Dual pump or no?

[I_Luv_Cheetos]

My rig:
CPU: i7-3770k
Motherboard: ASRock Extreme 4
GPU: ASUS GTX 680 (in SLI)
PSU 1050w XFX PSU
Case: Rosewill Thor V2

I'm trying to watercool both GPUs and the CPU and I plan on overclocking both GPUs and the CPU, I'm not sure how high I want the GPUs but I know I'll be aiming for 5.0Ghz on the CPU. I may not be able to hit 5.0 but I know I can do 4.8Ghz on my noctua NH-D14 atm so I'm hoping for 5.0ghz on the water cooling. I'm planning to have dual XSPC AX360 or maybe RX360 I haven't decided yet and the decision will most likely be based on which ever is cheaper at the time.
Since the Thor-V2 Isn't exactly the best thing in the world to mount radiators in, I'll be running it outside of my case, into an external box. I plan on the external box having the pump(s) in it also.

For the pump I planned on using the XSPC dual bay Res+pump combo with the D5(s) in it, two universal swifttech GPU blocks, and an XSPC Raystorm block for the CPU.
tubing and fittings I imagine don't matter a whole lot when it comes to this decision.

My question is whether or not I should get the Dual pump Res or just go with the single pump... Pumping through all of that I would be sure that the single pump would be fine if I wasn't pumping it through to an external box.
Anything else you need to know I'll be glad to provide the info.

Also, I'm doing this for fun, and as a hobby. It may not be practical to have all of that rad space and whatnot but what is overkill when you are already going down that road?

 

[P1nnacle]

TLDR: Yes, you'll be more than fine.

So if I read this right you're pretty much planning to run a WC system with on CPU block and 2 GPU blocks, a res/pump combo, and an externally mounted radiator. If so you really shouldn't need a second pump, the D5's flow rate was more than enough for my 3 GPU and CPU set up that ran through a 3x120 extra wide radiator. You really need just a fraction of the flow rate the D5 provides to cool what you're trying to do. My current system is 2 GPUs 1 CPU and a D% pump, and at full load on the GPUs and CPU (4.2 GHz) it only reaches 67 degrees C. You should be able to hit 5.0 GHz with plenty of room to spare heat wise


Also, personally I would never buy a combo res and pump. You're just asking for leaks. Just my opinion.

Rereading your post I wanted to add onto my statement. Don't worry about how far you need to pump, distance doesn't play a role in a closed loop system (at least not at this level). You could put your external radiator 20 ft away from your computer, and all that would change is the amount of water you need to add to your system.

 

[faalin]

I have a 2500k and 2 560 ti's running off a single pump on one of my computers and it does just fine. one pump should do fine its really up to you if you want 2.

 

[I_Luv_Cheetos]

TLDR: Yes, you'll be more than fine.

So if I read this right you're pretty much planning to run a WC system with on CPU block and 2 GPU blocks, a res/pump combo, and an externally mounted radiator. If so you really shouldn't need a second pump, the D5's flow rate was more than enough for my 3 GPU and CPU set up that ran through a 3x120 extra wide radiator. You really need just a fraction of the flow rate the D5 provides to cool what you're trying to do. My current system is 2 GPUs 1 CPU and a D% pump, and at full load on the GPUs and CPU (4.2 GHz) it only reaches 67 degrees C. You should be able to hit 5.0 GHz with plenty of room to spare heat wise


Also, personally I would never buy a combo res and pump. You're just asking for leaks. Just my opinion.

Rereading your post I wanted to add onto my statement. Don't worry about how far you need to pump, distance doesn't play a role in a closed loop system (at least not at this level). You could put your external radiator 20 ft away from your computer, and all that would change is the amount of water you need to add to your system.

Okay, I wasn't ever worried about pumping the distance, I grew up around aquariums, so I know about pumping distance. I was wondering about the D5 pump being strong enough to pump out and down about 3ft then pumping up and down in the box then pumping back up to the case.
I would be able to judge this myself but I've never experienced a D5 myself so I have no idea how well it pumps.

and thanks for the tip about the Res + pump combo, If I definitely only need one pump then it probably won't be an issue getting separate pieces.
Edit: Forgot to add that eventually I may want to cool my motherboard with it, would 1 pump still be enough? I read that the mobo covers can get pretty restrictive.

 

[P1nnacle]

You should just get a single D5 for the entire system. It will be more than capable of pumping for a 3 ft drop and back up. Remember, since its closed loop, the pump is pushing water out, which once it gets through all the components it's technically pushed back up into the pump, not so much sucked in. The D5 pumps something like 314 GPM, which is stupidly overpowered for almost any job. as far as restriction concerns with a motherboard, if you run the entire setup as a series you'll have no problem (the velocity of liquid will be higher through the motherboard block). You don't have to worry about restrictiveness unless you're planning on a parallel system (to which there is very little benefit over series). I think your main problem is just going to be filling the system.

 

[manofchalk]

Have the GPU's arranged in parallel to each other (lowers restriction to about 1/4 of them in Series), and I don't think you will have any issues in terms of pumping power.

 

[I_Luv_Cheetos]

Have the GPU's arranged in parallel to each other (lowers restriction to about 1/4 of them in Series), and I don't think you will have any issues in terms of pumping power.

How does this work with universal blocks? there wouldn't be a top and bottom outlet/inlet to the blocks

 

[manofchalk]

You could set up a parallel array with Universal blocks, you would just have to make the connections using fittings rather than the block itself.

............................. -> GPU1 ->.........................
Loop -> Y Piece ->................... Y Piece -> Loop
............................. -> GPU2 ->.........................

 

[P1nnacle]

You could set up a parallel array with Universal blocks, you would just have to make the connections using fittings rather than the block itself.

............................. -> GPU1 ->.........................
Loop -> Y Piece ->................... Y Piece -> Loop
............................. -> GPU2 ->.........................

We can go all day as to which is better, parallel vs. series, but having tried both I think he should stick with a series setup, it's easier and the restriction problem is rendered moot by basic laws of physics. Basically, he's just fixing a problem that doesn't exist by creating a parallel setup for his GPUs.

 

[manofchalk]

I think a parallel GPU array could be justified here, universal blocks are usually a fair bit more restrictive than their full-block counterparts. Especially if mobo blocks are going to be thrown in as well.
Do note that I am only saying parallel the GPU's, not the whole loop. The GPU array would be in series with the rest of the loop.

 

[P1nnacle]

I think a parallel GPU array could be justified here, universal blocks are usually a fair bit more restrictive than their full-block counterparts. Especially if mobo blocks are going to be thrown in as well.
Do note that I am only saying parallel the GPU's, not the whole loop. The GPU array would be in series with the rest of the loop.

My point is that in a full series setup, restriction is irrelevant. The equation of continuity states that for an incompressible fluid flowing in a tube of varying cross-section, the mass flow rate is the same everywhere in the tube. The mass flow rate is simply the rate at which mass flows past a given point, so it's the total mass flowing past divided by the time interval.

Simply put, the water will speed up in a restrictive area and slow down once it reaches the less restrictive tubing. All that adding a parallel section will do is give you an increased likelihood of air pockets sticking in the GPU blocks, and for no benefit.

 

[manofchalk]

My understanding is that the pressure of the pump interacts with the restriction provided by the loop to determine your overall flow rate.

Putting components in parallel significantly reduces the restriction offered by those components, hence your overall flow rate increases, even if the flow rate inside the array is half that of the overall flow. Assuming each block provides a resistance of 1, so you get a total resistance of 3 (one CPU and two GPU blocks) when in Series as resistance just stacks in Series.

When in parallel, you are reducing the resistance offered by those two GPU blocks to just a quarter of them in Series, or half that of a single block. Suddenly you end up with a total resistance value of 1.5, half of when your running in Series. Hence you get faster overall flow rate.

I could also make the argument that you dont want the heat of one GPU being directly dumped onto the next one, and that a parallel array means that each one is getting cooled water.

Some good reading.
http://dragonsgaming.wordpress.com/2011/04/06/water-cooling-video-cards/

 

[P1nnacle]

My understanding is that the pressure of the pump interacts with the restriction provided by the loop to determine your overall flow rate.

Putting components in parallel significantly reduces the restriction offered by those components, hence your overall flow rate increases, even if the flow rate inside the array is half that of the overall flow. Assuming each block provides a resistance of 1, so you get a total resistance of 3 (one CPU and two GPU blocks) when in Series as resistance just stacks in Series.

When in parallel, you are reducing the resistance offered by those two GPU blocks to just a quarter of them in Series, or half that of a single block. Suddenly you end up with a total resistance value of 1.5, half of when your running in Series. Hence you get faster overall flow rate.

I could also make the argument that you dont want the heat of one GPU being directly dumped onto the next one, and that a parallel array means that each one is getting cooled water.

Some good reading.
http://dragonsgaming.wordpress.com/2011/04/06/water-cooling-video-cards/

Thanks for the link, it's good knowledge. Here's where I disagree though:

Not to be all "hurdy hur I'm better, I am really just trying to help", I've actually had tests run on flow rates for a complete 1 cpu 3 gpu system (in parallel) with ASPEN, which is a large corporation engineery program thing (I'm not an engineer, my dad is and the software was availible via his contacts). This software designs some of the most sensitive systems used in fuel refining and other fancy far more dangerous and delicate stuff than a simple watercooling system. Anyway, this program determined that to keep my entire system running at 75 degrees F, I had to pump a total of 85 LPH. That's equivalent to 22.45 GPH, or roughly 7% of what the D5 puts out. So everything aside, the pump is so overpowered that it should be an almost negligible difference between series and parallel setups. Series is just easier and less prone to problems. The link below will take you to the results page so you can verify these results for yourself (note that these results are for full load i.e. max TDP)

https://docs.google.com/file/d/0B8J5LJlbVPUlNDUzM2I2ZTAtNjIzNy00YmIyLWEyMTEtMWE1NDBhZTJmNzAy/edit?usp=sharing

Additional note: this was all in parallel (all components had their own lines) and it caused me all sorts of trouble getting the flow rates to equalize since the CPU heatsink had a much more restrictive design. Changing over to series has fixed all the issues I had with the system while having no major effect on overall temperature (the CPU is actually cooler now than it was in a full parallel circuit).

 

[P1nnacle]

We have gotten so off topic, the main question was whether or not they needed two pumps. Which to reiterate OP, you do not.

 

[manofchalk]

I never proposed a fully parallel loop, only the GPU's, for the very reason of disparate flow. The parallel GPU array would be in series with the rest of the loop.
................ > GPU >.............
Loop -> <---------> Loop
................ > GPU >.............

Except the flow value that your working off is based on there being no restriction at all, the pump is just moving water between one bucket to another. If a D5 can pump .3GPM in a typical loop using only 1/1000th of its capabilities, then we should be running at flow rates of 300GPM which definitely doesn't happen.

But anyway, yes one pump will do you fine. IMO, running the GPU's parallel to each other will ensure that.

 

[P1nnacle]

I never proposed a fully parallel loop, only the GPU's, for the very reason of disparate flow. The parallel GPU array would be in series with the rest of the loop.
................ > GPU >.............
Loop -> <---------> Loop
................ > GPU >.............

Except the flow value that your working off is based on there being no restriction at all, the pump is just moving water between one bucket to another. If a D5 can pump .3GPM in a typical loop using only 1/1000th of its capabilities, then we should be running at flow rates of 300GPM which definitely doesn't happen.

But anyway, yes one pump will do you fine. IMO, running the GPU's parallel to each other will ensure that.

Yeah I screwed up and misread the rating. It's 317 GPH. Which is still much more than a necessary 85 LPH. And I'll reiterate why I think series is a better choice for the GPUs. It's simpler for someone who is new to watercooling, and the temperature difference is most likely going to be within 1-2 Degrees C since the flow rate of the D5 is so high. And if you are worried about restriction (which if you are you should be vouching for a full parallel system) I'll point out that once the system is full, basic physics takes care of the restriction for you.