GTX 980 TI 3 way SLI - Question to TTL

Trumpu

New member
Hey guys and Tom,

I was going to ask you this at i55 but I'm not going this time. and I've been struggling to find a way to direct message you, obvious spam prevention reasons :P

But I'm currently building a new RIG based on Skylake, and i have the money to include 3 of the following:

http://www.scan.co.uk/products/6gb-...hz-gddr5-gpu-1190mhz-boost-1291mhz-cores-2816

I wanted to do something differant with the watercooling that i've not seen done, because I'm a tidy freak. BUT i can only imagine it's not been done before because it can't be done.

in my Case, Corsair Graphite 780T:
http://www.scan.co.uk/products/corsair-graphite-series-780t-white-full-tower-case-w-o-psu

I want to route the loop in the back of the case with the cable management too, and what i would like to do, but I'm not sure it's possible is:

have the three cards in parallel and have the tubing on the bottom card's waterblock for the inlet and outlet, however i can understand that the water pressure could / would be too low to create circulation through the second / third graphics card waterblock.

the pump i have bought is:
http://www.scan.co.uk/products/ek-d5-vario-x-top-plexi-including-water-cooling-pump

If you can't understand the layout I'm trying to describe then i can attempt to draw a diagram to help describe what I'm asking, If it's possible.

I do predict the answer is no based on the lack of water pressure in the loop.

Cheers!
 
I dont like paralell graphics tbh

hose wise Im really not sure about up the back - thats going to be down to your abilities and what the case is capable of and not something I can really say too much about specifically


Pump wise its plenty btw - just do the GPU in series.
 
Last edited by a moderator:
EVGA's Classified cards use a proprietary PCB with custom power delivery so a reference 980Ti/TitanX block would not be compatible.

Here is a comprehensive list of cards compatible with the block you posted - http://configurator.ekwb.com/waterblock/3831109830574



Exactly why Tom doesn't like parallel blocks i'm not sure but I certainly wouldn't consider that a good enough reason not to do it. I wouldn't hesitate to link a matched set of blocks in a parallel configuration, including a CPU block with a different restriction is another matter although still entirely plausible.

You have a slight misconception when it comes to the pump setup lacking the 'pressure' to deal with 3 parallel blocks. By linking blocks in parallel the restriction of the loop is effectively reduced as the water can take 3 different paths, because the paths are close to identical an equal amount of water will pass through each. If the 3 blocks were configured in series then all of the water would pass through every block thus increasing the restriction and reducing the flow rate.

A single D5 will easily be able to deal with the restriction of three blocks in series so there is no real advantage to either setup, the cooling ability will be very similar.


Series seems more appropriate for your plans as would only need two ports entering the back of the case from the blocks rather than four. There is no reason why you couldn't route the loop in the back of the case, it could work out very neatly. I think it is done occasionally but you generally just don't really 'notice'.

JR
 
I agree with JR, a parallel config would suit three identical blocks quite well, in terms of aesthetics and reduced restriction.

It's a bit like Ohms Law. Consider three equal resistors (gpu blocks) with a restriction (resistance) of an arbitrary value of 5. In series this adds up to 15 (Rtotal = R1+R2+R3).

In parallel the equation is 1/Rtotal=(1/R1)+(1/R2)+(1/R3), so 1/Rtotal=(1/5)+(1/5)+(1/5), Rtotal=1.667

In summary:
Serial = 15
Parallel = 1.667

That is considerably less resistance (restriction).
 
Thanks for your input, you have overseen why i spoke of loop pressure though, i spoke of it because i had the idea of having the input of the loop on the bottom card and the out take on the bottom card, thus the coolant would be fighting gravity to circulate up to the second and first GPU.

I will draw a diagram tomorrow to explain my idea, although I'm quite sure it's not plausible
 
Gravity doesn't matter - the pumping height of the D5 is around 3 metres so unless you have a 10 ft tall case, the pump is more than capable of pushing coolant all the way to the top of it.
 
Yeah that's still not what I'm saying, the GPU's would be in parallel, so wouldn't the fluid enter GPU 3, and exit straight away and never fill up to GPU 2 and 1, because of gravity and pressure.

I'll draw a diagram in the morning so everyone understands 100%. I would love to create the loop this way as i think it would look great!
 
Yeah that's still not what I'm saying, the GPU's would be in parallel, so wouldn't the fluid enter GPU 3, and exit straight away and never fill up to GPU 2 and 1, because of gravity and pressure.

I'll draw a diagram in the morning so everyone understands 100%. I would love to create the loop this way as i think it would look great!

Not really because the restriction of the water passing through a GPU block is far greater than raising it 4cm so the restriction of each path is more or less the same.

JR
 
If you are concerned about your pumps ability to combat gravity, you need a different pump! However you have selected a D5 which is an excellent pump and more than adequate. Regarding "fighting" gravity - as your loop is (hopefully :)) a closed loop you needn't worry. Check this out:

https://www.youtube.com/watch?v=5P144FcPXzo

Different application but still relevant.

The only thing you must concern yourself with is that the pump is not self-priming and must be supplied water, usually by having it fed water from a reservoir.

A diagram never hurts but I think I know what your intention is - water in and out on the bottom graphics card.
 
If this is a gaming rig you are building and if you plan to run @2160p you could hit the 6gb VRAM limit on the cards with 3 way SLI.
 
piDUeGn.png


This is what i would like to do, i think it would look great. But my fear is once the liquid is in the loop, it would never enter GPU 1 and 2 and would enter GPU 3 and Exit GPU 3 as the GPU blocks would be in parallel and gravity would prevent the fluid entering gpu 2 because it would exit the waterblock of gpu 3 and fall back into the tubing on the other outlet of GPU 3.

If my fear is wrong, would there be enough flow over GPU 2/ 1 when the loop is filled?
 
Back
Top