Custom liquid cooler for LGA 775 question

[merz.38]

Does the PC12800 ram have a separate form factor than say PC17600? Like conceivably could I switch my 1333 MHz ram for like 2200 MHz ram?

 

[RGE]

Nice machining, pretty cool to make your own stuff!

The retention scheme is made to support up to 1000 grams, so clearly ok with that weight. But since you likely will be using your own support mechanism, then the attachment force is really the limitation. The suggested force is 160N/mm.

The retention scheme employed by CEK is designed to support heavy
heatsinks (approximately up to 1000 grams)

from intel cpu white papers, like one here page 49.

Intel® Xeon® Processor 5300 Series CEK Reference Heatsink Design clip and fasteners assembly have a stiffness of around 160 N/mm [915 lb/in]

pg 85

Since using aluminum and copper in direct contact, I understand the need to use dielectic oil, but you will lose some cooling effect. Best oils have thermal conductance of ~ 0.2 W/MK about 1/3 that of water 0.6 W/MK, and spec heat of about 1.6-2 Kj/KgK less than half of water 4.18.

Just out of curiosity, are you using aluminum instead of all copper for weight consideration, or cost, or machining ease?

Also, if you could mill thin straight channels on copper block, and run the water across them with inlet on one side, outlet on other, more similar to internals in top performing blocks, you would get more surface area.

 

[merz.38]

Nice machining, pretty cool to make your own stuff!

The retention scheme is made to support up to 1000 grams, so clearly ok with that weight. But since you likely will be using your own support mechanism, then the attachment force is really the limitation. The suggested force is 160N/mm.


from intel cpu white papers, like one here page 49.

pg 85

Since using aluminum and copper in direct contact, I understand the need to use dielectic oil, but you will lose some cooling effect. Best oils have thermal conductance of ~ 0.2 W/MK about 1/3 that of water 0.6 W/MK, and spec heat of about 1.6-2 Kj/KgK less than half of water 4.18.

Just out of curiosity, are you using aluminum instead of all copper for weight consideration, or cost, or machining ease?

Also, if you could mill thin straight channels on copper block, and run the water across them with inlet on one side, outlet on other, more similar to internals in top performing blocks, you would get more surface area.

Okay, So I'm a little bit confused by the 160N/mm, is this the amount of deflection that the board incurs do to the weight of the heat sink?

Thanks for the Intel paper, I'll go through this in detail when I start to design the mounting brackets.

As far as using the oil, I will not be using oil to separate the board and heatsink, there will be no contact between the heatsink and anything else but the top of the processor so I probably don't need to worry about that. I also don't want to need to perform any maintenance on this over the life of the cooler, so, oil is a far better choice than water because I don't want to have to worry about oxidation on the internal surfaces.

I was talking about using oil as a coolant because of its dialectic characteristics, i.e. in case a fitting leaks or I screwed something up, I don't ruin my 2500$ computer.

The aluminum was a good choice because of weight, cost, and ease of machining.

I'm a little bit confused by your last remark, Could you maybe elaborate on it a little bit more? I chose the layout because I could easily apply a seal design that I had used for another project in the past and knew worked well.

 

[merz.38]

Nice machining, pretty cool to make your own stuff!

The retention scheme is made to support up to 1000 grams, so clearly ok with that weight. But since you likely will be using your own support mechanism, then the attachment force is really the limitation. The suggested force is 160N/mm.


from intel cpu white papers, like one here page 49.

pg 85

Since using aluminum and copper in direct contact, I understand the need to use dielectic oil, but you will lose some cooling effect. Best oils have thermal conductance of ~ 0.2 W/MK about 1/3 that of water 0.6 W/MK, and spec heat of about 1.6-2 Kj/KgK less than half of water 4.18.

Just out of curiosity, are you using aluminum instead of all copper for weight consideration, or cost, or machining ease?

Also, if you could mill thin straight channels on copper block, and run the water across them with inlet on one side, outlet on other, more similar to internals in top performing blocks, you would get more surface area.

Also, the copper is 15X more expensive than the aluminum in that form factor.

 

[RGE]

15x..ouch.

The heatsink clamp + weight of the heatsink should apply ~ 160N/mm or 950lbs/inch of force. The sink weight limit of 1000 grams is only because intels heatsink attachment mechanism may not hold more weight than that once motherboard is in vertical position. If building your own attachment for yours, which you will do, then I dont think you need to worry about weight limits for what you are building. As for 160N, just means screw it on tight .

I meant your heatsink is part aluminum and part copper. Swiftech made one long ago with copper bottom and aluminum top, it corroded horribly even with inhibited glycol. But with dielectric, should have less of an issue. And yep, dont have to worry about leaks killing computer. Here, scroll down to aluminum/copper title under corroded rad is pic of swiftech copper base, aluminum top with inhibited coolant.

As for the surface area/internals of block, I just picked this one at random at skinneelabs, happened to be nickel plated, but can see high surface area. Wasnt thinking of that many channels but similar design.

Though will be interesting to see something different, and how it performs.

 

[merz.38]

15x..ouch.

The heatsink clamp + weight of the heatsink should apply ~ 160N/mm or 950lbs/inch of force. The sink weight limit of 1000 grams is only because intels heatsink attachment mechanism may not hold more weight than that once motherboard is in vertical position. If building your own attachment for yours, which you will do, then I dont think you need to worry about weight limits for what you are building. As for 160N, just means screw it on tight .

I meant your heatsink is part aluminum and part copper. Swiftech made one long ago with copper bottom and aluminum top, it corroded horribly even with inhibited glycol. But with dielectric, should have less of an issue. And yep, dont have to worry about leaks killing computer. Here, scroll down to aluminum/copper title under corroded rad is pic of swiftech copper base, aluminum top with inhibited coolant.

As for the surface area/internals of block, I just picked this one at random at skinneelabs, happened to be nickel plated, but can see high surface area. Wasnt thinking of that many channels but similar design.

Though will be interesting to see something different, and how it performs.

When you say corroded horribly, what do you mean by that? From the inside out? Outiside- in? Both? I was going to apply some sort of clear coating to the exterior surfaces of the cooler.

Also, the skinnelabs thing is not something i feel like I could make, at least not with my experience, availability of tools, or free time. I'm not even sure how they made the fins on that. Some type of forging? CNC plasma cut? I believe my design to be a fairly efficient use of space in the fact that the walls separating the coolant will also act as fins. Also, I plan on using something like a marine-grade bilge pump with like 1000 gph of flow. The deep channels will permit for high volumes of fluid to flow easily.

Also, side remark, I need to make several circular gaskets out of silicone rubber, does anyone know where I can find a punch that will make a perfect circle of about 1.5 in OD?

I'm also considering adding some sort of galvanic protection to the whole apparattus, but I need to whip out my old materials science book to figure out what i'm going to use. Maybe i'll bolt a magnesium plate on it somewhere.

I'm going to be lucky if I can finish this in a reasonable amount of time bc I have finals coming up in 2 weeks and then next quarter is going to be pure hell. 21 credit hrs of 500+ level engineering classes :-(

 

[RGE]

I was just describing the corrosion in the pics in link I posted.

one thought if your cooler is 1000 grams or even close to it, especially if your mobo is going to be vertical, you might want to look at making a mobo back plate (with rubber insulator) to minimize mobo bending.

 

[merz.38]

Ah okay, I totally didn't see that link before.

So here's a fun question, hasn't anybody considered using galvanic protection to preserve their equipment?

Also, does everyone use water as a coolant as opposed to oil?

Also I was going to base the mounting strategy off of the design used for the V-8. I I was going to use either delrin or if I have some silicon rubber left over some of that i'm using for internal gaskets. I was then going to install some sort of cantilever type support on the bracket for it to rest on to eliminate shear stress on holes.

 

[RGE]

From 8-9 years ago til about 4-5 years ago, I used distilled plus antifreeze, as did most others back then. But with todays choice of all brass/copper rads, all copper waterblocks, you can get buy without inhibitors, and many now just run distilled water plus biocide. Plus most update equipment at least every 3 years, ie new cpu socket, so new mobo, new cooler, new vga card so new waterblock, so only has to last few years at most.

The downside of antifreeze is some inhibitors, esp silicates, etc in antifreeze cause gunking in the pins in the waterblocks like pic here. Many had that issue, and multiple posts with pictures of crapped up waterblocks exist on this and other forums.

One professional cooling site states:

However, automotive glycol should not be used in a cooling system or heat exchanger because it contains silicate-based rust inhibitors. These inhibitors can gel and foul, coating heat exchanger surfaces and reducing their efficiency.

Performance wise you only lose 1C (skinneelabs testing) with 15-20% gycol though, and you can find a few without silicates, but no proof they wont cause gunking either. Or you can use concentrated inhibitors, and not lose any performance (still 98+% water), but they dont tell you if they contain silicates or not.

Also the coolant skinneelabs tested that had sp heat about 25% lower than water, and thermal conductance 50% of water, and higher viscosity, the cpu temps were 8C higher than for water. That wouldnt fly with most of the OCers on this forum.

According to a survey on EK waterblocks site, about 48% people use inhibitors, about 52% just distilled/biocide. But on this and few other forums, more skewed towards plain distilled and biocide.

 

[merz.38]

Interesting, and gross. Can that type of buildup be eliminated by incorporating a filtration system into the loop?

Also, One thing I noticed is that the fittings on that block had some sort of coating.

Is it possible to minimize this sort of buildup by using fittings without any type of coating i.e. brass or steel?

 

[RGE]

Filtration would likely help. Would hinder flow too much in our loops, but if you are using a strong pump...but then flilter would get clogged, so either way will be cleaning maintenance.

Most fittings are nickel plated brass, nickel is supposed to reduce issues with corrosion/oxidation, though may not always work that way. My nickel plated (inside and out) brass fittings are still pristine after 3+ years of use.

Also not everyone gets gunking with antifreeze, the majority dont, I never did. That one person was folding 24/7, ie computer on full load with high temps 24/7, probably increases risk for getting it. But sucks when you do get it.

And for running with just distilled + biocide, here is what a typical all copper block looks like after 1.5yrs. Only affected performance 0.5C, comparing before to after cleaning. I am now running with distilled 98% with 2% concentrated inhibitors, and see if get more or less fouling or oxidation, ie just for experimental value.

 

[merz.38]

Filtration would likely help. Would hinder flow too much in our loops, but if you are using a strong pump...but then flilter would get clogged, so either way will be cleaning maintenance.

Most fittings are nickel plated brass, nickel is supposed to reduce issues with corrosion/oxidation, though may not always work that way. My nickel plated (inside and out) brass fittings are still pristine after 3+ years of use.

Also not everyone gets gunking with antifreeze, the majority dont, I never did. That one person was folding 24/7, ie computer on full load with high temps 24/7, probably increases risk for getting it. But sucks when you do get it.

And for running with just distilled + biocide, here is what a typical all copper block looks like after 1.5yrs. Only affected performance 0.5C, comparing before to after cleaning. I am now running with distilled 98% with 2% concentrated inhibitors, and see if get more or less fouling or oxidation, ie just for experimental value.

Okay, I was thinking using an automotive oil filter, that'd be the most economical way I can think of at ~5$ for a filter.

That's good to know about the nickel plated fittings, what about the pure brass ones? I already spent like 25$ on a set of heavy duty 3/4" NPT fittings.

How powerful of a pump are you running in your setup?


Also that's an interesting study. I'll pay attention

 

[RGE]

pure brass works fine as well. Pumps I and most others use are here.

 

[merz.38]

pure brass works fine as well. Pumps I and most others use are here.

Yike's that isn't cheap. I was thinking about using something like this http://www.amazon.com/Volt-Marine-U.../ref=sr_1_182?ie=UTF8&qid=1330829629&sr=8-182

I could splice some connectors onto it and run it off of my comp cpu. Its fully submersible so it could just sit in the reservoir and release heat into the reservoir so i wouldn't have to worry about the pump overheating.

 

[Conumdrum]

Ack, No, thats a short use pump, not meant to run for years with no problem.

It's silly huge. I 'm sure it's loud. 7.5 Amps? A huge heat dump problem. NO, please NO.

There is a reason we use very few designs of pumps to WC a PC. Quiet, reliable, small, low wattage heat dump, good flow and pressure charecteristics. And the 35X can be bought for $80 at other places.

 

[merz.38]

That just seems so expensive for not a whole lot of flow. It doesn't need to fit inside of my case.

 

[SolidxSnake]

Ack, No, thats a short use pump, not meant to run for years with no problem.

It's silly huge. I 'm sure it's loud. 7.5 Amps? A huge heat dump problem. NO, please NO.

There is a reason we use very few designs of pumps to WC a PC. Quiet, reliable, small, low wattage heat dump, good flow and pressure charecteristics. And the 35X can be bought for $80 at other places.

How much heat does a 10W DDC-1 or 18W DDC-2 put into a loop? If you tell me 10W and 18W respectively, I can safely tell you that you should look deeper into power and efficiency.

I'm not disagreeing with you entirely (I'm sure the marine pump has a higher heat dump than an MCP35x), but I'd be surprised if the pump he picked would noticeably raise loop temperatures. It consumes 90W of power, but that doesn't mean it dumps 90W of heat into the loop (and judging by just how much more flow/head it has than the MCP35x, I'm sure the majority of that 90W goes into useful energy, not residual heat).

edit: merz.38: You'll notice you're going to get a bunch of suggestions to use PC-oriented parts and such (for good reason, mind you), instead of other ones. I would take a look at these parts and think about why they were designed the way they were. For instance, the MCP35x is one of the best, if not the best computer-designed pumps. It has plenty of flow and head pressure for almost any WC loop, it's very quiet, and it's extremely compact. It also uses very little power. All things that are a huge plus for PC watercooling. What you'll have to think of is why the parts you may initially think to use would not be beneficial in a computer loop, and then judge which parts to get from there.

 

[merz.38]

How much heat does a 10W DDC-1 or 18W DDC-2 put into a loop? If you tell me 10W and 18W respectively, I can safely tell you that you should look deeper into power and efficiency.

I'm not disagreeing with you entirely (I'm sure the marine pump has a higher heat dump than an MCP35x), but I'd be surprised if the pump he picked would noticeably raise loop temperatures. It consumes 90W of power, but that doesn't mean it dumps 90W of heat into the loop (and judging by just how much more flow/head it has than the MCP35x, I'm sure the majority of that 90W goes into useful energy, not residual heat).

edit: merz.38: You'll notice you're going to get a bunch of suggestions to use PC-oriented parts and such (for good reason, mind you), instead of other ones. I would take a look at these parts and think about why they were designed the way they were. For instance, the MCP35x is one of the best, if not the best computer-designed pumps. It has plenty of flow and head pressure for almost any WC loop, it's very quiet, and it's extremely compact. It also uses very little power. All things that are a huge plus for PC watercooling. What you'll have to think of is why the parts you may initially think to use would not be beneficial in a computer loop, and then judge which parts to get from there.

Hmph, I haven't actually considered that, I suppose I need to do a bit of research on pumps. I was hoping to get something with a robust stainless steel or cast iron housing.

It just seems to me like they're charging 100$ for a pump, just bc they know what people using it for a wc loop will pay a premium for it. Then again I could be totally wrong.

I could also add an intercooler after the pump to reduce the heat added to the flow by the pump.

 

[merz.38]

How much heat does a 10W DDC-1 or 18W DDC-2 put into a loop? If you tell me 10W and 18W respectively, I can safely tell you that you should look deeper into power and efficiency.

I'm not disagreeing with you entirely (I'm sure the marine pump has a higher heat dump than an MCP35x), but I'd be surprised if the pump he picked would noticeably raise loop temperatures. It consumes 90W of power, but that doesn't mean it dumps 90W of heat into the loop (and judging by just how much more flow/head it has than the MCP35x, I'm sure the majority of that 90W goes into useful energy, not residual heat).

edit: merz.38: You'll notice you're going to get a bunch of suggestions to use PC-oriented parts and such (for good reason, mind you), instead of other ones. I would take a look at these parts and think about why they were designed the way they were. For instance, the MCP35x is one of the best, if not the best computer-designed pumps. It has plenty of flow and head pressure for almost any WC loop, it's very quiet, and it's extremely compact. It also uses very little power. All things that are a huge plus for PC watercooling. What you'll have to think of is why the parts you may initially think to use would not be beneficial in a computer loop, and then judge which parts to get from there.

Okay, this is good advice.

You'll have to forgive my skepticism about my hardware selection. I haven't dabbled with liquid cooling at all before.

 

[SolidxSnake]

Okay, this is good advice.

You'll have to forgive my skepticism about my hardware selection. I haven't dabbled with liquid cooling at all before.

No worries. You're going to get a bunch of "read the stickies" and advice/comments that may not make sense. Just keep an open mind, is all I can say. You seem very ambitious and willing to learn, so just spend some time learning about why certain things have been so common in the PC watercooling field, and what it is you personally want to achieve and how the two relate.

Things like impingement designs in waterblocks are commonplace for a reason. Understanding why certain design choices have become (near) standard will help you design and build your own loop.