128 point high temperature cooling design.
 

I have designed a cooling system for a high density cluster server. It provides cooling for 128 points in a volume of .39 cubic meters, or about the size of a 44 gallon drum, to give you an idea.

I have no idea were to go from here. If you have any suggestions, I would be glad to hear them. The included image is a comparison to a 64 node Beowulf cluster built from standard ATX cases.

Re: 128 point high temperature cooling design.
 

wait. I don't get what you're trying to do. Would you mind explaining in greater detail please?

Re: 128 point high temperature cooling design.
 

What he said. Are you looking to manufacture the design, sell it or what?

Re: 128 point high temperature cooling design.
 

Take 64 dual processor rackmount units, and replace them with 64 SFF boxes, arrange them in a drum and run the cold air in from the middle.

Not sure how you got dual core mobos in an SFF though; they make those?

This is simply an optimization of air flow.

Correct me if I'm wrong.

Re: 128 point high temperature cooling design.
 

You are wrong. :laugh2: He wants to liquid cool this beast according to an e-mail he sent. He has a design it seems and want info on how to go about getting it into production as he doesn't have the funding to do so.

I honestly have no idea were to start with getting this into production. The main thing is making people think it is a must have and worth buying or no one in their right mind will consider putting up cash for production.

Re: 128 point high temperature cooling design.
 

All server rooms are designed for rectangal things, not round ones.
Why round?
Motherboards are rectangals, HDDs are rectangals, expansion cards are rectangals.
The picture shows lots of wasted space. Why do you need to get to the front of any of the nodes?

Re: 128 point high temperature cooling design.
 

Server rooms are also designed for front to back cooling with hot and cold aisles - vertical heat stacks are problematic.

Re: 128 point high temperature cooling design.
 

I work as tech support and we have something around 200 servers in a server room. They range from mainframe, passing through windows, linux, unix up to Open VMS. One thing that I can tell from personal experience is that the last thing you want to worry about is if the cooling system in a server box is running or not.

It's cool (pun intended) to run a watercooling setup on a machine at home, but it's far from practical to even consider the same on a multi-server environment.
Too many variables and too many failure points to consider.
I would never buy a server with a watercooling kit unless it came with the box and it was fully supported by the vendor.
Another thing is that it's a tight fit inside a rack server, specially the new Intel/AMD boxes with 2 or 4 Us, when you pop them open you can see there's no space available.

Anyway good luck with your project.
_________
TaTs

Re: 128 point high temperature cooling design.
 

already been done, there are several to choose from, just google them.
not of those can cool as effectively as our current projeect though :P

Re: 128 point high temperature cooling design.
 

The main failing of liquid cooling is "what's next?". I personally like Suns's approach fixing the problem on the electronics side with more efficient cooler running processors and other electronic parts.

That will the be the key to the future, not liquid cooling. Once liquid cooling is tapped out then the next options are ridiculously expensive.

In my opinion the need for water cooling is taking a step back wards in progress. However you have to deal with the hand that is dealt. Server manufactures see some serious $$$ to be made in liquid cooling aswell. Look at Mac for example. Charging $1,000 for a replacment liquid cooling modual..... That thing costs less than $100 to make...

Anyway carry on.