DIY CPU Waterblock (pics inside, 56k warning) and results!

DIY CPU Waterblock (pics inside, 56k warning) and results!Update(8/23)

This was originally posted over at THF, not many interested in WC on that forum lol; probably 4-6 real WCers there). So here we go (PS: 1st post)...........

\\\\\\\\\\\\\\\\\\\\\\\\\
Here are the final results. Block was CNCed (Aluminum 6061); couldn't do on Copper C110 due to cost. The water inlet fitting has some "fittings" inside of it to change/improve turbulent flow. The outlet fitting is untouched.

Leak testing:

Sorry for the mess. Took over my sis's bath room.

A quick rendering of the block:


The rad with a 120mm 110CFM Scythe Slipstream fan:


All assembled on a Celeron D 315 @ 3.6Ghz (CPUZ valid: http://valid.canardpc.com/show_oc.php?id=261994 )


Calculated the ~ Wattage of this thing:

uch: 141W!!!! with OC. No wonder these were called Preshot's :lol:

The first 10 minutes of P95 small FFT

Temp2 is the CPU temp

After 35 minutes:

Temp2 is CPU.
Note: Voltages shown in SpeedFan are wrong.
Sorry, I had to take pics of the screen with a camera as the test PC was not hooked up to the net (couldn't print screen as the mouse wasn't hooked either ). Heck doesn't even have Win 2k SP4 loaded lol

Not bad eh? With a Thermaltake P4 SPARK7+ (lapped) the results were 64C under load.

The block was not lapped yet (waiting for 400-1000grit and papers to be back in stock at the local Autozone). CPU was previously lapped. The CPU block is a quite concave so lapping should improve results by ~3-4C.

The Brass fittings are coated with 4+ coatings of clear automotive paint so not much to concern as far as mixed metals goes.

============
What I need help in:
1. What would be the best way to go about making a bracket/bolt in for LGA775?
2. Can I use moded a bolt in bracket from a S1283?
3. I lost my old LGA775 data sheets with all the measurements,guide lines,etc, if you can link me that would be great.

==========
Edit

Ill tell you again ! NICe job !

wow did your sis raise hell?

looks like a great job !

Thanks guys. So any one got any tips on building a DIY bolt in/bracket for this for LGA775?
I'll probably move this (if possible) to a E5200 @3.4Ghz or E2180 @3Ghz (limited by heat right now) for more testing after lapping, so any help on making a bracket would be vary appreciated

Nice job, and excellent first post. Welcome to OCF!

to the forums

Like it!
its a beauty... love the wads of hot melt stuck to the fan, great (I love to hot melt in a bodge it fasion)

think I have to hand my ghetto crown over to you for that one... much more home brew modding effort.

heres my out shined effort


looks like it was fun.

i think a H bracket with some thumb screws and springs would do the trick...

you can use this page for the white papers, wouldnt matter if you even went to the core 2 section. the data for LGA775 is going to be linked to the same PDF's.
http://www.intel.com/design/Pentium4/documentation.htm
check out the data sheets section

Evilsizer said:

i think a H bracket with some thumb screws and springs would do the trick...

you can use this page for the white papers, wouldnt matter if you even went to the core 2 section. the data for LGA775 is going to be linked to the same PDF's.
http://www.intel.com/design/Pentium4/documentation.htm
check out the data sheets section

Click to expand...

yes, try looking for someone selling a TT block and you should be able to make use of the H bracket from it.

and well done, i would see if you can find some aluminum barbs, so your not running a mixed metal loop.

Spawn-Inc said:

yes, try looking for someone selling a TT block and you should be able to make use of the H bracket from it.

and well done, i would see if you can find some aluminum barbs, so your not running a mixed metal loop.

Click to expand...

You could just go to the local HW store and pick up some vinyl barbs for a few bucks.

Looking at the pics it looks to me like an H bracket wouldn't fit inbetween the barbs there
If it was me i think i would just make a sqare ( or however that is spelled ) bracket and you might even be able to use the stock ring around the socket to attatch the retentionmechanism to
Not sure if retentionmechanism is the right word tho

If you are looking for sandpaper, try this link-->http://www.ocforums.com/showthread.php?t=245224

A lot of good reviews (me included) for prompt shipping and good quality stuff.

Got me sand paper today (@Super Nade: Thanks for the link, it's cheaper than buying in bulk esp. when you don't need 5 sheets of each grit lol)

Looking at the pics it looks to me like an H bracket wouldn't fit inbetween the barbs there
If it was me i think i would just make a sqare ( or however that is spelled ) bracket and you might even be able to use the stock ring around the socket to attatch the retentionmechanism to
Not sure if retentionmechanism is the right word tho

Click to expand...

Yeah, probably would have to do something like that as H bracket probably won't work due to barb placement.

Thanks all for the input so far!

would see if you can find some aluminum barbs, so your not running a mixed metal loop.

Click to expand...

Yeah, I was looking for them, but couldn't find the "right shaped (as in the inside) barb that's aluminum. Got 4+ coats of Clear Coat on the barbs (inside & out) so not much risk in mixed metal.

Super Nade said:

If you are looking for sandpaper, try this link-->http://www.ocforums.com/showthread.php?t=245224

A lot of good reviews (me included) for prompt shipping and good quality stuff.

Click to expand...

heat sink = heat spreader on the CPU ?

schmide said:

You could just go to the local HW store and pick up some vinyl barbs for a few bucks.

Click to expand...

lol, right! completely forgot about those.

actually now that i looked closer at it, you may want to try the swiffy hold down plate for a 6000/6002 water block. bard placement seems nearly the same to your home block.

Lapping time!

120grit. About 2minutes in. You can see how un-even it is. Mainly at the center.


All major problems gone:


220 grit:



400 grit:


600grit:



800grit:


1000grit; W00t! Mirror finish!


The Mess:


Close up of the block (sorry; the pic wasn't steady)

you may want to try the swiffy hold down plate for a 6000/6002 water block. bard placement seems nearly the same to your home block.

Click to expand...

I will look in to that. Thanks!

Shadow703793 said:

Got me sand paper today (@Super Nade: Thanks for the link, it's cheaper than buying in bulk esp. when you don't need 5 sheets of each grit lol)

Click to expand...

I suppose bulk ought to have worked better for me given that I am a bit challenged when it comes to polishing stuff. How hard can it be to muck up a polishing job? Well, I'd rather not say...

I think flatness is more important than a mirror finish. Given that any irregularities show up as a concave surface (predominantly), you can shine a laser pointer on the surface. Once you do that, take a piece of paper and look for the image (a white wall about 4~5 feet away would work). If the image is distorted or enlarged, you have a concave surface. The premise is that the concavity is very small.

ricflairw000 said:

heat sink = heat spreader on the CPU ?

Click to expand...

It could be any surface that needs to be polished.

I think flatness is more important than a mirror finish.

Click to expand...

Agreed, but for my use it wouldn't change much (as in temps; may be run .2-.5C cooler which I won't be able to measure) imo.

I should have posted this earlier but I designed this block based on actual 2+months of research and design. I designed this as a project to be entered in the 2009 SIEMENS COMPETITION (see: http://www.collegeboard.com/siemens/ ) and Intel International Science and Engineering Fair (see: http://www.societyforscience.org/ISEF/ ) and Intel Science Talent Search (see: http://www.intel.com/education/sts/ )

I transferred designs like the Swiftech GTZ on to CAD and ran simulations (which were quite accurate as it was +/- 2-7C from actual real world results) and looked at what made them "tick". Then I came up with my own designs (10-15) in CAD that were 90* turn from usual designs (GTZ,etc). Then I picked the best design based on virtual simulations (aka CFD) and CNCed/and a bit of hand work with a Dremel the final design you see here. Depending on what goes down, I plan to release all my research and designs to the public so others will benefit and maybe even make some money

Software used:
Thermal Desktop
FloWorks
SolidWorks
Inventor
CF Design

Update time!

Here are the temps after lapping:
10 minutes of P95:

Temp2 is CPU. A nice 28C! W00t!

30minutes of P95:

Tem2 is CPU. A pretty good 33C.

With out load, it sits at around 25-28C.

Load temp dropped by 1C for 30minute test.