pump placement

[basserdan]

Opinions please-

I'm putting a loop together with this config: res-pump-rad-gpu-gpu(4870x2 in series)-pump-rad-cpu

pumps= mcp655 V, mcp655 B
mcw60 (x2)

Is the pump after the gpu's a bad placement? Will I be starving the pump right after the gpu's? In my extremely confined mid-tower, this is the closest thing to an elegant arrangement I can come up with.

Before you ask, the mixed pumps are the result of my want of an rpm sensor (on the basic). I bought the vario a while ago.

[MpG]

If you're running two pumps, just make sure that the other pump is positioned right after the reservoir, and that you don't turn on the other pump until the water's actually reached it.

[FrostyPanda]

I would put another small res after the 4870x2 before the pump, just in case the flow is poor enough where you would want a little extra fluid to not "starve" the pump. I know you're cramped, but there's always room for some small res...a bay res may work well. I'm just worried that your pumps will have to suck the water rather than push it...

[basserdan]

I would put another small res after the 4870x2 before the pump, just in case the flow is poor enough where you would want a little extra fluid to not "starve" the pump. I know you're cramped, but there's always room for some small res...a bay res may work well. I'm just worried that your pumps will have to suck the water rather than push it...

I like this idea but, it will likely require a couple of 90's right before the pump to make it. I'll search for a possible small res and see what I can come up with. Any res suggestions?

thanks so far

[exe163]

Since you already have enough components for two loops, why not just make two loops with T-Line on the second loop?

[FrostyPanda]

I like this idea but, it will likely require a couple of 90's right before the pump to make it. I'll search for a possible small res and see what I can come up with. Any res suggestions?

thanks so far

well...the koolance 80mm ones are pretty small and will fit virtually anywhere. another option is to cut holes in the top of your case and place a res up there and then run tubing back down. Other decent ones are basically any bay res that will fit in a 5.25 bay. Check out PPC's for those...

[basserdan]

Just a thought but, if I have insufficient flow coming from gpu's, would a reservoir even very slowly, lose ground? I have a lack of imagination when it comes to trying to picture the dynamics of a cooling loop. For instance, I don't see how variations in flow rate could exist in a continuous path of a loop. I mean, the first molecule of water to pass through any given point in a loop is immediately followed by the very last molecule, if they could so be arranged, right? Ultimately the whole system moves no faster or slower than it can pass through the most restrictive point in the loop - again, right? I've never really considered it before. I really don't know the answer, it just occured to me. Comments please

[MpG]

You're quite correct. Liquid flow is constant throughout a closed loop, which makes a second reservoir redundant. Get the first pump running off the reservoir until the system is full and the coolant is circulating, then get the second pump working as well for higher flow rates.

[hotdun]

Just a thought but, if I have insufficient flow coming from gpu's, would a reservoir even very slowly, lose ground? I have a lack of imagination when it comes to trying to picture the dynamics of a cooling loop. For instance, I don't see how variations in flow rate could exist in a continuous path of a loop. I mean, the first molecule of water to pass through any given point in a loop is immediately followed by the very last molecule, if they could so be arranged, right? Ultimately the whole system moves no faster or slower than it can pass through the most restrictive point in the loop - again, right? I've never really considered it before. I really don't know the answer, it just occured to me. Comments please

You're quite correct. Liquid flow is constant throughout a closed loop, which makes a second reservoir redundant. Get the first pump running off the reservoir until the system is full and the coolant is circulating, then get the second pump working as well for higher flow rates.

It seems logical enough but it's just not true. Flow is not constant with pressure coming into play. Think about a loop using a Fuzion block with 3.5mm nozzle. If flow rate was static throughout, it would be extremely slow moving. That being said, the shortest loop always wins. Minimize bends and tubing length first. Component placement is less important.

Suggestion for the OP; Use a T-Line, small length of tubing and stopper after the GPU before getting to the pump.

[twwen2]

Personally i'd put the second pump straight after the first. You should get the same benefit (not sure there'd be any with just GPU blocks) but won't risk starving the 2nd pump.

If you do use both in the one loop, make sure you set the vario to the same speed as the non-adjustable one. I think that's setting 3?

[bigslappy]

Since you already have enough components for two loops, why not just make two loops with T-Line on the second loop?

YUP this is best !

it best NOT to have the CPU on the same loop as the Vid cards ,all yer doing is heating up yer CPU putting it on the same loop as the GPU's
the idea is to cool the CPU

[skinnee]

It seems logical enough but it's just not true. Flow is not constant with pressure coming into play.

Flow is constant, pressure changes. The pump starts out with the maximum amount of flow, as restriction is added (blocks, rads, reservoirs, and tubing) the flow rate lessens. Pressure can be altered anywhere in the loop, but flow cannot. When pressure is changed via a nozzle the flow rate does not change.

Edit:
Think of a garden hose... When you turn the spicket on half way, that is your flow rate. Now put your thumb on the end of the hose so it sprays further, does the handle on the spicket magically open, no...you only altered the pressure. In a closed loop the amount of restriction is constant, therefore flow rate is constant.

[twwen2]

Skinnee isn't it the other way around?

When you say "flowrate", do you mean "velocity"? Because the velocity of the fluid will most definately change throughout a loop, depending on the diameter of the hose/fitting/passage. That's how nozzles work. The pressure (push from the pump) is constant but the flow (velocity of fluid) changes with the smaller diameter, speeding it up and slamming down onto the baseplate.

Anyone willing to back me up on that?

[alacheesu]

Skinnee isn't it the other way around?

When you say "flowrate", do you mean "velocity"? Because the velocity of the fluid will most definately change throughout a loop, depending on the diameter of the hose/fitting/passage. That's how nozzles work. The pressure (push from the pump) is constant but the flow (velocity of fluid) changes with the smaller diameter, speeding it up and slamming down onto the baseplate.

Anyone willing to back me up on that?

Volumetric flowrate, Q, is a function of velocity, v, and cross sectional area, A.
Q = v*A

Q is constant throughout the loop, but v and A can vary. A decrease in area (like a nozzle) would mean an increase in v.

[Bobly]

+1 for dual loops if it fits, that way you can keep a nice cool loop for your cpu and a slightly warmer one for your gpu.

[inCore]

I was just going to add that flow is a measure of the amount of liquid that passes per unit time.

If there is a small hole to pass through, the liquid will go faster but the number that comes out of the equation will be exactly the same.

[basserdan]

I am aware of the dual loop concept and it makes sense. I just felt like it might require even more from my critical shortage of space. But, I am now reconsidering whether I can make it work. I have an mcr220 and a 320. I'd assume my 4870x2 should go to the 320 because, I think it runs on plasma. Does anybody recommend the reverse- 220 on 4870x2 and the 320 on my cpu(3.46GHz @ 1.36v)?

[AndrewZorn]

I'd rather have the bigger radiator on the CPU, the small temp differences are not as important on a video card.

Definitely do two loops.

[basserdan]

I'd rather have the bigger radiator on the CPU, the small temp differences are not as important on a video card.

Definitely do two loops.

Not from doubting your advice but, is it possible that there is enough cooling capacity in an mcr220 for a 4870x2? It's on air for now and I've never seen anything put out so much heat. I do have a place to hide a 120(rad) maybe this would do it.

[skinnee]

An MCR220 can dissipate around 300w efficiently with 3000rpm fans and the 4870x2 appears to generate around 280w (google searches). Here are the charts from Martinm210 on the heat dissipation on the MCR220.

[AndrewZorn]

I don't get it... looks like stock D12SM12s should be enough dissipation for the 4870x2 for 'average performance', which is still leaps and bounds over the stock cooler... right?

Maybe switch them around... I guess I was just hoping that you could get the better temps on the CPU and not worry about the GPU as much.

[skinnee]

Yes, my mistake in typing, the D12SM-12's would work fine as well.

[hotdun]

My bad.....flow rate is a result of the accumulation of restriction and pressure drop of the system as a whole. Flow rate is constant while flow speed and pressure changes.

[skinnee]

My bad.....flow rate is a result of the accumulation of restriction and pressure drop of the system as a whole. Flow rate is constant while flow speed and pressure changes.

[basserdan]

Thanks Skinnee

If I understand the graphs, we get: Graph2- between a 14 1/2 and 5 3/4 deg change in temp. based on fan speeds, while cooling 300 watts. This would seem acceptable. I'll need to look at Martinms testing and try and familiarize myself with the terminology and methodology so I can understand the parameters here. But from what I can gleen so far, I should be fine for my gpu's on the 220 without the addition of the 120.