SLI WC block connection: parallel vs series

[R-TYPE]

Hi all, a noob question: I've seen watercooling link diagrams between SLI waterblock cards as having Parallel and Series links. Parallel seems to have two connections between the cards. Series has one connection between the cards.

What is the difference between the two connection diagrams?

Am planning a one loop SLI setup (ie RES>RAD>CPU>SLI>SLI>RAD>RES..) Will have a dual pump res.

Thanks

 

[andybird123]

With series, the restriction of each block is additive, affecting the flow rate of the whole loop
With parrallel it becomes divisional, so the gpu blocks as a whole present less restriction to the loop, meaning flow rate will be higher for the cpu but for each block in the parallel section the flow rate is halved (or thirded or quartered for 3+ multi card setups)

Gpu blocks tend to be big, e.g. Large thermal mass and contact area, so are much less critical of flow rate, where as cpu's need high flow

With the right pump(s) either will be fine, i started off with serial and later switched to parallel to see, and it only made about 2C difference either way so not exactly critical

 

[R-TYPE]

Thanks a lot, AndyBird. Pumps will be dual D5 vario inbuilt to XSPC resevoir.

 

[JeffyB]

I've run both on multi gpu setups, here were my thoguhts.

Serial
Second card will be about 2-3 degrees warmer
much easier to bleed

Parallel
Identical temps
More awkward to bleed as the air can go down 2 tubes rather than 1.

I preferred serial for ease, It took much longer to bleed parallel as air would not be simply pushed out the as with a serial loop. The couple of degrees difference will have no bearing given most gpu waterblocks drop temps by 30-50 degrees from a stock heatsink.

 

[R-TYPE]

Thank you for the comprehensive responses all, appreciated. May opt for Serial if bleeding is a particualr problem.

 

[R-TYPE]

Just out of a technical interest, does the parallel method present more turbulence to the flow? With serial the water is contained and only has one route it can travel (hence the better flow).

With parallel the water presumably has to divide between the routes, half the water goes straight through the first connector block and on to the second card's connector block, and half goes down into the actual water block of the first card.

This strikes me as slightly counterintuitive. I understand parallel has lower flow rate, but how is a good clean flow maintained across the first water block itself given the split and the ability of the water to effectively bypass the first card?

Does that make sense?

 

[ubersonic]

With parallel the water presumably has to divide between the routes, half the water goes straight through the first connector block and on to the second card's connector block, and half goes down into the actual water block of the first card.

This strikes me as slightly counterintuitive. I understand parallel has lower flow rate, but how is a good clean flow maintained across the first water block itself given the split and the ability of the water to effectively bypass the first card?

Does that make sense?

It can be a bit counter intuitive when you look at a parallel/serial block, when you see it done with tubing (old school) it becomes more obvious:

You see now how it splits evenly? (water follows path of least resistance and if resistance is equal...)

If you still don't get it, then here it is as a kind of circuit diagram:

 

[R-TYPE]

Thanks a lot übersonic, yeah I get the diagram, I guess it was more that I couldn't picture how flow was maintained in the first block as adequately as if it were series

Despite the concern mentioned above about drainage, I am sort of interested in parallel, as the flow reduction may offset the fact that my loop will have three rads, two GPUs, and 5-6 QuickDisconnects which overall may be more restrictive than less.

To offset this the pump is dual d5 vario and the tubing is 1/2 3/4 inch (13mm/19mm) which apparently gives marginally better flow.