2 pumps inline

[rpalamara]

I have a Storm and 2 Silverprop Fusions HLs that I am putting in a system along with a 2x120 radiator. I was looking at the MCP655 as I wanted to hook it up the the PSU. I am afraid that it will be right at its limit and might decrease performance.

Will installing 2 pumps in series help me with my results? Or do I HAVE to upgrade to a large Eheim (1260).

Thank You

[9mmCensor]

I belive it is best to run one powerful pump.

Look into an Iwaki.

[mikoto]

I want to refer you to this article:
http://www.procooling.com/index.php?...s&disp=38&pg=1

Eheim 1250 --> 79" @ 28 Watts power consumption. AC pump
Eheim 1060 --> 122" @ 50 Watts power consumption. AC pump
MCP655 --> 86" @ 9 Watts power donsumption. DC pump
I cannot comment on 1260.
Iwaki MD20RZ --> 169" @ 50 Watts power consumption. AC pump.

[bobo5195]

two pumps in seris act like a single pump with double the pressure.
two pumps in parallel act like a single pump with double the flow albiet some more pressure drop due to the split tubing.

[rpalamara]

Bobo, from what I understand it wont increase the max flow rate, but for a given system the flow rate will increase over 1 pump? If you look at a graph of the flow rate vs pressure of a pump and a water block system, the system will run at the intersection of the two functions. Since the pressure is increased, the two lines will intersect at a later point of increased pressure and flow rate. Please correct me if I am wrong.

http://www.procooling.com/index.php?...ticles&disp=91

[Long Haired Git]

My approximator works on pressure, but with the PQ graphs unadjusted for two pumps. I trust this is right.
I do not have any PQ information for the HL fusion GPU blocks. I have assumed they are similar to the LRR block, as they employee the same technique for cooling (restrictive fins with impingement over core).

Two LRR's in series each on 70W (I assume you're cooling high end GPUs?)
MCR220QP radiator with 2 fans on 7v
Approx 1.5m of 1/2" tubing

Flow rate = 6.1 LPM
Water temp at rad exit = 7.1°Caa (degrees celsius above ambient)
GPU 1 = 19.7°Caa
GPU 2 = 19.9°Caa

Add in a second MCP655 @ 5:
Flow rate = 8.1 LPM
Water temp at rad exit = 7.7°Caa (degrees celsius above ambient)
GPU 1 = 19.9°Caa
GPU 2 = 20.1°Caa

So, I get it as being worse. This, in my ever humble opinion, is because the MCP655 is already working at a high flow rate where it dumps the most heat into the coolant. As we know, pressure head is related to the square of the flow rate, and so the tiny extra flow rate results in much more head, hence only an extra 2 LPM.

Swapping to an Iwaki 20RT gives the best temps according to the figures I have (heat dump for the Iwaki is an estimate as none available) being 7.2LPM and GPU temps of 19.3°Caa and 19.5°Caa.

To give you an idea of the restriction:flow-rate curve and how little flow rate matter with modern blocks and radiators, a single MCP350 pump gives "only" 3.8LPM but the GPU temps are 20.1°Caa and 20.3°Caa

If it were me, I'd take the single MCP655 and be darn happy.

BTW, with the single MCP655 @ 5 I have the following for the LRR block. Anyone who can confirm/deny and compare/contrast to the Fusion would be great to hear from!

Flow = 6.1078 LPM
Head from one block incl Y joiner and two 5cm sections of 1/2" tubing: 1.2315m of H20

[rpalamara]

Thanks for the info. A few questions for you. What thermal values did you use for the Swiftech radiator and do you have a graph for it. Did you factor in the CPU when you calculated everything out? What do you have as the wattage for the CPU? I have a AMD 3700 that I figured will produce 150-200 Watts of heat. That combined with the pressure drop of the Storm made me think that the added pressure would help.

[rpalamara]

I was considering three pump setups: a single MCP655, 2 MCP655s, or a Iwaki RD-30. The Iwaki and two MCP655s dump a similar amount of heat, but the Iwaki is a more powerful pump (needs a separate power supply though). I did some calcs and came up with this:

Tell me if you find anything wrong
(based numbers off reviews and specs from manufacturers)

Using a Storm and 2 Silverprops (used Swiftech MCW55 numbers) and a MCR220

MCR220 (from swiftech site)
@ 1 GPM = .0375 C/W
@ 1.5 GPM = .035 C/W
@ 2 GPM = .034 C/W

Storm GPU block
@ 1 GPM = .13 C/W
@ 1.5 GPM = .12 C/W
@ 2 GPM = .11 C/W

1x MCP655 will run at 1 GPM and produce 15 Watts
2x MCP655 will run at 1.5 GPM and produce 30 Watts
1x Iwaki RD-30 will run at 2 GPM and produce 35 Watts

Now I am estimating that the CPU will produce 150 Watts and the two GPU cores (volt moded) will produce 180 Watts combined

1x MCP655
150 Watts x .13 = 19.5ºC
345 Watts x .0375 = 12.9ºC
19.5 + 12.9= 32.4ºCaa for CPU

2x MCP655
150 Watts x .12 = 18ºC
360 Watts x .035 = 12.6ºC
18 + 12.6= 30.6ºCaa for CPU

1x Iwaki RD-30
150 Watts x .11 = 16.5ºC
365 Watts x .034 = 12.4ºC
16.5 + 12.4= 28.9ºCaa for CPU

If the CPU is producing 200 Watts

1x MCP655
200 Watts x .13 = 26ºC
395 Watts x .0375 = 14.8ºC
26 + 14.8= 40.8ºCaa for CPU

2x MCP655
200 Watts x .12 = 24ºC
410 Watts x .035 = 14.4ºC
24 + 14.4= 38.4ºCaa for CPU

1x Iwaki RD-30
200 Watts x .11 = 22ºC
415 Watts x .034 = 14.1ºC
22 + 14.1= 36.1ºCaa for CPU

Due to the amount of heat I was even thinking about getting a 120.3 PE. From what I had read the numbers for this radiator are approx 30% better than the 120.3 HE. So these are estimations off of Cathars numbers for the 120.3 HE.

If the CPU is producing 200 Watts

1x MCP655
200 Watts x .13 = 26ºC
395 Watts x .023 = 9.1ºC
26 + 9.1= 35.1ºCaa for CPU

2x MCP655
200 Watts x .12 = 24ºC
410 Watts x .022 = 9ºC
24 + 9= 33ºCaa for CPU

1x Iwaki RD-30
200 Watts x .11 = 22ºC
415 Watts x .021 = 8.7ºC
22 + 8.7= 30.7ºCaa for CPU

These seem to be big differences in temps through the different pumps. I realize that it would be a waste if the heat produced wasn’t so high, but from asking around and estimating off of some white papers this was what I came up with. The final CPU block is going to eventually be a G7 Storm but these are the numbers I have for now.

Any input as to whether these numbers are in the ballpark or way off, or whether I am calculating things correctly would be greatly appreciated.

Thank
Ryan

[bobo5195]

Originally Posted by rpalamara

Bobo, from what I understand it wont increase the max flow rate, but for a given system the flow rate will increase over 1 pump? If you look at a graph of the flow rate vs pressure of a pump and a water block system, the system will run at the intersection of the two functions. Since the pressure is increased, the two lines will intersect at a later point of increased pressure and flow rate. Please correct me if I am wrong.

http://www.procooling.com/index.php?...ticles&disp=91

Agreed but i was talking about the pumps not the system. Doubling the pump flow rate will not double the system flow rate. I was trying to give a picture of pump not system performance.

Its a theoretical thing. The effective pump graph made by two pumps in parallel is the same as the same as the pump graph made for one except the effective pump flow rate is doubled. the other case of two pumps in seris produces double the effecitve head but no extra flow. In real life this is not the case because the pump theory is not perfectly in line with how real pumps function.

In fact under certain cases the extra resistance of parallel tubes and extra head may lower system performance.

[rpalamara]

Yeah I understand that the max flow rate will not improve.

Thanks

[Long Haired Git]

Originally Posted by rpalamara

What thermal values did you use for the Swiftech radiator and do you have a graph for it.

Its on their site.

Originally Posted by rpalamara

Did you factor in the CPU when you calculated everything out? What do you have as the wattage for the CPU?

Clearly stated as:
Two LRR's in series each on 70W (I assume you're cooling high end GPUs?)
So nope, didn't include a CPU at all. Whoops!

Originally Posted by rpalamara

I have a AMD 3700 that I figured will produce 150-200 Watts of heat.

Based on what? I saw most AMD CPUs produce 89W of heat. You'd need to increase the freq by 40% and the volts by 40% to double wattage to 188W. How about I use 110W for the CPU and you let me know why you think it would be so high.

Originally Posted by rpalamara

That combined with the pressure drop of the Storm made me think that the added pressure would help.

Yes it would.
Let me digest your figures and I will re-approximate mine and see.

[rpalamara]

Originally Posted by Long Haired Git

Its on their site.Clearly stated as:
Two LRR's in series each on 70W (I assume you're cooling high end GPUs?)

I have 2 7800 gt's that are volt moded, bringing the Vcore up to 1.6 volts. I was told that it could have brought the heat disipation up to 90 Watts per core, but 70 Watts is probably in the ballpark.

Originally Posted by Long Haired Git

So nope, didn't include a CPU at all. Whoops!Based on what? I saw most AMD CPUs produce 89W of heat. You'd need to increase the freq by 40% and the volts by 40% to double wattage to 188W. How about I use 110W for the CPU and you let me know why you think it would be so high.

I estimated a graph off of AMD's white paper using an exponential function. This graph was a quick estimate and really had no hard data on overclocked numbers. I have spoken with a few people that put the estimate more in line with your numbers. Since my CPU runs at 3000 Mhz stock, I figure that 3.1-3.2 Ghz at 1.6 volts is reachable.

The fact that my estimates on the CPU are wrong really makes me happy.

[Long Haired Git]

Next issue is that Swiftech Storm C/W graphs do not include the TIM. I thus use Robotech's testing from systemcooling.com (with permission).

So the C/W for the Storm you're using is out.

Right.

Single MCP655. 4.4 LPM. Dumps 13.3W into the loop.
Rad MCRP220QP with 2 fans @ 12v: C/W = 0.0361. Coolant at exit is 8.65°Caa
Storm on 110W. C/W = 0.1404. CPU temp = 24.1°Caa
GPU 1 on 70W. C/W = 0.186. GPU temp = 22.0°Caa
GPU 2 on 70W. C/W = 0.186. GPU temp = 22.25°Caa
Approx 1.5m of tubing

Add in a second MCP655
Dual MCP655. 6.0 LPM. Each dumps 13.73W into the loop.
Rad MCRP220QP with 2 fans @ 12v: C/W = 0.0348. Coolant at exit is 9.0°Caa
Storm on 110W. C/W = 0.136. CPU temp = 23.9°Caa
GPU 1 on 70W. C/W = 0.18. GPU temp = 21.8°Caa
GPU 2 on 70W. C/W = 0.18. GPU temp = 22.0°Caa
Approx 1.5m of tubing

So, 0.2 degrees better. W00T!
The coolant temp is still hot, even with dual fans @ 12v.
Instead of the MCP655, lets add in another MCR220QP with fans @ 12v.

Single MCP655. 4.25 LPM. Dumps 13.29W into the loop.
Rad 1 MCRP220QP with 2 fans @ 12v: C/W = 0.0363. Coolant at exit is 4.54°Caa
Rad 2 MCRP220QP with 2 fans @ 12v: C/W = 0.0363. Coolant at exit is 4.12°Caa
Storm on 110W. C/W = 0.141. CPU temp = 19.6°Caa
GPU 1 on 70W. C/W = 0.187. GPU temp = 17.5°Caa
GPU 2 on 70W. C/W = 0.187. GPU temp = 17.8°Caa
Approx 1.5m of tubing

You can see that makes a much bigger difference.

In fact, I'd drop the fans back to 7v and suffer the one degree drop and instead enjoy the quieter fans:
Single MCP655. 4.25 LPM. Dumps 13.29W into the loop.
Rad 1 MCRP220QP with 2 fans @ 12v: C/W = 0.05. Coolant at exit is 6.4°Caa
Rad 2 MCRP220QP with 2 fans @ 12v: C/W = 0.05. Coolant at exit is 6.0°Caa
Storm on 110W. C/W = 0.141. CPU temp = 21.5°Caa
GPU 1 on 70W. C/W = 0.187. GPU temp = 19.4°Caa
GPU 2 on 70W. C/W = 0.187. GPU temp = 19.7°Caa
Approx 1.5m of tubing

[rpalamara]

When I figured everything out I had estimated off of Cathars graph of the G4 and the MCP 655 numbers. So when it worked out it came to:

1 MCP655 ran at 1 GPM= 3.8 LPM
2 MCP655 ran at 1.5 GPM= 5.7 LPM

At those numbers I get a difference of 1.25 C. Where did you get your numbers from. They seem to be much more precise than mine. Are you estimating them off of charts or do you have a graph that you are tracing?