[Gooserider]

Also, since you don't speak from experience, I have returned your disrespectful sarcasm with some more.

First off, I would like to start by apologizing for coming accross as sarcastic. That was not my intention, as I don't normally go out of my way to upset others. I will admit to a sometimes caustic sense of humour, sorry if it rubbed you the wrong way.

You did NOT make clear in your original post that you intended the upside down operation for leak testing only. I thought that your suggestion was for permanent operation, which seemed to me like a very bad idea for the reasons I mentioned. If you are suggesting just flipping the system for leak testing, then I have far fewer problems with it. My biggest concern would be whether the plumbing was designed for upside down operation (especially if the plumbing was all in the case). For instance a res that fed the pump from the bottom normally would be feeding from the top when flipped; if the res wasn't totally full, then the pump would be sucking air which isn't good... Also was all the hardware secured enough to handle a flip. Some builders put pumps, HDD's, and other such things in the case setting on a layer of foam without much securing - if the case were flipped they might dangle and do damage.

If those concerns (which I think you'd agree make sense) aren't a problem, then the flip isn't a bad idea. However, all reports suggest that for a non-conductive coolant, a spill onto a non-powered board is not a big issue as long as it's dried completely.

My objections were mostly targeted at the idea of running a flipped PC on a continuing basis.

1. I think some case fans solve the heatbuildup problem. You test it with a fully watercooled pc (CPU,GPU,North) and then get back to me. I want hard data, not your opinion.

I don't have hard data, however I do know the conventional heat flows in a case, and that it is FAR better from a noise standpoint to design a system to avoid thermal issues to begin with rather than trying to solve them by adding fans. I also haven't seen many setups that address the layer of stagnant air between the board and the case. On a vertical layout, that air gets changed by convection flow at least. In an upside down case it would be trapped unless extra work was done to remove it.

2a. Everything is NOT below the mobo. Water will go straight down in the case of drips, and there is nothing directly over my sockets except for the case lid.

I'll buy that, especially since I haven't seen your case, but if there is any accumulation can it run to somewheres that it could be a problem? Is ALL the plumbing only above the lid, including the pump, rad, hoses, etc.? (a drip might run down a hose and along it before falling...) I also wasn't replying just to you, but to anyone else that might read the thread and consider your idea as well.

I am well aware of gravity. I have not only thought about mounting things upside down, but I have also done it.

I have seldom done it long term in any system that I have designed, but I have had to work on many a setup designed by others that had things mounted upside down. One of the things I've learned from long experience is that upside down socketed parts are a frequent cause of problems. A standard PC card is a particularly bad potential problem because the only things holding it in are the spring tension of the contacts and the single screw in the mounting bracket. Cards (especially the old 8 and 16 bit ISA cards) have a tendency to come loose even in regularly positioned systems. One of the things you'll see long time techs do as a first step when setting up a new box is take the cover off and push on all the boards to make sure the seating is still tight. It has always amazed me that so few people do this considering the number of times I've found problems that way before I've even powered up a system. The tendency of many people to hang big heavy heatsinks and waterblocks off their video cars doesn't help either.

I see in your signature you WANT to have a watercooled system, so I don't respect your opinion.

Yup, and I'm using 20+ years experience in the PC business, (including 5 years working in systems for the telecom business where 99.999% uptime is a contractual requirement) to design and build my box, mostly from scratch, including component selection, machining my own blocks, etc. Right now I WANT to have a system. When I'm done, I WILL have one that is as rock solidly reliable and bullet proof as I can make it. I am also quite respectful of those who have gone before me in this area, and who share their knowledge. I started this poll as the result of a discussion of failure detection and handling where I realized there wasn't a real good accounting of the actual problems people had. My goal was to find out what sort of problems were found in the real world so that I could focus my design efforts back on preventing them. I respect you as a person, regardless of what I may think of your ideas (which as I said weren't all that bad, just perhaps a bit problematic). While I would appreciate your respect in return, I don't require it, and won't loose any sleep if I don't get it. I might ask you to consider that if I thought you were a fool I wouldn't be spending the time it takes to make this reply...

4. Put your pump-res-rad in a separate case ABOVE the PC, and the CPU blocks are no longer the highest point. Case closed (pun intended)

Not so fast... You are correct in an absolute sense, but there is still a problem. I make the assumption that your blocks follow the fairly standard design of barbs sticking out of the side opposite the die surface. (I'd normally say the 'top' but since this discussion is dealing with an upside down case, I wanted to be clearer) If so, then the hoses coming to and from your pump/res box HAVE to go at least some distance below the block, make a bend and then come back up in order to make the connection. This means that the die side of the block is still the high point in that section of the loop. Any bubbles going into that section will stay trapped in the block unless the flow pressure and volume is high enough to push them down the exit hose and around the bend going back up.

4a. Will the heat transfer suffer because the block's weight and the water within pulls it away from the die? I have definitely thought about the mounting system, and my waterblocks are going nowhere.

While that is a possibility (especially long term) it was not the primary issue I was concerned about - I assume that a well designed mount can handle being upside down w/o major problems.

My main concern was bubbles getting caught in the block. (see above) since the block is upside down, any trapped bubbles will form a layer of air between the block's die contact surface and the water. This would be a major barrier to heat transfer. Think about the following ascii diagrams, and you'll see why...

Upside down:
 __die_surface_______
++++B+L+O+C+K+++++++++
+****A*I*R***********+
+~~~~W~A~T~E~R~~~~~~~+
++ I +B+L+O+C+K++ O ++
 + N +          + U + 
 +   +          + T +
=================================
Right side up:
 + I +          + O +
 + N +          + U + 
++   +B+L+O+C+K++ T ++
+****A*I*R***********+
+~~~~W~A~T~E~R~~~~~~~+  
++++B+L+O+C+K+++++++++
 __die_surface_______

As should be obvious, in the top case, the air isn't going anywhere unless either the system is tipped, or there is enough flow volume to force the air out. OTOH, in the bottom case, the air will naturally exit.

However if one assumes that the air does stay in the block for some reason, the bottom situation is still better, as the coolant which has better heat capacity is in contact with the cooling surface of the block. In the top case the air acts as a high thermal resistance insulator and will block most direct heat transfer to the coolant. Instead the heat must travel across the block and up the sides in order to reach the coolant. This increased thermal resistance could easily be enough to fry the chip.

Hope this clarifies things, and again, I apologize for the unintended upset.

Respectfully,
Gooserider