Series VS parallel: tri-fire watercooling

[vinister]

I am putting together a pretty complex loop. Its not that big, it is just tricky because of the single spacing between the video cards. I am using stubby barbs between the cards, and the barbs almost touch each other. A very short, straight peice of tubing will connect the cards to each other.

3 x EK 3870 full coverage blocks
1 x MPC 655
1 x Swiftec double rad w/res
all on a DFI 790x board.

So, there are at least 2 possible ways to plumb this setup.

in parallel:

in out
_|__|_______card 1
| |
_|__|_______card 2
| |
_|__|_______card 3

... or in series:

in
_|_________card 1
|
___|_______card 2
| out
_|__|_______card 3

Both methods have been recommended to me. Parallel takes more fittings, but I have ordered the correct amount anyways. Many people warn against parallel, because more water might take one path than another. The difference in this situation is the 3 blocks are identical, so the flow should be equal in the 3 branches. Others that have tried, from what I read, are getting nice even temps.

I am worried about the series method overheating the bottom GPU, and also the reduced flow caused by 3 series blocks.

Anyone have experience with a system like this? I'm also concerned about how I am going to assemble/fill/bleed the loop. I'm worried about working on the cards while they are in the motherboard, I don't want to mess up the pci slots or anything... but its going to be hard to get all the proper tubing lengths without them being on the board, or in some kind of a jig.

Last question: If I was to add a CPU block and a single rad for it, should I run a seperate pump as well? Or could the 655 push through both loops?

[MpG]

Just to make sure from your posts, these are the two options you're considering, right?

Parallel Tri-Fire

Series Tri-Fire


The parallel setup works in with two cards, but I don't know for certain if the middle card gets an equal share in a triple setup. Theoretically, it should, since you're using identical blocks.

An MCP-655 (at the high setting) will push through either loop without too much trouble, and the water shouldn't pick up more than a degree or two through each block.

If you want, it'll probably push through most CPU blocks as well, while still keeping flow above the 1.0gpm mark, but you'll want to make sure the CPU is the first block in the loop. And your radiator is definitely going to start to choke a little unless you're giving it some good, strong fans. An extra single rad wouldn't definitely be a good idea in that case.

[vinister]

Ya I'm going to go with the top diagram you have posted, except the in and out are both at the top, since I have no room below the bottom card for an outlet. I'm just waiting for all the stubby fittings to arrive. I wish it was easier to try both methods, but I have a feeling that once I get it going I'll be leaving it alone.

[MpG]

If you do a parallel setup with the outlet at the top, you WILL get a flow imbalance, because there will be more resistance for the water moving through the lower cards.

The exact degree is debatable, and may not matter too much in the end - once you've got the card on water, a few extra degrees rarely means anything for overclocking. I don't know how flow divides across paths with different resistances (I don't think it works like electricity, unfortunately), but if you see vastly different temps with the final product, that'll be your reason why.

[vinister]

I really don't think the flow direction will make any difference, if anything more flow would get to the bottom card because of the water's inertia, and it not wanting turn 90 degrees when it could just go straight.

We'll see how it works.

So far, finally got the mobo and phenom setup and running perfectly with 1066. I'll post pics here of the final setup!