Martin's custom "Thin Line" waterblock worklog

[TopherTony]

cos copyrighted the diamond 45 square drawing i submitted to gabe many moons

how is this relevant?

[Martinm210]

Thanks guys! Havn't figured out yet if I'm going to sell them, just testing and looking for input on improving it right now...

Looks like I have room for improvement. I've included all of the CPU testing I've done to date: Stinger V8 with artic silver 5, Estimated stock fuzion AS 5, Bowed/nozzled Fuzion AS5, then Bowed Nozzled Fuzion Coolaboratory Liquid Pro, and finally this block also using Coolaboratory liquid pro.

The "Estimated" is simply taking my bowed/nozzled test and added 4C load, 3C idle, because I never went through the rigorous testing I do currently with my fuzion when completely stock, but I did see about 4C gain under load and 3C on idle after bowing/nozzling/and sealing the inner chamber.

Bottom line, the block does well. It does do better than my stock fuzion runs, but not quite as well as my showerhead custom nozzle and bowed base fuzion.

My testing consists of running 5 tests for each idle load scenario and averaging them. During each test I let the PC temps stabilize then using TAT logger I start logging continuous core temps, and while these core temps are logging I manually start logging the ambient temp entering the radiator and water temp from a probe in the base of my T-line every 30 seconds. In the end I have about 100+ CPU loggings for every test that I average, and I average 10 recordings of ambient/water temps. My water/air temp sensors are just digital aquarium temperature probes that are specd within 1C, but give readings down to .1C. I figure with all the logging and averaging, it's a fairly good method and I've done it the same for each test except for my noted "Estimate Stock Fuzion AS5".

I have questioned my Stinger results, but even remounting and remounting doesn't seem to improve it, it's probably just the poor LGA775 socket that messes with it.

Anyhow, this is what I have, I think I beat a stock fuzion:thumb:, just not a modified nozzle/bowed variety:o:

I'll leave this block running for a while to see if any additional curing makes a difference, but the CLP is fairly fast at curing, instructions say 48hours, and this is after 24 so I don't expect much more.

Hmmm...what's in store for version 2??...

[tiro_uspsss]

nice results!!

I have the chance to buy a *slightly* damaged mill-lathe... usually retails AUD$2600, I can get it for AUD$1000 - Dad says its easily fixed.. should I??

[Martinm210]

nice results!!

I have the chance to buy a *slightly* damaged mill-lathe... usually retails AUD$2600, I can get it for AUD$1000 - Dad says its easily fixed.. should I??

Sure if it's in good shape, you can usually see how worn a mill/lathe is by inspecting the ways very carefully and look for differences. It should also have good oil fittings and preferably oil channels cut into the ways.


I will always give a to investing in good tools. Taken care of can last a lifetime..

Wish I could say the same for my video card..or anything in my computer for that matter..

[tiro_uspsss]

Sure if it's in good shape, you can usually see how worn a mill/lathe is by inspecting the ways very carefully and look for differences. It should also have good oil fittings and preferably oil channels cut into the ways.


I will always give a to investing in good tools. Taken care of can last a lifetime..

Wish I could say the same for my video card..or anything in my computer for that matter..

its brand new - just tool a spill of a fork-lift - my Dad started work as metal-work guy, he inspected it, says its too easy ... its 750mm BC size...

[Martinm210]

its brand new - just tool a spill of a fork-lift - my Dad started work as metal-work guy, he inspected it, says its too easy ... its 750mm BC size...

Oh yeah, that would be a nice investment, you won't be sorry and if you decided later you could probably sell if for nearly what you paid new.

[perfection]

hey, in the graph, you use AS5 and CLP, what TIM is CLP?

[Baleful]

Awww man nice results. How many times have you remounted, because you should be able to get 1.3 C easy w/ a really good mount. Awesome job man!

[Martinm210]

hey, in the graph, you use AS5 and CLP, what TIM is CLP?

It's Coolaboratory Liquid Pro - Liquid soldering, total PITA to install it, and pretty much requires lapping to remove it, but it performs really well so I put up with the suffereing..

Just a guess, but I think it works best on lapped CPUs where you have direct copper contact because it bonds better with it.

http://www.sidewindercomputers.com/colipro.html

Awww man nice results. How many times have you remounted, because you should be able to get 1.3 C easy w/ a really good mount. Awesome job man!

Thanks! this is just one mount, and I know to do it right I really should repeat 4 or 5 times. I'm hoping my rocker arm mounting plate minimizes mounting variability, but I don't know.

[bobo5195]

Measure the inlet and outlet water temperature and the water flow. Therefore watts outputted by the processor is known roughly by

W = Mass flow water (m^3/s) * 4200 * (Temp out – Temp In)

This allows you to calculate to a degree of accuracy what C/W is without resorting to my CPU is running about something. Note that this method is highly prone to variances in your test data though.

[Solarfall]

martin must say that your water block looks just fantastic.. awesome craftsmanship mate

[Clockwork_Apple]

Amazing work, I wish every thread were like this. It's rare to see someone with your skill and patience create something as impressive as that.

Results look good aswell, beats a stock fuzion?

[hecktic]

stinger v8 w/ CLP ?

[nealh]

I hope some who thought perfromace would not be up to par...chime in and say they were surprised

Excellent results so far...stepped base, impingment jets, dual outlet..wow

Your skills are excellent, top quality look and performance



I am interested to see how pin spacing and shape have an effect if you decide to change these parameters

[Jedda]

stinger v8 w/ CLP ?

Stinger V8



CLD answered above.



Yes Neal, it'd be interesting to see which feature has the most influence or if it's purely a combination of increments.

[Duh]

I really like you block martin..

[Martinm210]

Thanks guys,
While I didn't outperform my modded fusion, it did pretty good considering, and I'm happy

I didn't run my stinger with CLP, I assumed that because of the LGA775 socket wierdness that I'd get better results from the fuzion and it's ability to bow. I suppose I could mill in a step on the stinger(again thanks to Niksub1 for his work on this ), but I'd have to be really careful, I believe the base is .080" and some of the dimples leave as little as .040. I also should probably make a run with my fuzion completely stock and log it all, my estimated columns is just a rough estimate and CPU testing is exhaustively boring.. .

I do really like the stinger's build quality though, I just think this socket needs bowing or otherwise to make the best of it. All copper and brass goodness is great though

Thanks bobo5195, I should set up some temperature sensors and a flow meter to do that, that would be a great way to at least come up with some sort of measuring baseline in addition to temps. Even if I don't specifically know what heat output the processor is producing as long as it was all relative it would be a helpfull look particularly to see how performance changes with flow rate.

[Martinm210]

stinger v8 w/ CLP ?

http://www.stingerwaterblocks.mysite.com/

CLP was just my Thermal paste, Coolaboratory Liquid Pro..

[Deux]

looks great, you can probably beat the fuzion with a tweak or two. What copper alloy did you use and how tough was it to mill?

[Martinm210]

looks great, you can probably beat the fuzion with a tweak or two. What copper alloy did you use and how tough was it to mill?

http://www.petrastechshop.com/2x2x1cobl.html

I believe it's C110, the milling is good using WD40 as a coolant. I started trying to do some dry milling, but after enough practice I won't do anything without coolant. The slot cutting is impossible without coolant, but you get so much better finish cuts water cooling, I even used it for milling the lexan, with the WD40 it eliminated the heat and gumming up problem.

I also have a few bigger slabs of C110 1/8" and 1/4" I picked up from ebay.

Sort of ironic that you need "Watercooling" to make "Watercooling" parts..

[Cupcake]

NOOOO !!!! I have had almost the exact same idea floating around in my head for a while now , Anyway I doubt it would have turned out that nice

Wonderful job man

[Start]

Great job so far Anymore updates?

[Martinm210]

Great job so far Anymore updates?

I did a few checks on temps the next day, and I may have gained fractions of a degree with curing of the CLP, but nothing significantly different than my posted temps. It's really close to a bowed/nozzled fuzion, next I've been thinking I need to focus on only very small incrimental and tested changes.

For starters I think I'd like to expand the center nozzle length clear across the block. This would force the flow through the pins to be very balanced, but it may also take some emphasis off the core. I'll give it a try, it's easy to do...will post pics when I make that small change.

I've actually been working on a custom block for someone else. It's their own design, and I'm just having fun milling it for them. I also have some other work for someone else also lined out, so I think I'm getting a much appreciated response to the work. I'll never be able to compete with a commercial block manufacturer in terms of fabrication time & costs, but I think I do have something to offer in terms of a custom service. Commercial block can't offer you the option of selecting materials or little tweaks here and there to make it a one of a kind block, but that's something I can easily do.

PM me if anyone is looking for a custom block or part. I don't have alot of capacity, but I'm interested in making different things. And since I don't really know how much time anything will take me, I'm keeping cost down to basically just cover materials and tooling...I'm cheap for now..

On the side I'll continue some of my own projects like this block and some different designs and testing of their performance. I already have a full time regular job, so I really don't need to make any money from this work, and I don't want it to become work. I'm fortunate to have the opportunity to work as much overtime as I could ever want and make really good money, but I like playing too much..

Besides...with two young kids... a little milling in the very quiet garage is just soo relaxing right now..

[FallenCow]

Wow... great looking piece of kit you got there! Hopefully you'll considering putting it into production at some point! And props for offering to do custom work as well!

[Martinm210]

Update on Version 2... 11-2-07, deburred and dished base between pins, removed nozzle holes in favor of a continuous slot.

I decided to try very small changes to my CPU block at a time and test results thoroughly.

For this round I did the following:

Deburred the copper pin base with a 1/32" drill bit. There were some burrs left from the slotting saw that this cleaned up and it also left the base between the pins with a more efficient rounded bottom in between all the pins.

Second, I noticed that the block was fairly restrictive, almost more than my nozzled fuzion, and that the nozzle holes didn't necessarily line up with the pin openings.
The nozzle change:


New results...getting closer one degree at a time... Actually it's doing very very well for how unrestrictive of a block as it is. The nozzles on the fuzion add alot of restriction, and this is much less restriciton.


Might try milling the base down thinner for the next round..