parallel and serial

Quite a few different views on this, but it seems there isn't much difference by way of temperatures.

I went with serial and have decent temps, seems a more simple way of directing the flow of coolant.
 
parallel feed coolant to both cards at the same time. Like using a Y splitter
serial is one card, followed by the next - like the rest of the loop or a string of beads.

Parallel can improve flow rate.
 
But it doesn't matter too much if the bottom card isn't getting as much flow long as the temperatures are decent.

With serial cooling loops, the temperature throughout the loop varies very little. With a parallel loop, it is almost impossible to get equal flow through all parts and this can lead to extreme temperature variations.

We don't watercool to get "decent" temperatures. :p
 
I went with the parallel EK sli bridge between my 2x 670GTXs both my temps are even at around 39 on full load, if I decrease the pump RPM the temps increase evenly, so flow seems to be pretty even for me.

On air they were hitting around 80ish
 
I went with the parallel EK sli bridge between my 2x 670GTXs both my temps are even at around 39 on full load, if I decrease the pump RPM the temps increase evenly, so flow seems to be pretty even for me.

On air they were hitting around 80ish

same here using ek temps are the same flow the same

like some one posted earlier if your getting drastic temp variation then some thing wrong with your loop or your card and not because u chose to go parallel
 
If you have a strong pump (D5) then I think serial works better as the pump can deal with slight flow restrictions and maintain a decent circulation through the system. If you have a weaker pump then parallel can help as you maintain total circulation by losing some local flow through restrictive blocks/connections on the parallel circuit.
 
The aquacomputer gpu block I just got states serial only with their blocks, presume this is down to the flow paths in the blocks.
 
Parallel is more difficult but potentially better. Serial is always simpler. Provided you're sure the two parallel sections are the same you're safe. If you're trying to send 1/3 water down one tube and 2/3 down another, you'll probably need to measure the flow rate. I started down the road of calculating pressure drop in various configurations but concluded life is too short.
 
The aquacomputer gpu block I just got states serial only with their blocks, presume this is down to the flow paths in the blocks.

I think so, I think that to get good cooling they have to be a bit restrictive allowing for surface area in contact with the fluid so if they are on a parallel loop it could limit the flow as the fluid will try to take the easiest path.

Don't think I'd ever use a parallel loop as I'd always be concerned that you'd get preferential flow at least with serial you know that all of the flow goes through all of the blocks.

The other thing I thought about parallel is that if you use a fluid with a suspension (pastel or aurora) if the flow drops through the parallel loop would you get issues with drop out ?
 
I don't understand this fear of parallel.

I run my 7950 blocks in parallel and both GPUs would have identical temps if one wasn't running a slightly higher voltage than the other.

If both blocks are the same there should be no flow issues at all and it should be a better setup. It certainly works for me but folk seem to have an irrational fear of it.
 
The problem lies with restrictive blocks (which I suspect the aquacomputer blocks may be).

Like electricity water will ALWAYS follow the path of least resistance, so the fear with running parallel is that a block may get directly by-passed, with the fluid finding the 'easy' path being back out of the loop rather than down to the next card.

I did run parallel for a good few months on my old 6950's with absolutely no issues though, that said I was using EK blocks, EK FC blocks have the highest flow/least restriction on the market.
 
I have used both serial and parallel. I started off with serial when I setup a 2way SLI. Temps were good but 3-4C higher on the last GPU in the loop. I did have a fairly restrictive CPU block at the start of the loop then though.

When I moved to three way SLI I switched to parallel. I did not want three GPUs in serial and I also wanted the least restrictive flow as I am "only" using a res/pump combo, the XSPC 750. Initially temps varied about 6c between the cards but after installing a new CPU block temps are now within 2c on all three cards.

I was worried about parallel too at first - serial is easier to understand and visualise but parallel isn't as easy to get to grips with.

My recommendation would be parallel for low to medium flow systems as it should give less flow resistance and serial if using high speed pumps.

Does that make sense?
 
in serial, you reduce the speed of flow to the whole loop when you add a new component, the solution to this is to over-spec your pump in the first place

in parallel, you keep the speed of flow high in the loop as a whole, but you halve or third the speed of flow in the components that are paralleled - which might be fine, but for me personally it's less than preferable

when I built my loop I had the plan to try serial with a view to going parallel, however it worked absolutely fine in serial... the last GPU in my loop is the one that runs the coolest and it makes no difference to the temps of any of my components which direction I run the loop in (cpu first or cpu last), which shows that my pump is up to the job

if you run in serial and you get ever increasing temps across components then it usually means that your pump is not up to the job or you have some very restrictive blocks (same thing really)

I did go with a 24v pump though and it does have something like 7m of head :D

if you are using dual D5's then I would just do serial as it will be easier and those pumps will easily be man enough for any loop you can throw at it!
 
Back
Top Bottom