PUMP: Thermaltake p500
RADIATORS: Thermaltake big water 760i & Swiftech MCR220
Coolant: UV reactive thermal take kit coolant + 50/50 antifreeze (needed extra fluid)
Project Start date: mid to end 2007
Project Delays: Finishing my BSME degree, having to work & not caring about project
Project Finish date: mid 2009
Lesson learned: Although i believe that my DIY water block out performs current market solutions and can be easily mass produced. According to my models the best water cooled thermal design and cost is simply an finned heat sink where water runs through. I'll release an model with temps later but I don't plan to produce it..... thermaltake or swiftech perhaps will make it one day and then I'll buy it.
Have a great day extreme systems!!
Last edited by Mr. Morello; 09-19-2009 at 05:04 PM.
Reason: Technical Clarification
Nice. However, alcohol might prevent corrosion, but it cracks acrylic, which I believe is what the thermaltake crap is made of. Get rid of it quick! Best not to mix aluminum and brass radiators. If it's solidworks, can we get screenshots of the matrix to see how it's designed? It seems interesting.
What software package is that? I have a little CFD and SPHD experience. I do GN&C now. To me, what the thermal guys do at work is an art.
I like the classic heat-exchanger concept. Out of curiosity, what made you go with a design which has velocity increasing by restriction through the heat exchanger only to immediately diffuse it afterwards. Did this show increased heat transfer to the fluid? I'd think it would be exactly the opposite due to the non-linear dependence on time.
I've been thinking about taking a stab at this myself when time and funds allowed. One of the things that I liked about swiftech was a blurb I read about them designing their diamond pin matrix using CFD. I hope the others use some sort of engineering, but in the back of my mind I've always suspected that it was primarily trial and error.
Boogerlad, I run the fluid Thermaltake provided + a little extra 50/50 radiator fluid because I needed more. Everything seems to be fine... no rusting or corrosion occuring.... Car's use glycol based antifreeze with aluminum, brass and copper and are problem free.
What I really meant to say about the alcohol + Benzotriazole was that the Benzotriazole was dissolved into alcohol and the parts were soaked in and then dried before assembly as for anti corrosion.
Meanmoe,
Forty 1mm holes drilled into a roughly 3/8inch diameter piece of copper is not easy or cost worthy and only provides minor gains in heat transfer. Even when overclocking (changing the model to simulate 130 and 200 watt). It also is restrictive and creates a larger pressure drop, which ultimately results in lesser flow through the system. Since it was a 1 off deal, I decided to have the machinist put 40 holes in it.
Yes, it was my intention to have turbulent flow (high velocity across a small cross sectional area) through the heat exchanger and then go back to laminar flow because of better thermal transfer.
I'm not going release prints or designs, as you may or may have not noticed everything is plastic deformed together (to have this type of design work correctly it requires experience) and it would take quite a bit of effort explain and even then you may not know GD&T and machining, assembly and tooling processes.
As stated before, I plan on releasing a different model of essentially a standard copper heat sink with better performance both for dual core and quad core systems. because of far better heat transfer (when overclocking) and lower pressure drop.
Last edited by Mr. Morello; 09-19-2009 at 05:19 PM.
Reason: Minor reword.... don't wanna assum anything about someone
Forty 1mm holes drilled into a roughly 3/8inch diameter piece of copper is not easy or cost worthy and only provides minor gains in heat transfer. Even when overclocking (changing the model to simulate 130 and 200 watt). It also is restrictive and creates a larger pressure drop, which ultimately results in lesser flow through the system. Since it was a 1 off deal, I decided to have the machinist put 40 holes in it.
Yes, it was my intention to have turbulent flow (high velocity across a small cross sectional area) through the heat exchanger and then go back to laminar flow because of better thermal transfer.
I'm not going release prints or designs, as you may or may have not noticed everything is plastic deformed together (to have this type of design work correctly it requires experience) and it would take quite a bit of effort explain and even then you may not know GD&T and machining, assembly and tooling processes.
As stated before, I plan on releasing a different model of essentially a standard copper heat sink with better performance both for dual core and quad core systems. because of far better heat transfer (when overclocking) and lower pressure drop.
I know very little with regard to fabrication, dimensioning and tolerancing. Only what I learned in school actually.
I understand that boundary layer reduction is critical due to the wall boundary condition for inviscid flow. I guess my question should've been, given a constant velocity, would you ever actually reach a steady flow through the heat exchanger given entrance effects and all. Or in otherwords, is the high velocity necessary if the length is small. And likewise, is the pressure drop necessary? It's a curiosity and you're a guy with a model and an engineering background so I thought I'd ask.
I would like to know what software package that is. It reminds me of a package called unigraphics that I used once for CAD and stress analysis. I know you can buy other plugins for it but that's probably not it.
In the original post, did you say that the model showed that it is better or worse than a finned design?
Quote:
Originally Posted by Mr. Morello
Lesson learned: Although i believe that my DIY water block out performs current market solutions and can be easily mass produced. According to my models the best water cooled thermal design and cost is simply an finned heat sink where water runs through. I'll release an model with temps later but I don't plan to produce it..... thermaltake or swiftech perhaps will make it one day and then I'll buy it.
Or in otherwords, is the high velocity necessary if the length is small. And likewise, is the pressure drop necessary?
I would like to know what software package that is.
In the original post, did you say that the model showed that it is better or worse than a finned design?
I like your work. Thanks for posting it.
-MM
I don't believe that turbulent (inviscid) flow is required to improving heat transfer. But it can be noted that greater flow does improve the block heat transfer.
In school, Bernoulli's equations, moodys chart and pressure drop were used for calculating and accounting for hydraulic losses. In a different model I used bigger holes six 0.125 inch to be exact. In this revised model there was less pressure drop but slightly higher temperatures 2C... When using double the power 130watts (overclocked model) there was a 4 C difference.
The surface area of the six 0.125" holes was half of the 40 1mm holes. I believe that the added surface area affected the temperature difference the most.
The software package is Solid Works.
Using the same model parameters the model showed big improvements with the finned heat sync style design, especially when overclocked on a quad core cpu model.
*ADDED Finned Concept*
Same parameters as classic heat exchanger as shown above, but as we can see there is a 6.5C improvement in temperature. But when overclocked...... the improvement temp gains are greater.
Last edited by Mr. Morello; 09-22-2009 at 09:24 PM.
Reason: Added Pic
I don't believe that turbulent (inviscid) flow is required to improving heat transfer. But it can be noted that greater flow does improve the block heat transfer.
I believe boundary layer reduction typically improves heat transfer... I think of it as improved flow at the wall (this is probably over simplified). I know that it becomes very apparent in tubes as d/L becomes small. The heat tends to build up at the wall because the velocity at the wall goes to 0. I think that this may be why pin configurations are used instead of fins but I don't know...
Quote:
Originally Posted by Mr. Morello
In school, Bernoulli's equations, moodys chart
...memories...
(Skinnee should review Bernoulli..."Total dynamic head pressure" )
Quote:
Originally Posted by Mr. Morello
The software package is Solid Works.
Thanks. I didn't know that it could be used for this type of analysis...
Quote:
Originally Posted by Mr. Morello
Using the same model parameters the model showed big improvements with the finned heat sync style design, especially when overclocked on a quad core cpu model.
Is this difference also due to surface area? There looks to be significantly more wetted surface in this design. Would it be worth your time to make some additional cuts in the model to make those fins into rectangular pins and re-run it? If it's done right, the wetted area will remain constant.
Again, compliments on your work... I've always wanted to do something similar.
I can't say jack really 'cause I'm not burning my free time testing components and then providing results to the community out of the goodness of my heart... Thanks Skinnee, sorry for the low blow...
Yeahhhh..... the total dynamic head pressure calculation BTW, having a reduced boundary layer as described for improved heat transfer makes sense.
Actually, I'm unemployed at the moment.... automotive industry has always been shrinking, but really tanked in 2008 :| Theoretically I should have plenty of free time....
Let me tell ya, there's nothing like being unemployed while phone screening with small companies for less than desirable positions and then being told you're entry level because you graduated from at the end of 2007 all the while they're really talking to me because I have a total of 5-6 years of experience completing mechanical engineering projects and also manufacturing engineering & supervision experience at large companies I guess I like larger preferably global work environments.
Yea, Solid Works can with add-in packages (I'm not an expert)
The model shows the same chip temp result with added cuts.
I'm still interested in making that standard square heat sink fin model
Moe, thanks for the complement I enjoyed sharing the project. To be honest a lot of this project was a learning experience because at work I usually don't usually get this technical.
Okay, it's been awhile.... I did some work on and off on the finned model and the results seemed to be too good to be true.
There were some bugs in my old model that I changed.
1.) The CPU die mesh was too course and needed to be refined, I thought this was selected with the heat spreader, but it wasn't!!!! This skewed my model a bit toward the high side. (model moved die temp to 84 degrees compared to 98)
2.) The model did not include lead free solder (http://www.intel.com/pressroom/archi...070522comp.htm) between the Die and heat spreader. Skewing model toward lower temps (This moved die temp from 84 to 88 and validates that removing IHS lowers temp because the solder has poor thermal conductivity. Copper is roughly 400W/M-K vs Sn-Ag-Cu solder 57W/M-K) Don't quote me here but I think my e6850 actually uses Sn-Pb, but what the hay a 7W/M-k difference is close enough (http://electronics-cooling.com/artic...g_techdata.php)
3.) The die was previously considered to be 1mm thick instead of 0.75mm (I scaled off of intel drawings and I used 0.25mm of lead free solder between that) Sure enough the 0.2grams stated is what the model calculates for the solder.
4.) Due to having a far better computer compared to 2007 and 64 bit CFD (64 bit means more ram is able to be used) software <3 The model was changed from a partial symmetrical simulation (half of the model is calculated to save memory and computation time) to a full simulation.
I also forgot about the previous test, it turns out the water was heating up, so the pic previously released was not entirely accurate. So that night I took a 5 gallon bucket and filled it with tap water, let it sit next to my comp for a day and then put the radiator in it to keep the temp stable.
So now were looking at a more accurate model now.... approaching 95% .. I'm guessing the rest of the error would be the supposed perfect contact between the IHS and the block as well as the assumed 3.33L/min flow rate.
Last edited by Mr. Morello; 10-13-2009 at 09:27 PM.
)
BTW, having a reduced boundary layer as described for improved heat transfer makes sense.
I guess I like larger preferably global work environments.
Linear Mode
