Desdicado wrote:
I pretty much use Danger Den parts exclusively and with apologies to Jeremy Burnett, from my personal experience I believe he is wrong when he says that you don't need separate cooling loops. I would've expected him to be correct, however, previous experience, even with more than enough cooling, proved him wrong.
Basic science, though, proves him correct.
Calories in and calories out. That is all that a water loop is .. and so long as your loop can shed more calories than it picks up, it is doing its job.
Consider two loops, A and B.
On A, you have X calories produced by the components it cools. On B, you have Y calories produced by the components that it cools. If A is capable of shedding up to X calories, your components won't overheat. If B is capable of shedding up to Y calories, it won't overheat.
If A and B are linked into a single loop, then the number of calories the new loop, C, can shed is equal to the sum of the calories shed by A and B. The number of calories created by the components in A and B is X+Y. If A and B were capable of shedding X and Y, respectively, the C is capable of shedding X+Y.
Summary: basic science tells us that if A and B can do their jobs, C can do it just as well. There is no advantage to multiple loops.
The advantages to a single loop are manifold. First is simplicity. A simple loop is easy to clean, repair and maintain. Second, if you want to add a component, it is much easier to add it to a single large loop. Not only is the loop simpler, but the extra calorie potential from A and B are both present in C. If A can handle 100 more calories than it produces and B can handle 150, then C has 250 calories of additional potential available. Adding a 175 calorie component is easy.
Quote:
The biggest problem I see with your configuration though is the GPU cluster itself. Having already done that configuration with only two 8800 Ultras on my previous machine I found that over the long term the second (or third) card in the serial chain never actually gets cool water to it. It is always subjected to the hot output of the first card before it even begins to try and transfer its heat to the water. And because the water is already at a higher temperature
the second card doesn't get a chance to transfer as much heat away from itself and so it just sits there and cooks. Over the long term that will kill the second, and in your case, the third card. I used/use laing D5 pumps with 1/2 in IDA tubing so I'm moving plenty of fluid there just wasn't enough temperature differential left when the water reached the second card for it to take away as much heat as was removed from the first card.
That is a function of your loop design and it would easily be solved with multiple rads in a single loop. You need to put a cooling solution between the heat sources.
Quote:
You should use a splitter to divide the the input to all of the cards so they all start out with the same input temperature and yes I know that the pressure will be reduced but with 1/2 inch IDA tubing and either a D5 pump or an additional pump in the line just prior to the splitter that won't be an issue. The problem is you guys build the thing and then it just sits there looking all shiny so you never do long term testing and see how your creations last. I am willing to bet money that if you were to use this monstor you have created as a heavy-duty gaming machine for the next year until the new Dream Machine is built you will find that one or more of those graphics card will fail because of that configuration.
Your solution might work but it is inelegant. Much better to add cooling between heat sources.
Interesting discussion.
Login
Register
FAQ
Search
© 2010 Future US, Inc. All rights reserved.