Data on single vs dual loop?

[jdcook]

I'm sure it's in here somewhere but my search-fu is leting me down (to much noise from .sig files I think). I'm looking for some actual data on single dual-radiator loops vs dual single-radiator loops. I gather the consensus is that dual single-radiator loops perform better but I don't really believe it as it doesn't make sense to me (which is not to say it isn't true, only that I can't figure it out). If the total heat output is the same and the cooling capacity of the radiators is the same, why would chopping it up into different loops improve things? The only thing I can think of is that the radiaotrs are more efficient at certain temperature ranges.

Anyway, I know there must be gloriously detailed tests in here somewhere. Can somebody point me the way?

[nikhsub1]

I could be wrong but I don't think specific testing has been done to meet your criteria... With that said, I have never been a fan of dual loops, I like to build it big or go home In your scenario above I don't think it would make much difference either way but I would have to give the edge to the single loop if you are only going to use 1 pump in that loop... in the dual loops you would need 2 pumps which would make 2x the head dump of 1 pump.

[millertime359]

Can't think of where that data is either. As far as I have been lead to beleive, the main reason to do a dual loop is too much restriction for the pump to handle. Then there was also the concern of heat dump and water temps hitting an equilbrium.

If you are only using running a single CPU, GPU, and rad. A single loop should be fine for the most part. If you are running multiple rads and GPUS, CPU and board. You might need to look into dual loops.

This also depends on choice of blocks. If you try to use the most restrictive blocks and rads, the amount you can use with suffer. Same thing, you use real free flowing blocks, you can use more.

[Ketzer7]

Well this leads to an interesting situation that I posted something about briefly in the T3 thread.

I'm looking at setting up a new WC build that would include CPU, motherboard NB/SB/mosfets, and 3x GTX285s all being water cooled.

With the advent of the T3 I had considered basically doing a CPU + mobo + red1 off of one outlet of the T3, and then doing the GPUs + rad2 off the other outlet.

Then someone mentioned this was not a good idea due to the heat dump of the GPUs into the same reservoir shared by the CPU, which I can kind of see their point. I've seen some post on here though saying that they didn't notice any temp increases running multiple GPUs and CPU in the same loop/reservoir, so I am

Now I'm back to looking at doing two entirely separate physical loops for each, but what's really the more ideal of the possible setups?

Everything on a single loop with multiple rads?

Two loops with a shared reservoir?

Two physically separate loops?

[millertime359]

That seemed to be the case before the T3. One of the golden rules was excess heat dump. I am in the same boat as you are with this.

From what I have heard about the T3 though, is seeing the water gets turned over something like 17 times a min, it doesn't get as effected as bad as a regular setup. The only way to be truely certain is to wait for more testing to be done. Your setup might even be too much for the T3. The data is just not out completly yet though

[Scubar]

Single loop and Dual loop single res will be the same temps. The only thing that will change is the flow as the split loops allow better flow rate in some strange way. The temps will be just the same.

[alacheesu]

I'm sure it's in here somewhere but my search-fu is leting me down (to much noise from .sig files I think). I'm looking for some actual data on single dual-radiator loops vs dual single-radiator loops.

A dual loop is just two single loops, and a single dual-rad loop is of course also just a single loop. Thus knowing how single loops behave should be enough to answer your question.

I gather the consensus is that dual single-radiator loops perform better but I don't really believe it as it doesn't make sense to me (which is not to say it isn't true, only that I can't figure it out). If the total heat output is the same and the cooling capacity of the radiators is the same, why would chopping it up into different loops improve things?

The main reason you run dual loops is to separate the heat. Let's try a small, simplified example:
2 equal rads, each dissipating 300W at 10C.
2 GPU's, dumping a total of 400W heat.
1 CPU, dumping 200W heat.

Everything on one loop would give you an air-water delta of around 10C.
Running 2 loops would give you a delta of 6,7C on the CPU loop and 13,3C on the GPU loop. This is good, because you want the CPU as cool as possible. GPU's usually don't care at all, and the increased temp is most likely irrelevant. If you wanted the same CPU temp in a single loop you would have to add a third rad.

[rge]

Also depends on whether the cpu and gpu will be loading at same time, usually they dont for my uses, and how many gpus you have and their idle watts.

Currently I have dual loops for cpu and gtx295 (dual gpu), each loop has ddc3.2, 360 rad, 6 fans push/pull.

CPU prime load measured by k/w meter and everest 185Wcpu+18Wpump has 5.7C delta air to water in my 360rad with current fans, for .02807 c/w. During prime load my other loop gtx 295 idles at 80W + 18W pump, also cooled by 360rad. If I combined loops into one, during prime, c/w.014x(185Wcpu load+36Wpumps + 80Wgpu idle (k/w meter) new delta air to water would be 4.2C, so would lower cpu load temps by 1.5C.

GPU loop would also benefit, as during furmark or gaming, cpu puts out less than 70W (everest), but even during full cpu load, gpu loop benefits since OC gpu full load ~370W.

However if loaded both cpu and gpu, then cpu loop does suffer. But since I cant think of anything I do that fully loads gpu and cpu at same time, I will be combining mine soon, so each loaded component gets 2x360 rads.

I would think if you know idle, loaded watts for each component, and not dealing with huge amount restriction problems, you could start with how you use it and figure it out using data in liquid cooling rad stickies. And yes there will be tntc variable differences, but there are many of those comparing different setups anyways...you wont know for sure till you test on your own setup.

[jdcook]

Thanks everybody. I hadn't considered head pressure issues at all. That makes a lot of sense. I can also see splitting the loop for particular delta-t reasons. (By background, my only previous venture into the wonderful world of watercooling was in the glorious days of the Celeron 300A. Things have changed a bit since then. I'm planning a new box and have decided to give it a go again.)