Water loop advice.. safe to split and merge flow?

[davidm71]

Hi,

I want to cool the gpus and cpu with the same pump but split the flow to different rads and even use the top rads filport to merge the flow. Can anyone comment on my design?

Thanks

There are three rads. A 240 rad on the bottom, a 280 in the front, and 360 slim on the top. The red bar is reservoir. The green lines are an alternate plan where the water goes straight and returns to the res as opposed feeding into the top rads fillport. I designed it like this to minimize extra tubing length.

Thanks

 

[timta2]

Splitting the flow seems pointless to me, since it's all one system and combined at the pump anyway. You should have enough cooling there to cool just about anything, doing it normally, without having to worry about any temps.

 

[Frode B. Nilsen]

Hi,

... Can anyone comment on my design?...

There are three rads. A 240 rad on the bottom, a 280 in the front, and 360 slim on the top. ...

For this system, there is only two components with equal resistance, which is the GPUs. Those you could run in parallel, as the total resistance of your rig, might be high. You just have not told us.

As for your drawing, it only shows that you have not understood distribution of pressure. Since you do not, run the rig in series, and the GPUs in parallel or series.

Also, you need to device a way to drain the system of water, not just filling it. For filling, you simply just need a flexible tube and a fitting, so usually there is no need for a filling port at the top of the chassis. A valve at the pump for draining, is probably a good idea.

Since this is a crammed setup, a combined reservoir and pump, is probably smart. You still need a drain valve, at the lowest point of the loop.

Adding an extra rad is always smart. In the bottom you could. Just move the pump up. That would give you about 1120, which is not to bad. With two GFXs this will not be a silent rig, but it might be a quiet one.

A super big res is not needed, 15cm is just fine.

 

[davidm71]

Splitting the flow seems pointless to me, since it's all one system and combined at the pump anyway. You should have enough cooling there to cool just about anything, doing it normally, without having to worry about any temps.

I dont see how theres enough cooling power to go around? If you chain one component, gpu or cpu or vica versa together, the heat would get carried to the next component wouldnt it and not give you the lowest possible temps? Not only that but the flow rate would diminish on the second component?

As far as the distribution of pressure no I dont understand. All I understand is that its not good to have restrictive rads such that im planning on returning the top Black ice stealth 30mm rad for another brand.

I thought of using the bottom rads third extra port as a drainport . Its just not drawn in.

Thanks,

 

[Erasmus354]

I dont see how theres enough cooling power to go around? If you chain one component, gpu or cpu or vica versa together, the heat would get carried to the next component wouldnt it and not give you the lowest possible temps? Not only that but the flow rate would diminish on the second component?

As far as the distribution of pressure no I dont understand. All I understand is that its not good to have restrictive rads such that im planning on returning the top Black ice stealth 30mm rad for another brand.

I thought of using the bottom rads third extra port as a drainport . Its just not drawn in.

Thanks,

1) Flow rate is constant throughout the loop. If it wasn't constant you would have water backing up into another dimension somewhere along the loop

2) "Distribution of pressure" is in my opinion a really awkward and bad way of understanding how flow and restriction works in a closed loop water cooling system. It actually works exactly the same as electrical circuits (if you are familiar with those). The analogs are: Pressure = Voltage, Flow = Current, Restriction = Resistance. Parallel and Serial works the same way that it does in an electrical circuit as well. So if you split the water flow, it splits with most of the flow taking the path of least restriction. If you do not balance the parallel paths to have roughly the same restriction you end up with the more restrictive path getting very little water flow (and therefore worse performance).

3) If you chain a component together you *do* get slightly worse performance on the next component. However when you actually run the math you see it doesn't matter. In a typical water cooling system, the water heats up by around 0.5C in a worst case after going through a hot component like a CPU or GPU. Therefore the next component is getting water only 0.5C hotter. The difference is therefore too small to worry about and doesn't really have much impact on the cooling performance of later components in the loop. The reason this happens is twofold: water takes a lot of energy to increase in temperature (specific heat), and the water is moving very quickly through the loop. So don't worry about this!

4) As for your plan, I can't comprehend your drawing in the slightest. I tried but I cannot figure out how you are trying to split the flow at all. My best guess is:

RES -> Bot Rad -> PUMP --> Front Rad -\
		       \	 -------> Top Rad --> CPU -> RES
			-> GPU -/		  \
			       \		   -> Endless loop back to Front Rad
				-> RES

In this setup, aside from some of the paths being pointless or completely broken the main problem is that you are essentially putting the Front Rad and the GPU's in parallel with each other. This means that the GPU's are going to see almost no flow, because they are much more restrictive than a radiator. So most of your flow is going to go through the Front Rad instead of the GPU's. EDIT: Actually looking at it you are putting the Front -> Top -> CPU in parallel with GPU, which means the CPU won't get much flow. Bah, regardless this is an impressively complicated setup that hurts my mind trying to figure it out. It will probably turn out very poorly.

I would strongly advise sticking to a 100% serial configuration and not trying to split the flow, it simply isn't needed.

 

[davidm71]

Hey Erasmus,

Thanks for your explanation. I've done some research and found a very guide guide describing the differences between serial and parallel setups on Koolance's web site. Their end point was that it doesn't matter either way and serial loops will have one component 1 degree hotter than the next and recommended experimentation if going another route.
I have also found out that my pump is extremely powerful and should be able to handle splitting the flow if I wanted but I'm open to your recommendation. Heres my path:

RES ->PUMP --> Front Rad -> Top Rad --> CPU -> RES
           \
            -> Bot Rad -> GPU 1 +2 -/-> RES

 

[Tsumi]

Again, it comes back to the problem of balancing restriction so that each loop has an equal (or similar) flow rate. If you have a very powerful pump, you're better off keeping everything in serial.
Parallel loops are most ideal with pumps that have high flow rates but low head pressure. Serial loops are most ideal with low flow high pressure pumps.
I run a parallel loop much like you posted. However, I have a pump on each loop with a combined reservoir.

 

[davidm71]

Well I have another reason why I wanted to split the flow in parallel keeping cpu and gpu separate. At the moment I don't have the resources to buy the gpus I want and thought I could get it off the ground and later add the gpu loop when I'm ready in a couple months. So so far I have everything but the cpu and gpus. I have a 290X in another older system I was going to use on air stock cooler (read these Asus OC 290X's are so overclocked that little benefit to be had water cooling it anyhow). I really want to get a 980 Ti Hydro. Maybe just one or two if I can max out the credit cards. Anyhow maybe a serial loop will do. What do you guys think of the Hardware Labs Nemesis 360 GTS Stealth rad? I heard its very restrictive on reviews and some people said that's good. Reviewer didn't think so. Don't understand why. I have very limited space up top that's why I'm going with that rad. Second choice is the Coolgate 360HD.

Thanks

 

[Tsumi]

If you don't have the financial means of doing it right the first time, I would highly advise against watercooling and getting a high end air cooler or AIO for your CPU, and getting GPUs with beefy air coolers. Watercooling has an extremely poor $/performance ratio, and is a hobby that will only be a black hole in your wallet.

Additionally, you will have to drain your loop to add in splitters. If you want an easy method of adding and removing components, you would need a pair of quick disconnects sets for each component you plan to change, at a cost of 50+ a pair.

 

[Nimisys]

You can get cheap plastic barb lossless disconnects at McMaster for about 11$ per Male/Female set.

 

[Frode B. Nilsen]

... I've done some research and found a very guide guide describing the differences between serial and parallel setups on Koolance's web site. Their end point was that it doesn't matter either way and serial loops will have one component 1 degree hotter than the next and recommended experimentation if going another route. ,,,

I would need a link for that page. This does not sound right.

I made a quick search, and only found something on GPUs and not loops.

The configuration you use depends on your individual cooling system. A common misconception is that running parallel paths in the same loop is always better. Experimentation is usually suggested. Keep in mind that parallel lines reduce coolant pressure, so results will depend upon the pump and tubing configuration. In a serial system, only 2-3°C is typically added between video blocks.

To illustrate this point, let's say two video blocks are connected in series. The first video block is operating at 45°C, while the second is at 47°C. This is disturbing to some users. They want both video blocks at exactly the same temperature, so the blocks are reconnected in parallel. However, due to the pressure decrease of parallel paths and additional restrictions in the loop, both cards may now operate at 48°C.

You have already been told that parallel for GPUs work because of equal resistance. So surely, you are not pointing to this article, claiming it is about loops. I just did not find the article you referred to.

 

[Erasmus354]

If you don't have the financial means of doing it right the first time, I would highly advise against watercooling and getting a high end air cooler or AIO for your CPU, and getting GPUs with beefy air coolers. Watercooling has an extremely poor $/performance ratio, and is a hobby that will only be a black hole in your wallet.

Additionally, you will have to drain your loop to add in splitters. If you want an easy method of adding and removing components, you would need a pair of quick disconnects sets for each component you plan to change, at a cost of 50+ a pair.

If you are on a tight budget, sticking to just water cooling the CPU for now might be the best bet. You can always do a second loop for the GPU's at a later date if you decide the noise or performance is not acceptable to you.

I would say if you decide to go with a CPU only loop you should stick to a single radiator (360 or 240) and leave the two other radiator spots in your current plan for the potential separate GPU loop. Typically these days a GPU loop (especially dual GPU) will have more heat dump than a CPU.

 

[Tsumi]

I would need a link for that page. This does not sound right.

I made a quick search, and only found something on GPUs and not loops.



You have already been told that parallel for GPUs work because of equal resistance. So surely, you are not pointing to this article, claiming it is about loops. I just did not find the article you referred to.

Coolant temperature variance between components is highly dependent on flow rates. The higher the flow rate, the lower the temperature variance. At an infinitely high flow rate, there will be 0 variance. Obviously an infinitely high flow rate is not possible, but 1-1.5 GPM is the generally accepted flow rate for minimal (0.5-1 C) variance in computer watercooling.