Does using a sli bridge vs a T fitting matter?

PCnooberson · Jan 24, 2014

I was wondering if, temperature wise, flow wise, or anything wise, using a sli bridge (parallel) such as like this:

Would be any different than splitting the flow into 2 like this:

manofchalk · Jan 24, 2014

It shouldn't, their both a parallel configuration just done in different ways.

toolmaker_03 · Jan 25, 2014

It makes me happy to see the interest in parallel configurations but the same issue exists in both of the examples shown the CPU block is in a series configuration with the loop making that block the determining factor for the total system flow. This means that even thought the video cards are in parallel with each other, the flow through them is only half of what is going through the CPU block. this is a major no no. A MPC655 pump has a total flow of 8LPM today the major system components blocks are designed to handle around 2.2LPM through each one. this means that a 3-way parallel configuration is the largest that you can create, this will require a flow of 6.6LPM with is all most the max that the pump can provide. as for the system with the T's in its design, I would have done a few things different. I would have paralleled the two radiators with each other, and piped that in series with the paralleled video cards, and piped that in series with the CPU and mother board, that would also be paralleled together. that way I could obtain a total system flow of somewhere around 4.4LPM. I hope this helps, let me know if there is any thing else that I could help with.

PCnooberson · Jan 25, 2014

toolmaker_03 :

Thanks but im kind of confused on what your saying.

I see how the gpus are getting half the flow of the CPU (picture 2). Is it that important though? Beucase using a parralel bridge is the same thing isnt it? How would one go about doing this efficiently?

And you lost me on the rest with the LPM. are you saying that the pump cannot hand anymre components after the GPUs? or?

You said that the CPU is running in a serial way loop-wise, but it goes from gpu to rad to cpu. Isnt serial when one component is heated up, and that hot water is used to cool another component?

Sorry for not understanding, but im a bit of a newb

toolmaker_03 · Jan 25, 2014

https://imageshack.com/i/b50k5lj
https://imageshack.com/i/j6j1srj
https://imageshack.com/i/0v8ocbj

currently I am running a 3-way parallel setup I will try to explain I have 3 radiators in parallel with each other in series with a 3-way parallel hardware setup containing the two video cards and the CPU as a result I can achieve a total system flow of 6.6LPM

https://imageshack.com/i/n5done001j
https://imageshack.com/i/0dpicks001j
https://imageshack.com/i/0zmodular004j

before this I had a 2-way parallel setup I had two radiators in parallel with each other in series with two video cards that where in parallel with each other the total system flow on that setup was 4.4LPM

at that time my CPU was on its own loop one pump, one radiator, one block in series with each other and that loop ran at 2.2LPM

so basically it is simple 1 component one radiator, 2 components two radiators, 3 components 3 radiators, and that is the max for todays water cooling pumps, not to say that they can not handle more components I have more components attached to mine, I am only saying that the largest parallel configuration that you can create with major components is a 3-way parallel setup, that is, if you intend to supply it with total flow possible.

I suck at communication, but I am willing to try and explain what I have learned in many different ways, so I hope this helps and let me know if there is any thing that I can further explain.

manofchalk · Jan 25, 2014

Basically those pictures have a flow order like this.
-------------------------> GPU1 \
---> Loop -> CPU <------------> Loop
-------------------------> GPU2 /

While toolmaker is suggesting.
----------> GPU1\
Loop -<---CPU---> Loop
----------> GPU2/
Basically having all of your blocks in parallel with each other.

I can see the merits of a system like that, but I personally wouldn't. The way parallel systems works is that flow is split between the different paths according to how much resistance each offers. If you have two identical blocks, your going to get the same flow through each of them as they will both offer the same resistance. Once you add in different blocks to a parallel setup, you could find that more water flows through one than the others. In a worst case scenario where one block offers significantly less resistance, you could starve the other blocks of flow.

toolmaker_03 · Jan 25, 2014

correct MOC now with that I have also tested each component individually on all the builds that I have done, and have come to the understanding that todays blocks are all made with about the same restriction, regardless of weather it is a CPU block or video card block. the reason for this is simple a flow rate through the block of around 2.2LPM is the optimum flow rate for the water to remove as much heat as possible.

believe me or not, but even $30,000 dollar pumps with chilled water running through them to keep them cool, have a flow rate of 2.2LPM through the units water block, so I am pretty shore that there is some data somewhere that will support this clam, and that data will give reason for all the companies to start making this the standard design of all there blocks..

PCnooberson · Jan 25, 2014

manofchalk :

Basically those pictures have a flow order like this.
-------------------------> GPU1 \
---> Loop -> CPU <------------> Loop
-------------------------> GPU2 /

While toolmaker is suggesting.
----------> GPU1\
Loop -<---CPU---> Loop
----------> GPU2/
Basically having all of your blocks in parallel with each other.

I can see the merits of a system like that, but I personally wouldn't. The way parallel systems works is that flow is split between the different paths according to how much resistance each offers. If you have two identical blocks, your going to get the same flow through each of them as they will both offer the same resistance. Once you add in different blocks to a parallel setup, you could find that more water flows through one than the others. In a worst case scenario where one block offers significantly less resistance, you could starve the other blocks of flow.

Im sorry but i still dont understand the flow order. What do mean with loop? and the slashes?

manofchalk · Jan 25, 2014

By loop I just mean all the stuff like rad, pump and reservoir. Just its not important to what I was saying so its "loop". The slashes and Greater/Lesser Than symbols indicate where the flow goes, splits and converges.

toolmaker_03 · Jan 25, 2014

http://imageshack.com/a/img268/2112/bb0x.jpg

so this is the setup I have right now it is a 3-way parallel configuration of the main system composites.

the radiator setup is also a 3-way parallel configuration but the radiators and blocks are in series with each other meaning that there is a single line connecting the in's and out's of the radiators to the component blocks.

does this make any more sense.

Edit
https://imageshack.com/i/f2r244j

PCnooberson · Jan 25, 2014

toolmaker_03 :

Ah yea, i see what you mean by parallel, but would it be better to have the components in serial? Like gpu to rad to cpu to rad to res?

Because running the gpu and cpu in parrallel means that they each get a fraction of the flow, arent you suppose to want greater flow? When its in serial, only having one way to go, wouldnt that give it the highest flow?

manofchalk · Jan 25, 2014

Parallel configurations offer a much lower resistance to flow than a serial config would. A parallel dual GPU array offers about ~1/4 the resistance of those blocks in Series. The flow rate is halved within the array, but overall flow rates are much higher.

PCnooberson · Jan 25, 2014

manofchalk :

So it gives it less restriction, resulting in a higher flowrate?
However, do you think that it is necessary? Like is the diffrence in performance significant? Or is it one of those things were its good to have it, but yo dont need it.

toolmaker_03 · Jan 25, 2014

well if you had a pump after each component then yes that setup would have the greatest flow.

what I have done is tested each of the components that I have as a single loop. than I started paralleling the components together, while still maintaining the total flow needed for each component.

so if a CPU loop with a pump, reservoir, and radiator has a flow of 2.2LPM

and a single video card loop with a pump, reservoir, and radiator also has a flow of 2.2LPM.

than by paralleling the two radiators together and paralleling the CPU wilt the video card you should be able to obtain a total system flow of 4.4LPM by doing so you have eliminated the need for one pump and one reservoir that where used to create each loop individually.

https://imageshack.com/i/b59mduj

without paralleling the components in this way it would take 2 to 3 pumps to maintain a good system flow of a system that was configured to be in series with each other, that utilized that much hardware, does this make sense.

PCnooberson · Jan 25, 2014

toolmaker_03 :

I think im starting to get it. Hows this?

The gpu are parralled, so split into 3 like you said, they then meet up and cool off at the intake rads. Then they go into the second pump, split into three, cooling the mobo, ram, and cpu in parallel. THey meet up again after, get cooled by the exhausts, then back to the first pump.

THanks for the help too.

toolmaker_03 · Jan 25, 2014

https://imageshack.com/i/0y957fj

wow with something like what you have proposed I would do something like this give me a little time and I could clean it up but, I hope that you can get the general idea of what I am seeing this would be a good candidate for a two, on two, on two, on two, parallel setup.

PCnooberson · Jan 25, 2014

toolmaker_03 :

Yes, yours definately looks alot cleaner, and alot more efficient, but im a little stumped on the GPUs and CPU. You drew them all in parrallel right? But its split into 4. Didnt you say that the the most you can run in parallel was 3 components?

toolmaker_03 · Jan 25, 2014

sorry about that it wont work that way its getting late and I was not thinking

https://imageshack.com/i/b5xhbkj

here is a better one

PCnooberson · Jan 25, 2014

toolmaker_03 :

I dont see how letting them converge at the top rather than the bottom is any different. If you want, we can continue this tommorow, youve been tons of help already.

toolmaker_03 · Jan 25, 2014

thanks I will talk later after some rest

PCnooberson · Jan 26, 2014

Any new ideas?

toolmaker_03 · Jan 29, 2014

Ok so yes, I have some info about the flow rates on the motherboard blocks and memory blocks, these my be different that the regulated flow through the major components water blocks. that is why I have linked them to each other after the CPU and video card blocks, that way they will not affect the system in a negative way.

PCnooberson · Jan 29, 2014

toolmaker_03 :

But you cant have the GPUs and CPU in parallel because thats 4 components? And are you sure that running the ram and mobo blocks in parallel? Im sure one is more restrictive than the other, and if water takes the path of least resistence, it might mean one block will end up with not enough water and lower performance. Im not sure if there is any other way to go for the blocks other than serial?

toolmaker_03 · Jan 29, 2014

it might be easier to split this into two loops you have two pumps already for this build so on one loop have the two lower radiators with the larger reservoir and the three video cards. on the second loop you could have the motherboard, memory, and CPU, with the upper two radiators and smaller reservoir.

PCnooberson · Jan 29, 2014

toolmaker_03 :

Honestly, splitting into two loops doesnt seem to be much of a change. All that would be doing is moving tubing a few inches to connect to the res's. Also, i prefer to have one loops, so that all the Rads are shared. with two, one loop might not be getting more rad space than is needed, leaving the other one not getting enough, if that makes sense.

Does using a sli bridge vs a T fitting matter?

Honorable

Glorious

Honorable

Honorable

Honorable

Glorious

Honorable

Honorable

Glorious

Honorable

Honorable

Glorious

Honorable

Honorable

Honorable

Honorable

Honorable

Honorable

Honorable

Honorable

Honorable

Honorable

Honorable

Honorable

Honorable

Share this page