are meny videos on how to lap a cpu on youtube, the main thing is not to apply pressure to the cpu as you are lapping it, keep rotating the cpu, and just take your time.

As you start lapping look at the cpu heatspeader now and then, and you will soon see how bad the heatspeader is, as you will get the copper showing round the outer edges of the heatspeader 1st, once you have the copper across the hole of the heatspeader it's time to move onto the finer grades.

Lapping my CPU knocked around 5c at load off my temps.

So if you was to lap both your cpu and water block I would hope to see you running in the low 50's when stress testing.

When I was running the stock cooler and before lapping the cpu, I ran IntelBurnTest, and the temps where in the mid to high 60's.

wiatroda
06-04-2011, 09:12 AM
I'm going to make cooler for chipset (I finished cutting out another one for graphic card and junction bars yesterday) ,so possibly I'll need to remove MB. This will be perfect opportunity to lap my CPU. Doesn't static electricity damage a CPU (in case touching the pins)?

M250cnc
06-04-2011, 09:34 AM
Jonathan got it right in an earlier post, and it looks so nice i didn't want to say anything negative.

Firstly i wouldn't touch the face of the CPU

You have three flaws in your design, it seems the wall thickness is to great the face touching the CPU should be in the order of 1mm or less

The flow of water should be controlled in a backwards and forwards motion in that it forms a single cavity from start to finish forcing the water to travel over all the surfaces

The size of the cavity should match the area of the pipe and within the cavity's should be fins for maximum heat transfer

For best heat transfer it should be made of copper

Air water etc will flow in the easiest route, in your design you can go from start to finish ports bypassing the edges allowing heat build up.

Phil

wiatroda
06-04-2011, 11:44 AM
Hi Phil
I found my old 12V water pump from water steriliser (long forgotten adventure with cichlides) and it will be circulating the cooling water. Currently CPU is water cooled, and the graphic card and motherboard chipset will be as well soon. I'm waiting for Peltier modules to come and I'm going to drop coolant temperature as low as possible without condensation. uC is going to control Peltiers by PWM in 8 bit mode. PWM duty cycle will be dependent upon comparing set temperature with readout.
For example coolant temp = set temp then duty cycle = 0 ( peltier off). Temp difference x2 = PWM register (up to 255) so coolant temp variation going to be around 1.28 centigrade creating from 0% to 100% peltier duty cycle.
Looks like this small silencing project going to be a monster one :smile: .
At last that's the theory, maybe somebody have some suggestions ???

Graphic card and manifold parts :
3897