Got a cool design?

[Tens0r]

Hey, havent checked this place in ages . lost the bug a while ago building blocks in highschool. Im now in my final year of mechanical engineering and since my project is in the field of computational fluid dynamics and specifically relating to heat transfer i though id pop on in and see if anyone had some funky arse designs that they would like a quick benchmark against a current design that is used.

It would require you to send me a CAD model of what you wanted to analyze. probably another one of a standardish design to benchmark your new design against.

depending on complexity might take upto a week to get results out , but i have some spare computeres here to work with now so no prob.

cheers!

[bobo5195]

Direct die cooling. Yes i know its been done but

• Use very small laser etched holes etched in plate - 100 microns or so
• Small holes mean very close to the block - 0.3 mm
• Small holes mean lots of holes about 0.3mm apart
• Need for a big ass pump to up performance
• exit points all the way around the side to prevent flow restriction
• Direct die with trying to optimise for hot points so stagger the holes.

Would suggest that if you mounted thermocromatics on the cpu surface with a clear top you could see where the hot areas were and optimise accordingly. Or just optimise for hte thermal sensor and cheat. You will half your C/W this way.

[ben333]

Didn't JD conclude that direct die and direct IHS cooling didn't beat using a block?

[bobo5195]

Originally Posted by ben333

Didn't JD conclude that direct die and direct IHS cooling didn't beat using a block?

he did. I disagree with his results. You need smaller holes with a nicer spread.

[jaydee]

Originally Posted by ben333

Didn't JD conclude that direct die and direct IHS cooling didn't beat using a block?

More so on a practical/enthusiast level. What bobo describes here will never be seen in the enthusiast market. I see no point in arguing over things that will never happen. Direct die cooling, even if effective, would require the CPU itself to be designed for it to be properly implemented anyway.

Also with todays CPU's the minimal gains you would get over a good watercooled or even air cooled system would be minimal. CPU's these days don't react to cooling as well as in the past. Not to mention they are being designed for lower power consumption. We will be running passive heat sinks again sooner or later.

[jaydee]

Here is a simple idea if someone wants to waste a perfectly good Swiftech Storm... First test it thoroughly as is. Then finish drilling the holes through the base of the block so the water hits the top of the IHS (I can hear tiny drill bits breaking). Figure out a way to seal it. Test it again and see what happens.....

[billbartuska]

Originally Posted by jaydee

Here is a simple idea if someone wants to waste a perfectly good Swiftech Storm... First test it thoroughly as is. Then finish drilling the holes through the base of the block so the water hits the top of the IHS (I can hear tiny drill bits breaking). Figure out a way to seal it. Test it again and see what happens.....

OOOHHHH! That's interesting!

[bobo5195]

Sorry for the general tardiness of posting.

I would disagree with the assertion that CPUS themselves have to designed for it. Heat spreaders already add a nice layer of packaging. Agreed it is risky to an extent but no means impossible.

The central question is if performance is there. With a big pump I believe it is. (As it happens me and my supervisor on my current project are looking at this, he is betting on Lamina flow heat exchangers. Note that he considers turbulence to be wasteful in heat sink designs from all his experience)

Swifttech storm idea is good, will perform crapish though as the storm is laid out for confined jet impingement not what we are talking about and some of the other geometry is not what you would want.

A flat plate with laser or similar holes mounted less than 1mm off the surface and sized so it won’t deflect is perfect. Thinking holes of the order 0.1mm or less.

Cheap to make in small to medium quantise me thinks. Besides has the added bling of having to know what paterns suits your CPU better.

* would add this is all drunken flu medicine inspired ideas. Based on the idea that 4 half diameter jets are better at cooling with lower flow than 1 and optimum stand off is about 5 diameters from the surface is memory serves. The question then becomes how small can we go.

Give it 5 years (probably a bit more) and watercooling will be popular enough to do this as our CPU/GPU chips will be straining against barriers.

[jaydee]

Originally Posted by bobo5195

Sorry for the general tardiness of posting.

I would disagree with the assertion that CPUS themselves have to designed for it. Heat spreaders already add a nice layer of packaging. Agreed it is risky to an extent but no means impossible.

I have done it twice direct die and direct IHS so it is sure possible. But this why companies like Swiftech do not use it. How can you seal it properly with out the possibility of the tubing getting tugged on and breaking the seal? Also when you take it off how do you keep water off the parts? The block will keep water in it no matter how hard you try to drain it. These are things keeping it from mainstream and why I suggest the CPU needs to be built around it. If you can mount the cooler directly to the CPU itself instead of the socket (like clamping it right to the IHS or something) and can remove the CPU with the cooler then you are golden... Until that happens (and I suggest it never will) it will be to much of a liability for any company to do this on a general user level.

The central question is if performance is there. With a big pump I believe it is. (As it happens me and my supervisor on my current project are looking at this, he is betting on Lamina flow heat exchangers. Note that he considers turbulence to be wasteful in heat sink designs from all his experience)

Performance will not be much greater if not worst. Not enough to justify the cost. Especially after you take into count the need for special water additives to keep the water clean. With holes that small any tiny bit of corrosion, dirt or whatever will quickly clog those holes. This also means adding a filter and adding a bigger pump which will add more heat to the water than what you may have gained with the cooler itself.

The best blocks out now keep the water around 10C above water temp on a hot CPU like the 125watt variety... With a more reasonable CPU that is running in the 65watt or even the newer 45watt ones that drops to around 5C over water temp. You will not get below water temp or even within a few C no matter how well you cool it. With even the most optimal cooling system you are only going to gain a few C cooler.. What good is that few C? What are you going to gain? Nothing... Overclocks will not be much. The quiet angle will be negligible and being you need a big pump it will be loud anyway. I just don't see why it would be worth the questionable seal.

Swifttech storm idea is good, will perform crapish though as the storm is laid out for confined jet impingement not what we are talking about and some of the other geometry is not what you would want.

A flat plate with laser or similar holes mounted less than 1mm off the surface and sized so it won’t deflect is perfect. Thinking holes of the order 0.1mm or less.

I am not sure this is possible. Once the water flows with that many small holes that close together isn't it going to create back pressure and keep the holes in the middle of the grid from flowing outward? I think that is what you meant by deflect.

Cheap to make in small to medium quantise me thinks. Besides has the added bling of having to know what paterns suits your CPU better.

Who has the equipment to make such accurate laser cut holes that small and is cheap? The $50,000 laser I used to run would not come close to that small. Hell it wouldn't even cut thin metal. I know it is possible but for cheap? What is cheap?

* would add this is all drunken flu medicine inspired ideas. Based on the idea that 4 half diameter jets are better at cooling with lower flow than 1 and optimum stand off is about 5 diameters from the surface is memory serves. The question then becomes how small can we go.

Even in a drunken state you are probably well more qualified than I am for this. I am just stating my opionions based on my experience. I am by no means qualified for a real debate on the issue and in no way suggesting this not be tried. I would love to see it even if it failed.

Give it 5 years (probably a bit more) and watercooling will be popular enough to do this as our CPU/GPU chips will be straining against barriers.

Doubt it. In 5 years Notebooks will rule and pocket PCs will be the new in thing. Desktops will be the size of modern hard drive enclosures. CPU's are getting much much cooler. 5 years ago when the AMD athlons were in their prime and running hot as a fire cracker I suggested this will not continue and CPU manufactures would have to start lowering power consumption. I was laughed at but here today we have CPU's 4 times as fast and using less than half the wattage as then. Today we have CPU's just as fast using less than 1/10th the power consumption.

The "Green" movement will rule. The world is in a energy crisis and manufactures are demanded to make lower power using parts. Passive heat sinks will return and water cooling will continue to become less popular and harder to install as things get smaller.

This is the future: http://www.overclockers.com/index.ph...ls&Itemid=4259

[bobo5195]

Your more quailified than I am. I have never handled a waterblock as yet which puts me at a major disadvantage and frankly everytime I go into my PC I break something! If my day job of crazy engineering to do big changes in performance has shown me anything some practical experience is all you need (although too much makes you think nothing is possible). Someone laughing at power not being an issue in CPU design clearly knew nothing about what they were talking about.

Im rather enjoying long posting and will poster a longer reply on where I think water cooling is going. Saw this just now on realworldtech (a nice very geek forum) on a thread on if smartphones are more than enough.

http://www.realworldtech.com/forums/...93355&roomid=2

I agree that there is a huge market for Atom style processors which are good enough processing wise but the type of people who would and have watercooled are not that market and the overclockers article is a little simplistic. The future is multi core chips which are completely power limited. In such a case it is my believe that more cooling is better and hence there is a market for water cooling.