May I pick your brain on DIY/custom parts and general water cooling concepts?

[derracs]

First post so I'll do a quick background. I'm a 'teach yourself' type of guy and I started on computers early in high school. Unfortunately, after a rather successful venture in building my first PC, I hit personal problems and had to give it up, but the joy of tinkering with it still lived on (yes, I am still running a 478 P4 as we type). Fast forward through almost a decade and a half of intense civil litigation, paying for school, at one point working three jobs, and getting married and I'm finally in a place to pick up some of my old hobbies.

That said, please forgive the rather noobishness of my questions or understanding that may come out in some instances - I am still going through a crash course in the last decade of computers. Also, 95% of my ideas are born of my own head and probably some of the stupidest ideas you will hear. I haven't been reading topics and thinking "well what about this..." When I built my first computer, water cooling was considered cutting edge and only done by those CRAZY people who wanted bragging rights. I thought my ideas were going to be bleeding edge and pushing the envelope, but I came to find out I'm actually in the realm of retro with my ideas - which is a good thing though because it means there is experience I can draw off of.

Also, I'm NOT looking to be held by the hand and guided to my destination, but looking for more of a "That's the wrong direction; go this way" type of approach. I'm also not 100% against pre-fab, but as a machinist and a tinkerer I'd rather spend twice as much for a less effective and hideously ugly unit and KNOW exactly what's going on inside than have a store bought unit. I also have NOT set my mind on hardware so these are more general questions.

That said, I have several questions. Most are the results of reading and research that has addressed one side of an issue but not fully addressed the entire issue and I'm either curious or confused. I have tried using the search function but sometimes that was a dead end, the term(s) were so common I couldn't find an answer, or worse - it caused even more questions. So here we go:


1. If I go water, I'm thinking of making my own blocks of solid copper with threaded barbs brazed in. My initial thought was to mill a channel with a ball endmill in a spiraling type pattern with an inlet at the center and the outlet towards the edge, a thin 1/32 or thinner plate would be brazed of soldered to closer it air tight with a quick pass by a fly cutter and hand polished to ensure a flat surface. My logic was this would not cause a reduction in the line and pressure/flow would stay the same, the coolest coolant would be delivered to the hotest spot first, it would maximize time over the chip without allowing stagnant areas (such as corners if it were square in design), and would be thin walled to allow maximum heat transfer and minimal heat retention by the block itself.

However, I have seen a lot of designs that look like flowers, or mazes with dead ends, etc that seem contrary to my own logic. I also read a thread, I think though the actual extreme cooling FAQ, that stated something along the lines of you want flow rate in the lines to be high but in the blocks to be low. Just curious is there a fundamental flaw in my logic or my block design?


2. In a radiator type water cooled setup, would a heatsink on TOP of a water block be of an benifit, or would it act as a heater and push ambient heat into the block as opposed to pulling heat and putting it in the air (assuming proper air circulation in the case)?


3. I have read the thread on why mini-fridges won't work and it makes sense to me. However, that seems to address the issue of using the fridge itself as a cooling tank. What about stealing the compressor system and NOT using the box, replacing the expanssion side of it with copper tubing and splicing in a block of similar design as described above. If my understanding is correct, this would essentially be a VERY crappy phase change system. Could something like this work if you used a full sized freezer or AC unit or would it still be too much load for an effective cooling system on a PC?


4. I had thought of going hybrid and came up with an idea as a cross between the two above, which I later found out is actually a water chiller. Would using a method as described in question 3 but without a block and with the copper tubing wrapped as a coil and placed in a rather large reservoir (5-10 gallons) of a regular water chilled system be an effective way to bring down the temps below ambient? I'm thinking of leaving the unit always on, and then running the actual water system only when the computer is running (up to 12 hours a day). The large insulated reservoir would take a while to hit it's balanced temperature when the computer was off, but once there the compressor unit would only kick on and off every now and then to keep it at that temperature thus reducing the load on it considerably. Also the large reservoir would require large amounts of heat to increase it's temperature drastically so the compressor would not be under too much load to keep it within reasonable termpatures. Of course for the same reason if it did get heated up, the compressor would be put under that much more load as it had a large reservoir to cool as opposed to a small system.


5. Would the system described in question 4 benefit from a radiator at the end of it's cycle before being dumped into the large reservoir, or would that simply heat the line with ambient heat? Would it be a good idea to have such a radiator with a fan on a temperature setting so it only kicks on after extended periods of use or would it not really be necessary?


6. If I were running a LOT of blocks (CPU, RAM, mosfets, northbridge, multiple GPUs and their ram, etc) which system do you think would be more effective for 24/365 opperation: The system described in question 4 or a straight distilled water system on a smaller reservoir (say 1-2 gallons) using multiple radiators that are stationed with each one between sets of several blocks? In case that was confusing, something like this:
Pump - > Radiator 1 -> CPU -> mosfets -> Radiator 2 -> RAM -> Northbridge -> Radiator 3 -> GPU 1 -> GPU 2 -> Radiator 4 -> Reservoir


7. Would a water system as described in question 6 be hurting itself by drawing too much heat from items that could be otherwise air cooled with fans and heatsinks? Would it be better to focus on the major components, like the CPU, RAM, and GPU(s) and leave the others to more passive methods?


8. Lastly, what is the general concensus on using TIG welding coolant in place of water? If I use only copper in my system corrosion shouldn't be an issue, and the TIG coolant is specifically non-conductive (it has to be), has better heat retention than water, and has an anti-clogging oil agent to keep small pumps clean and lubricated. Any thoughts on the subject?


Thank you in advance for any help you guys can give and sorry for bombarding the topic with questions. Also, I apologize if any of these issues are addressed elsewhere.

-derracs

 

[thorilan]

1.the idea is to get water to flow past the hot spots pretty fast with low resistance so if you look up most of todays common blocks you will be able to figure out how this is achieved. it comes down to the total sums of bends and turns adding resistance.

2.no

3. yes a full sized fridge is what you would need for phase change but keep in mind you will have to head to that forum as there is a Lot to learn about phase change and the inherant issues with that style of cooling

4.bad idea for the initial reasons you will find out about in the phase change section ( think condensation) AND in-efficiency will actually cost you a LOT in terms of power bill and space.

5.not being rude but i would forget this one until you are a water cooling guru

6. IF you dont plan on ever turning off your comp then either a complete water cooled solution OR a phase change system would work but dont think about mixing them .
the ammount of blocks you are thinking about isnt generally doable because you have to custom build the blocks to the specific mobo and for our purposes its generally overkill unless you are going for an ultim4te style system

7. go to the stickies and understand #6

 

[m0r7if3r]

Firstly, to ocf and the watercooling board

1. If I go water, I'm thinking of making my own blocks of solid copper with threaded barbs brazed in. My initial thought was to mill a channel with a ball endmill in a spiraling type pattern with an inlet at the center and the outlet towards the edge, a thin 1/32 or thinner plate would be brazed of soldered to closer it air tight with a quick pass by a fly cutter and hand polished to ensure a flat surface. My logic was this would not cause a reduction in the line and pressure/flow would stay the same, the coolest coolant would be delivered to the hotest spot first, it would maximize time over the chip without allowing stagnant areas (such as corners if it were square in design), and would be thin walled to allow maximum heat transfer and minimal heat retention by the block itself.

However, I have seen a lot of designs that look like flowers, or mazes with dead ends, etc that seem contrary to my own logic. I also read a thread, I think though the actual extreme cooling FAQ, that stated something along the lines of you want flow rate in the lines to be high but in the blocks to be low. Just curious is there a fundamental flaw in my logic or my block design?

most of the blocks today use micropins/microchannels to remove the heat. The idea is that you have the greatest SURFACE AREA contact, not the greatest time, if you can increase the time, that might help, but surface area is the primary thing, since heat dissipation is w/m^2

2. In a radiator type water cooled setup, would a heatsink on TOP of a water block be of an benifit, or would it act as a heater and push ambient heat into the block as opposed to pulling heat and putting it in the air (assuming proper air circulation in the case)?

I have often wondered this myself, whether a hybrid block would work...I can't think of a reason it wouldn't, the cpu blocks get pretty hot on their own, and it probably wouldn't add but a few C of dissipation, but I'm an every last ounce of performance sorta guy...

3. I have read the thread on why mini-fridges won't work and it makes sense to me. However, that seems to address the issue of using the fridge itself as a cooling tank. What about stealing the compressor system and NOT using the box, replacing the expanssion side of it with copper tubing and splicing in a block of similar design as described above. If my understanding is correct, this would essentially be a VERY crappy phase change system. Could something like this work if you used a full sized freezer or AC unit or would it still be too much load for an effective cooling system on a PC?

check out some of the materials on making phase coolers, a couple people do it, and it sounds like you have the background for it.

4. I had thought of going hybrid and came up with an idea as a cross between the two above, which I later found out is actually a water chiller. Would using a method as described in question 3 but without a block and with the copper tubing wrapped as a coil and placed in a rather large reservoir (5-10 gallons) of a regular water chilled system be an effective way to bring down the temps below ambient? I'm thinking of leaving the unit always on, and then running the actual water system only when the computer is running (up to 12 hours a day). The large insulated reservoir would take a while to hit it's balanced temperature when the computer was off, but once there the compressor unit would only kick on and off every now and then to keep it at that temperature thus reducing the load on it considerably. Also the large reservoir would require large amounts of heat to increase it's temperature drastically so the compressor would not be under too much load to keep it within reasonable termpatures. Of course for the same reason if it did get heated up, the compressor would be put under that much more load as it had a large reservoir to cool as opposed to a small system.

imo, waterchillers aren't worth it, your temps don't get low enough for the risk of condensation to be worth the work

5. Would the system described in question 4 benefit from a radiator at the end of it's cycle before being dumped into the large reservoir, or would that simply heat the line with ambient heat? Would it be a good idea to have such a radiator with a fan on a temperature setting so it only kicks on after extended periods of use or would it not really be necessary?

probably wouldn't benefit from the rad, temps in the loop equalize, think of it as a body of water, not as a bunch of little bits of water that carry heat.

6. If I were running a LOT of blocks (CPU, RAM, mosfets, northbridge, multiple GPUs and their ram, etc) which system do you think would be more effective for 24/365 opperation: The system described in question 4 or a straight distilled water system on a smaller reservoir (say 1-2 gallons) using multiple radiators that are stationed with each one between sets of several blocks? In case that was confusing, something like this:
Pump - > Radiator 1 -> CPU -> mosfets -> Radiator 2 -> RAM -> Northbridge -> Radiator 3 -> GPU 1 -> GPU 2 -> Radiator 4 -> Reservoir

once again, temps equalize, so as long as your pump is right after your reservoir, your loop order matters none. That said, you'll need a lot of pumpage for that loop. You also should think about dropping the mobo and definitely drop the ram, ram cooling is entirely unnecessary and just adds restriction to the loop...mobo cooling about the same, though with less restriction.

7. Would a water system as described in question 6 be hurting itself by drawing too much heat from items that could be otherwise air cooled with fans and heatsinks? Would it be better to focus on the major components, like the CPU, RAM, and GPU(s) and leave the others to more passive methods?

It would be hurt by the extra heat, but if you've got the raddage for it, you can keep your cpu temps where you want them. Also, ram is not a major component (not for watercooling at least, high heatloads components are major).

8. Lastly, what is the general concensus on using TIG welding coolant in place of water? If I use only copper in my system corrosion shouldn't be an issue, and the TIG coolant is specifically non-conductive (it has to be), has better heat retention than water, and has an anti-clogging oil agent to keep small pumps clean and lubricated. Any thoughts on the subject?

Never even heard of the stuff. If you wanna try it (and potentially risk your gear), go for it...but everything will become conductive over time in contact with copper (it ionizes). This stuff sounds like good stuff, and i'd be interested to see a comparison between this and the tried and true distilled+ptnuke

 

[derracs]

Thanks for the responses.

Ok, so it sounds like a submerged water chiller is a bad idea. I REALLY like the idea of a phase change system but I'm a bit reluctant to go from passive air cooling to a full on phase system simply on the whim of 'I want to'. That said, it sounds like a straight water system is the way to go.

I think I am on the right track with my waterblocks design, however, they need to be modified. Maybe something along the lines of running 1/2" hose but instead of using a large channel in the block, use several parallel channels. Hell I could even make the channels extra wide, and mill mating walls in the plate floor so that it a cross section would look like interlocking teeth. All I'd have to do then is mill them deep enough so that the total surface area of open space was equal to the surface area of my inlet(s) and flow rate should not be effected.

I had a rather interesting idea earlier today. What if instead of using the head of a pump to try and force water through a system, we used instead gravity to PULL water through the system?

Imagine this; a collection reservoir with condensate pump at the base, the PC resting on top of it, with a resrvoir on top of the PC. The condensate pump floods the top reservoir with coolant. The top reservoir has as many outlets in the bottom as you would like. Hose travels from each outlet in the top reservoir, downward via gravity through any blocks that are in it's path. Afterwards, the hoses exit and drain into the condensate pump's reservoir and the process continues.

With this type of design, you would not have to worry about pressure so much because so long as you're pump could handle the head to deliver water ~4-5 feet up to the drain reservoir then it wouldn't matter if you had 1 block or 50 blocks. You might also be able to pressurize the system with a vaccum pump or even compressor air in the top tank to really move some water. The hardest part would really be figuring out the number of drains and/or drain sizes to find an equilibrium between water displaced and water drained.

 

[m0r7if3r]

That won't work because of how pressure changes in tubing of varying diameters, the instant you put anything that splits the flow (ie, microchannel) or restricts it, the pressure goes up, that pressure is then translated back to the pump, so with each block that had any sort of function you'd lose flow, so each block would perform worse...if you had A LOT of tubing for the water to accelerate through it'd work, but then you're back to needing a high head pump.

 

[Owenator]

Welcome to the forums! It sounds like you left watercooling about the time I started.

1) That block design was cutting edge at the time but now microchannels are the way to go. Contrary to what most people think microchannels are not about surface area but the fluid to block boundary layer. Basically the small channel makes a very fast flow that shears away the boundary layer and increases convective heat trasfer. It's kind of a complicated fluid mechanics concept. I can explain it if you like or you can google on "fluid boundary later".

2) As Thorilian said, generally not any gains from this. Water is just so much better the heat sink doesn't do much if anything. Once core concept to emphasize is that heat only flows from hot to cold so don't worrk about heating the fluid unless it is cooler thatn ambient which is not the case unless you do have a chiller of some form.

3) The problem with minifridges is that they are designed for a pretty small heat load. Nothing like a modern PC. They are also set up to cool gradually and the "top off" the cooling occasionally not run constantly.

4) You can make a chiller, they are used in labs all the time. But as was said they are not cheap and use a lot of energy. If you are purely into efficiency they are not as efficient as going water or phase completely. But they are real and can be made if you are so inclined.

5) Again heat goes from hot to cold so ambient won't heat the water. You can combine cooling concepts but like I said it get's expensive and the gains are small. Think of it like drag racing. The faster you want to go the cost goes up exponentially.

6) The fluid will all come to a steady state temp in a 24x7x365 system. Order of components is not as important as the total capacity. More rads are fine but each one adds flow resistance which may mean more pump. In general more components bends etc leads to more pump needed. Try to size things with the minimum parts to reduce pump head needed.

7) Depending on the sizing you can cool it all but with the heat output of modern video cards we tend to use separate loops for CPU and GPU. I am struggling with this myself. I added the GPU's ot the CPU loop and they are heating the CPU. I'll be separating them today or tomorrow.

8) This one I have no idea. Sounds intriguing though. Do you have more info/links on it?

 

[QuietIce]

1) That block design was cutting edge at the time but now microchannels are the way to go. Contrary to what most people think microchannels are not about surface area but the fluid to block boundary layer. Basically the small channel makes a very fast flow that shears away the boundary layer and increases convective heat trasfer. It's kind of a complicated fluid mechanics concept. I can explain it if you like or you can google on "fluid boundary later".

And I always thought it was tied to impingement (not surface area), which I'm sure isn't an ME term. In ME terms maybe it has something to do with making sure we have turbulent flow instead of laminar flow ...?

 

[Owenator]

And I always thought it was tied to impingement (not surface area), which I'm sure isn't an ME term. In ME terms maybe it has something to do with making sure we have turbulent flow instead of laminar flow ...?

No, you are right. Impingement is the proper ME term. It's the term for the fluid hitting the surface. Impingement is part of the effect of the high speed in the microchanels. The fluid essentially hits the surface hard, breaks down the boundary layer and then quickly moves away. There is also literature on "spray" jet impingement cooling that uses jets of fluid blasting at a flat surface. It's a similar effect but different application in microchannels.

EDIT: Yes, I am a bit of a cooling geek. And I do have the education to understand this stuff too! But my experience is limited to my own tests.

 

[GIXXERGUY6]

If you have plans on upgrading to the new quad series and plan on overclocking it and getting nuts then phase isn't for you either(so I have been told)

 

[Owenator]

derracs - I did some research on water cooled tig torches and it looks interesting. Can you get ahold of a water cooler from work just for testing? I wonder what's inside them -any ideas?

 

[QuietIce]

No, you are right. Impingement is the proper ME term. It's the term for the fluid hitting the surface. Impingement is part of the effect of the high speed in the microchanels. The fluid essentially hits the surface hard, breaks down the boundary layer and then quickly moves away. There is also literature on "spray" jet impingement cooling that uses jets of fluid blasting at a flat surface. It's a similar effect but different application in microchannels.

EDIT: Yes, I am a bit of a cooling geek. And I do have the education to understand this stuff too! But my experience is limited to my own tests.

That would be the same effect (taken to the extreme) the Storm water block uses? I'm still running two of those.


BTW: Thanks for the reading material - good stuff!

 

[Owenator]

That would be the same effect (taken to the extreme) the Storm water block uses? I'm still running two of those ...

Yep! That's exactly what the storm does. IIRC it was Cathar who first did this and he "licensed" the idea to Swiftech. Smart man that Cathar!

 

[derracs]

I'm actually not a welder by trade, though I do some here and there for fun, so I can't attest for the true potential of the TIG coolant. I did however run the idea by the head welder at work and I got a good response. I think at first he misunderstood my question because he made the statment along the lines of water being the best for cooling, and if I was going to high heat I could add some anti-freeze for the heat. However, once I mentioned computers and electronics his tone changed and it went to an ABSOLUTELY go with TIG coolant. He said in his experience, using something like anti-freeze works but it kills pumps quickly and he's had problems with sludge forming in the lines from over heating, causing pressure spikes, and then either hose ruptures or being pulled from their connections. He said he has NEVER experienced these problems with the TIG coolant, which as he understood was a Glycol based coolant like anti-freeze but with a smaller micro-structure (can't remember his exact wording but that conveys the thought) and it had the added benefit of a self lubricating oil in it to prolong pump life. I'll see if I can maybe steal a label off the bottle next time we get some.

I did also have the opportunity to tear apart our small 2 gallon pump and helped make the mounts and plates to put it onto a 55 gallon drum for a Plasma Transfer Arc system. In all honesty, it looked like a regular run of the mill pump system with a small collection system at the bottom and a few beels and whistles, like an overflow valve, circulation unit (I presume to prevent settling of anything in the lines), etc. It did run rather loud for it's size, but when we got it hooked up to the large drum it wen't near silent so that may be more revirbiration from the shallow sheet metal tank than the pump itself. Unfortunately, now that it's hard mounted in our PTA system, I couldn't take it apart or even move it if I wanted. Of course, that won't stop me from taking off a cover or two to get a better idea...

I'm liking this fluid boundary layer concept. It's VERY similar to a lot of the more modern concepts behind firearm suppressors - like K baffles and especially the Jet Z design. I hadn't really thought about using my knowledge of firearm suppressors and adapting it to a computer water block. It's essentially the same concept; a fluid passing through an enclosure where you are trying to rapidly dissipate heat and control fluid current and pathways. Interesting... interesting... I must ponder on this more.

I had actually planned on going with one of the hex cores just for the sheer WTH aspect of it, but in my reading I haven't heard very many good things about them in regards to customizing and it seems like more people point towards the Bloomfields for over clocking. Now I'm hearing the new cores aren't good choices for phase change at all. What's the deal? Anyone have any details on this?

 

[m0r7if3r]

man, owen, i just read all that...good call, why didn't you help out on the ocf custom cooler?

 

[Owenator]

I'm actually not a welder by trade, though I do some here and there for fun, so I can't attest for the true potential of the TIG coolant. I did however run the idea by the head welder at work and I got a good response. I think at first he misunderstood my question because he made the statment along the lines of water being the best for cooling, and if I was going to high heat I could add some anti-freeze for the heat. However, once I mentioned computers and electronics his tone changed and it went to an ABSOLUTELY go with TIG coolant. He said in his experience, using something like anti-freeze works but it kills pumps quickly and he's had problems with sludge forming in the lines from over heating, causing pressure spikes, and then either hose ruptures or being pulled from their connections. He said he has NEVER experienced these problems with the TIG coolant, which as he understood was a Glycol based coolant like anti-freeze but with a smaller micro-structure (can't remember his exact wording but that conveys the thought) and it had the added benefit of a self lubricating oil in it to prolong pump life. I'll see if I can maybe steal a label off the bottle next time we get some.

I did also have the opportunity to tear apart our small 2 gallon pump and helped make the mounts and plates to put it onto a 55 gallon drum for a Plasma Transfer Arc system. In all honesty, it looked like a regular run of the mill pump system with a small collection system at the bottom and a few beels and whistles, like an overflow valve, circulation unit (I presume to prevent settling of anything in the lines), etc. It did run rather loud for it's size, but when we got it hooked up to the large drum it wen't near silent so that may be more revirbiration from the shallow sheet metal tank than the pump itself. Unfortunately, now that it's hard mounted in our PTA system, I couldn't take it apart or even move it if I wanted. Of course, that won't stop me from taking off a cover or two to get a better idea...

I'm liking this fluid boundary layer concept. It's VERY similar to a lot of the more modern concepts behind firearm suppressors - like K baffles and especially the Jet Z design. I hadn't really thought about using my knowledge of firearm suppressors and adapting it to a computer water block. It's essentially the same concept; a fluid passing through an enclosure where you are trying to rapidly dissipate heat and control fluid current and pathways. Interesting... interesting... I must ponder on this more.

I had actually planned on going with one of the hex cores just for the sheer WTH aspect of it, but in my reading I haven't heard very many good things about them in regards to customizing and it seems like more people point towards the Bloomfields for over clocking. Now I'm hearing the new cores aren't good choices for phase change at all. What's the deal? Anyone have any details on this?

Great research on TIG coolant! Over the years I've seen lots of different cooling ideas slowliy incorporated into our hobby, it's great! You are right that fluid mechanics sounds like water but it applies to all kinds of things that just behave as a fluid. Air for example can be treated as an incompressible fluid in a lot of cases. It's usually not until you get into supersonic speeds that the compressible stuff comes into play IIRC.

Not sure about the phase change cooling I'd suggest posting that/those questions in there to get straight to the experts.

man, owen, i just read all that...good call, why didn't you help out on the ocf custom cooler?

I missed that one, sorry. There's a lot going on here and I am not as good at keeping up as I used to be. Do you have a specific thread you want me to look at?

 

[m0r7if3r]

nah, it was a tom's contest just recently sponsored by zalman's...our design was a good cooler, from the results, it seemed like zalmans was looking for gimicky, not performance.

back on topic, I'm REALLY fascinated to see how this TIG stuff does...

 

[Owenator]

nah, it was a tom's contest just recently sponsored by zalman's...our design was a good cooler, from the results, it seemed like zalmans was looking for gimicky, not performance.

back on topic, I'm REALLY fascinated to see how this TIG stuff does...

The TIG fluid does look interesting. It looks like it is used to cool a high heat load. But I'm nit sure how hot those welders get but they look hot. IIRC TIG stands for Tungsten Inert Gas. According to Wiki pure Tungsten's melting point is 3,422 °C, 6,192 °F. The welding rod is most likely an alloy but still, wow! That's hot!

EDIT: I looked some more and it shows a rod not a wire like a MIG welder. Plus the rod is what melts not the Tungsten so probably a lot lower temps than melting Tungsten. Tungsten is non-consumable unlike a welding rod. But I'm no welding expert.

 

[m0r7if3r]

the question is, will the tig coolant be efficient at a temp much lower than it was designed for? Basically, is it's behavior linear or otherwise?

EDITosted and immediately realized linear is not what I want...it must just outperform water at every point on a graph from 0-~60c (where pumps are rated) to be a fully viable replacement for everyone...it might end up being a viable replacement for worse performing loops and not great ones or vice versa.