I've been designing a few builds involving watercooling for a while now, more of a 'what if' past time than something that'll happen in the very near future. That'll be all air. for now.... Now I'd just assumed that the most efficient way to cool GFX in SLI/Crossfire would be in series, but I noticed in a couple of MOTY builds that their GFX were cooled in parallel: Silverstone TJ11 Carbon by keir White by alain-s And some in series: L3p D3sk by L3p Saturday Night Special by bulldogjeff (Incase you read this Jeff, even though I didn't read the thread till after I'd seen the finished article, that was still one hell of an emotional ride! Sometimes I'd forget myself and start screaming "no don't give up!" at my screen. I hope now after time has passed you can forgive her. She's an absolute beaut!) What's the what's the pro's and con's of both and is one more efficiant than the other?
I would imagine parallel would cause less flow through the blocks, whereas series, there is only one route so flow would be increased. If the pump can provide the flow though, parallel might offer slightly better cooling for the second/third cards. Although I suspect the differences are probably negligible. Would be interesting to see if it did have an effect.
I thought pretty much the same, but I thought in parallel most of the water would rush straight through to the last card as it'd be the path of least resistance.
Not really, because the block will offer a large amount of resistance, so the water would 'back-up' until it hit the 2nd, then 3rd and so on and so forth. You might gain a degree or 2 this way, but I'm not sure, I've never had enough GPU's or watercooling to try this out. Although if I do my plan for a loop would look rather sick
I switched from series to parallel and haven't noticed much of a difference, maybe the GFX cards are a degree hotter and it takes a lot longer to fully bleed the loop. I did mine for aesthetic reasons as I have enough pumps for decent flow even with 7 blocks. It's supposed to help if you have a lot of blocks or marginal flow rates as it decreases flow rates through the GFX blocks but increases it in the rest of the loop.
Thanks for the info. I'm glad to hear that it doesn't really add anything to the temps as i wanted to do it parallel. My planned route for the pipes would mean there's a fair bit of distance between each card. you think this would make much difference? I'm talkin 12" at least!
You have to bear in mind that not only am I seriously over-radded but until I can afford decent cards only have low wattage ones (220W max for the pair). That might make a difference of a few more degrees. I'm interested how (and why) you're going to have such a big distance between cards?
oh i won't have that big a distance between the cards. that would be quite some feat! The plumbing is just goin to go on a scenic detour for purely aesthetic purposes. lol
Water does not heat up and cool down between components in your loop. Please burn that fact into your mind - It's one of the biggest misconceptions in watercooling to think that the water temperature (or the water's capacity to draw heat away from a component) is affected by block arrangements. Think of your loop as a constant stream of steady heat moving around, not like a bus picking up and dropping off passengers (heat) at stops (blocks/rads). I doubt it makes much difference, if any, whether your cards are connected in series or parallel; as long as the same flow-rate is attained on all 3 blocks in a parallel situation, remembering that water takes the route of least resistance.
I have noted scenarios in which loops are not in equilibrium. L3p D3sk is one of them, his 1080 hidden underneath that freezing part of his house gives him a 4 degree difference in temps before and after the radiators.
That's an extreme case though, with far longer runs than is typical in 99% of setups. The point is that putting the CPU before the GPUs doesn't mean the GPUs are getting 'warmer water', or that the 2nd card in an SLI loop is receiving 'warmer water' - Not to any degree outside of experimental error. In any kind of loop with a reasonable flow-rate the water is moving too fast from one component to the next for it to heat up and cool down between components.
While some components give out more heat than others, after a few minutes of running the system will have hit an equilibrium where it doesn't matter. The plans I have is for there to be '2 loops' but they are connected like so: Res1>Pump1>CPU>Rad1>Res2>Pump2>GFX>Rad2>Res1....
The times I've used parallel I have higher temps on those parts. Serial doesn't look as good but I want max pressure in each block. My pump has very high head pressure and my rad has half the pressure drop of a normal triple 120 so flow rate is good.