WC Dual cards config right or wrong?

Associate
Joined
17 Mar 2009
Posts
2,475
Location
Scotland
This is my watercooled setup for my dual 780ti's and below it is a copy of a post on FB of some guys telling me I'm all wrong. After having spoken to many of your guys prior to my build the way I have it is perfectly acceptable. So is this chap full of wind and **** or does he have a valid argument? Your thoughts?

aWqp8H3l.jpg


VhlFSOY.png
 
You've got it right... You can plumb them in parallel instead which will give half the flow over each GPU but will have a more even temp. As it stands, you may have a ΔT over the two cards of 3-4ºC.

Exactly the same plumbing as mine :)
2DFAD86C-5ABA-4807-9B53-D22E31AD4E88_zpszgaltmix.jpg
 
Well I didn't want to jump on the chap but I agree the Delta T will be minimal and given that the system runs without issue I really can't see the argument. I just dislike people telling me something is totally wrong ITO. I had folk telling me I've got my bottom fans wrong to but that's their opinion, I know its breaking with traditional thinking but I didn't do it out of ignorance.

I suppose its something to get used to if you plaster your build around the net there will always be +ve and -ve feedback.
 
Exactly. The only pros/cons of going parallel over series for me is the convenience of the layout. Ie. if you want both the in and out coming out of the bottom. It has just made me wonder whether the direction of flow matter on GPU blocks though... I know my raystorm has an 'In' and 'Out'...
 
Exactly. The only pros/cons of going parallel over series for me is the convenience of the layout. Ie. if you want both the in and out coming out of the bottom. It has just made me wonder whether the direction of flow matter on GPU blocks though... I know my raystorm has an 'In' and 'Out'...

You mean if you put it in one side it gets the the hotter components first rather than the other port which may be less heat transfer? You'd think with the size of the block as a whole this would be not something to worry about, the card/block must find its equilibrium at some phase.

Sounds like the kind of guy that also thinks you need a rad between each block.

Yeah man, I kinda thought the same and wondered if there is anyone out there planning a triple loop just in case ! I even had a build idea moment based on that theory ! But quickly discarded it to the "builds I won't be doing" part of my brain.
 
his talking rubbish..

In serial generally the first card in the loop will be slightly cooler.. lets assume 50c and the second card will run at + 3 - 5c from the first so lets say 53 - 55c.. this will be more restrictive, higher flow rate..

In parallel both card will run at the same temperature but that will be higher than card 1 in serial.. so probably around 52-53c in the example numbers.. this will be less restrictive, lower flow rate..

ofcourse this is all generalisation and a lot of it depends on loop and pumps and the like..

Also reference rad between cards.. well everyone knows that the water temperature in the loop is pretty much constant.. for example in my last dual card loop (2x 4870x2) the temperature delta between the water after the 2nd card.. and the water after the rads was 1 - 2c..

Bottom line all things equal it makes not enough difference to worry about..
 
Last edited:
I was considering putting a rad between my cards as opposed to having both cards straight after each other in the loop. Loop would be...

Res -> Pump -> CPU -> GPU1 -> 240 Rad -> GPU2 -> 360 Rad -> Res

Would that be better, just the same or worse?
 
Going from CPU direct to GPU I'd be wary of mate, the temperature difference between inlet of gpu1 and gpu2 would be quite a bit more than if you had the CPU-240Rad-GPU1-GPU2-360Rad I'd have thought.

At the end of the day its all about your gpu core temps, just how much they will vary based on your proposed routing is anyone's guess. Are you planning a big OC to the system?
 
Going from CPU direct to GPU I'd be wary of mate, the temperature difference between inlet of gpu1 and gpu2 would be quite a bit more than if you had the CPU-240Rad-GPU1-GPU2-360Rad I'd have thought.

At the end of the day its all about your gpu core temps, just how much they will vary based on your proposed routing is anyone's guess. Are you planning a big OC to the system?

Yeah. 4670k sits at 4.4 and GPU's will be close to the max.
 
If you have the time and the patience then try both options, test and then decide. TBH there might not be a huge detrimental effect, depends if you are going for a certain cosmetic look also.
As I was saying recently some people will sacrifice function for form, its not always about raw performance for me.
 
Both are fine when you have 2 cards.... If you have 3 I'd do it like you have as doing them as he is suggesting can sometimes hinder results. I tried it on my cards and the flow just didn't seen to bother going through the bottom card as the fluid had an easier path through the top 2....
 
Just for info I go CPU > GPU1 > GPU2 in serial and the temps are excellent.

the Temp difference between GPU1 & GPU2 is 1 degree idle and 2-3 at full load.
 
Loop coolant will reach equilibrium temperature anyway.

The only reason you'd opt for parallel flow over serial is if you don't have high enough flow rate to counter the restriction.

If your flow rate is so low that the coolant in GPU1 is being heated up a significant amount to affect the temperatures on GPU2, your loop is already boned!
 
Loop coolant will reach equilibrium temperature anyway.

The only reason you'd opt for parallel flow over serial is if you don't have high enough flow rate to counter the restriction.

If your flow rate is so low that the coolant in GPU1 is being heated up a significant amount to affect the temperatures on GPU2, your loop is already boned!

I was thinking along the same lines. The water should be moving fast enough that the block doesn't have time to cook it that much.
 
I have my cards configured in serial (same as yours) and I have a temp sensor before and after the cards in my loop.

Even at full whack there is barely 1.5 c difference in the temps.
 
Back
Top Bottom