Single 2 Rad Loop or Y-Split to 2 Rads Loop

[bigpoppasmooth]

Hi All. Thanks in advance for your input. I consider myself to still be somewhat of a n00b. I have rebuilt my loop 4/5 times as I've upgraded PC parts and cooling parts. Most recently I've upgraded from a 240 & 360 rads to 2 420 rads.

I am currently running a single loop with said 2 420 rads.

Res/Pump -> Hardware Labs Black Ice SR2 420 MP -> CPU -> Chipset -> GTX 1080 -> GTX 1080 -> Hardware Labs Black Ice SR2 420 MP -> back to the Res/Pump

For the most part I am happy with my overclock & temps...but like every Overclocker out there I want every drop of extra (everything) that I can get.

My proposed new Y-Split loop would go:

Res/Pump -> Y-Splitter -< -CPU -> Chipset -> Hardware Labs Black Ice SR2 420 MP - back to Res/Pump

GTX 1080 -> GTX 1080 -> Hardware Labs Black Ice SR2 420 MP - back to Res/Pump​

Basically 2 loops after the Y-Split, then heading back to the same Res/Pump.
There is no Y-Split back to the Res/Pump. My Res/Pump has 1 Out ports, and 2 In ports.

I'm curious to know your thoughts, and/or results if anyone has tried this before.

 

[Spartacus09]

First what pump/res do you have?
Second based off your design I'm assuming you want to get more out of the CPU?
The best overall performance would be to run a dual loop technically, I'll let someone else chime in on change in loop performance of series vs split (I dont think it will change it much though).

That being said the biggest concern I see with a split would be the changes in flow rate and pressure since the CPU block and gpu blocks are different resistances and it wanting to take the path of least resistance.
If you were to do a split, (assuming you have the same waterblocks/GPUs, I would recommend doing this to keep the resistance equal:
Res/pump ->Rad1->cpu->chipset-> Rad2-> split-> (both 1080) -> Res/pump

 

[Zarathustra[H]]

The best overall performance would be to run a dual loop technically,

I fundamentally disagree with this.

All else being equal, a single combined loop will provide better temps for two reasons.

1.) The CPU and GPU are rarely both at full load at the same time. Because of this when you are playing a game that hits the GPU hard, but maybe only loads the CPU to ~30%, you have all that extra radiator capacity cooling your GPU. Then when you go to encode a video, and have almost no load on the GPU, the CPU benefits from both radiators. If you had a dual loop setup you would have one radiator completely heat soaked, and the other sitting there cold. This is not very efficient use of your radiators.

2.) With a large single loop you can get away with one pump. With two loops you need two pumps. Most common pumps like the D5 dump their heat into the loop, so now you have two pumps dumping heat into loops instead of just one.

I can't think of a single implementation where you'd get better temps from a dual loop than from a single loop if the radiator capacity is the same.

I'll let someone else chime in on change in loop performance of series vs split (I dont think it will change it much though).

I have never tested splitting a loop like this but I think you are right, the impact should be small, but I think that small impact is NEGATIVE not positive.

Here is my thought process. When you split the loop with a splitter, you halve the flow after the split, so the water flows slower through the components after that split. For the radiators this doesn't matter much, but for the blocks this could be an issue and reduce performance. Because the fluid is moving slower through the blocks, there is a risk that the water entering on one side is cooler than the water exiting the other side. This could result in higher GPU/CPU temps compared to a high flow situation where there is very little water temp change across the block, so the entire block sees the same low inlet temp.

Ideally in a water loop you want a high enough flow rate that there is a negligible temperature change across a block. This results in the entire loop entering more of a steady state, and it becomes more efficient.

Now, if your pump is powerful enough to keep a high enough flow after the split (tpyically little benefit is seen above ~1GPH), then there should be no adverse effect at all, but I can't think of any circumstance under which this would improve anything.

 

[bigpoppasmooth]

First what pump/res do you have?
Second based off your design I'm assuming you want to get more out of the CPU?
The best overall performance would be to run a dual loop technically, I'll let someone else chime in on change in loop performance of series vs split (I dont think it will change it much though).

That being said the biggest concern I see with a split would be the changes in flow rate and pressure since the CPU block and gpu blocks are different resistances and it wanting to take the path of least resistance.
If you were to do a split, (assuming you have the same waterblocks/GPUs, I would recommend doing this to keep the resistance equal:
Res/pump ->Rad1->cpu->chipset-> Rad2-> split-> (both 1080) -> Res/pump

I am using a Photon 270 Reservoir with a Laing D5 pump.

I understand what you're saying about flow rate, but wouldn't that be negligible because I'm bringing both flows back to the Res seperately? - Stupid question by ME answered by Zarathustra[H]

 

[bigpoppasmooth]

I fundamentally disagree with this.

All else being equal, a single combined loop will provide better temps for two reasons.

1.) The CPU and GPU are rarely both at full load at the same time. Because of this when you are playing a game that hits the GPU hard, but maybe only loads the CPU to ~30%, you have all that extra radiator capacity cooling your GPU. Then when you go to encode a video, and have almost no load on the GPU, the CPU benefits from both radiators. If you had a dual loop setup you would have one radiator completely heat soaked, and the other sitting there cold. This is not very efficient use of your radiators.

I think you just nailed the essence of what I was asking. With that logic staying with the single Loop looks like the way to go. Thank you!

 

[Spartacus09]

I fundamentally disagree with this.

All else being equal, a single combined loop will provide better temps for two reasons.

1.) The CPU and GPU are rarely both at full load at the same time. Because of this when you are playing a game that hits the GPU hard, but maybe only loads the CPU to ~30%, you have all that extra radiator capacity cooling your GPU. Then when you go to encode a video, and have almost no load on the GPU, the CPU benefits from both radiators. If you had a dual loop setup you would have one radiator completely heat soaked, and the other sitting there cold. This is not very efficient use of your radiators.

2.) With a large single loop you can get away with one pump. With two loops you need two pumps. Most common pumps like the D5 dump their heat into the loop, so now you have two pumps dumping heat into loops instead of just one.

I can't think of a single implementation where you'd get better temps from a dual loop than from a single loop if the radiator capacity is the same.

While I agree overall from an economical and efficiency standpoint, those rads independently should be sufficient to cool those items separately.
It also depends heavily on your blocks, every single item and bend in your loop goes through is going to drop slightly in flow or pressure.
So if you have two smaller loops each with their own pumps with the cpu + chipset and then the dual GPU in another it would theoretically perform better, be it marginally.

However you're then doubling the risk of failure/leaks, etc though. So an increase in flow and pressure could be achieved by a redundant pump in the single loop as well.
All that being said as I mentioned its not the most economical option.

Looks like OP got their answer though so, cheers!

 

[Tsumi]

The biggest issue with using a Y-splitter in such a manner is you will get uneven flow due to unequal flow restriction. GPU blocks will generally have less restriction than CPU blocks, and you're going to get significantly more flow through the GPU section than the CPU section. Only go two loops like that if you plan on running two pumps, one for each loop.