waterflow through dual VGA - Serial or parallel (images inside)

[Klimpen]

I'm currently upgrading my system with both chipset and mosfet cooling blocks from Koolance.
While I'm at it I want to re-arrange the tubes for a more nice look.

My current system:


And here is a classical paint-drawing of the not-so-good sort of my desired drawing:


So do you think that using a parallel route is fine?
I found this drawings and I think the serial one I use now is the most logical one but I want to use the parallel on since it will look better.
I wonder if the flow is gonna be equal in both blocks when using a parallel setup since there is nothing to force the water to both blocks.
In a serial setup the water has "no choice" but to go through both of them. As in the parallel system the water can "go wherever it wants".

Paralell (I'm using TWO cards)


versus

Serial (I'm using TWO cards)

[Church]

With two cards, yeah, imho parallel is preferable. Flow will be slightly worse in gpu blocks, but better in overall loop with cpu. GPU full cover blocks care a bit less about flow then cpu blocks - thus the choice.

[Edited]

Agreed with Churchy

[anzial]

I had both, didn't see much difference really. Do whatever you think is more convenient.

[Klimpen]

Thank you guys. I will do the parallel since it will look best and require minimum of tubing.

[gmat]

From your drawing, you will plug both top exits, am i right ? Otherwise it doesnt make sense...

[Klimpen]

From your drawing, you will plug both top exits, am i right ? Otherwise it doesnt make sense...

Yes of course, I should have drawn that too, in my "oh so good" paint drawing.

[gmat]

And to answer your question, yes the flow will be equal, provided both blocks have the same restriction (if they are the same blocks it's easy of course).
Flow in each block will be half of your loop flow, but consider that your loop flow will increase a lot due to the lower restriction.
For example in my own loop (very big loop, 3 radiators, lots of tubing, 2 pumps, 2 GPU and CPU) i get the following figures:
1 CPU + 1 GPU alone (so, in series basically) = i get around 200lph
1 CPU + 2 GPU in parallel = around 300 lph, so 150lph in each GPU block.

[Emperor]

parallel might be better than serial cooling

in parallel, water goes into both gpu blocks, cools the gpus individually then much hotter water exits the blocks. this is compared to serial cooling where warm water from 1 block goes and cools the 2nd block.

though the pressure in the 2nd block might be slightly lower.

[Section8]

With two cards, yeah, imho parallel is preferable. Flow will be slightly worse in gpu blocks, but better in overall loop with cpu. GPU full cover blocks care a bit less about flow then cpu blocks - thus the choice.

I agree with churchy as well.

[Sideroxylon]

Check this vid and draw your own conclusions. I think I'll be sticking with serial personally.

[gmat]

The vid has it wrong. First there is air in the loop, flow will actually become a lot better when air is cleared - second the flow cannot "jump the bridge" as the video uploader said...

[defect9]

lets also not forget that as flow rate increases, so does back pressure (friction/restriction, basically). water will seek the path of least resistance (ie: least restriction). this is what equals the flow between the blocks.

[Sideroxylon]

The vid has it wrong. First there is air in the loop, flow will actually become a lot better when air is cleared - second the flow cannot "jump the bridge" as the video uploader said...

That's the whole point of the vid. In parallel, in this example, you don't have enough flow to push the air out of the blocks. You lose control over that section of your loop. You get more flow through the loop because the water isn't being forced through those blocks, it's just going the path of least resistance.

Personally, I like to know that my blocks are full of water and are performing as well as they can.

[gmat]

You get more flow through the loop because the water isn't being forced through those blocks, it's just going the path of least resistance.

But what path of least resistance ? I dont get it: the only path (and there's only one) is through each block. There is no shortcut. So alright flow is divided by 2 in each block, as i said in a previous post in this thread it's still good enough (in my case i had 150 lph in each block).
Bleeding air from the loop is another subject entirely and i had air bubble problems in GPU blocks (and worse: long motherboard loops..) long before i used parallel and SLI...