Do I Need 2 Pumps Per Loop?

[JLambeth87]

Here is my build thread: http://www.overclockers.com/forums/showthread.php/787677-600lbs-Steel-I-Beam-Desk-Computer-Build

I decided to put the pump next to the radiators and am concerned with how much tubing there will be to get from the radiator to the block and back. Roughly 16ft worth of tubing per loop with 2 uphill sections at 3ft each. Should I add a second EKWB D5 pump to each loop on the opposite side as the first pump?

Thanks in advance.

Josh

 

[Lochekey]

I would say that you are ok with just one pump per loop.

Just curious but why are you doing 2 seperate loops.

Based on your other thread you plan to use 2x 480 rads. Each of these rads can typically dissipate 400 watts of heat. This means that your cpu loop will have excessive cooling capacity since your cpu is a 165w you even overclocked it will not exceed 250w so that means you have an extra 150w capacity to play with. On the other loop you are running 270x 1080tis that have a tdp of 250w each so your cooling loop on that side has 500w on a 400w rated rad. If I was building that I would combine the loops to take better advantage of the radiators to equalize heat load better.

 

[Nebulous]

From the looks of that build and the distance the water has to travel along with it being 2 loops, your best bet would be an Iwaki magnetic drive pump: https://www.iwakiamerica.com/products/MD.htm. The D5's, tho an excellent pump, will have trouble trying to cope with the distance & uphill travel for long periods. The D5 will eventually burn itself out.

Back in the early era of watercooling, lots of members using chevy heatercores and Cathar's blocks used the Iwaki pumps for their superior pressure/head and the ability to run 24/7 without so much as a hiccup. The Eheim pumps were another excellent choice as I used those for my early builds.

Of course alot of my colleagues will state the Iwaki pump is wayy overkill, lol.

 

[Dolk]

Uhhh Yeah Nebulous, why not just go with a Siftwech MP50/60?

 

[JLambeth87]

I would say that you are ok with just one pump per loop.

Just curious but why are you doing 2 seperate loops.

Based on your other thread you plan to use 2x 480 rads. Each of these rads can typically dissipate 400 watts of heat. This means that your cpu loop will have excessive cooling capacity since your cpu is a 165w you even overclocked it will not exceed 250w so that means you have an extra 150w capacity to play with. On the other loop you are running 270x 1080tis that have a tdp of 250w each so your cooling loop on that side has 500w on a 400w rated rad. If I was building that I would combine the loops to take better advantage of the radiators to equalize heat load better.

I am planning on way over-clocking the 7980XE. I have been reading in multiple places that the numbers I am planning on getting to (was binned at 4.2Ghz) will be pulling around 600w of power. Wouldn't this make the 480 rad a good choice to be on it's own loop? I am also taking into account the 16+ ft of copper piping that will be in open air which would also help with cooling correct?

@Nebulous I don't think I want to go that route, but if I did, what version of the pump would you recommend? Thanks!

Josh

 

[Dolk]

Lol What?? 600W of power through a CPU? Nah that ain't happening. Maybe 300W at Peak, but not sustainable.

1.45Vcore @ 150% (45A Typical) current capacity would be roughly 163.125W, Add in PCI-E and Mem Controller, and you may get another 100-120W. So again maybe 300W total.

One of the reviews I found, and others could help me if they find more:

https://www.gamersnexus.net/hwreviews/3066-intel-i9-7980xe-7960x-thermals-power-review

There is no description of their test methodology and how they procured the measurement of the reading from the ESP12V cables. You can't just take a DMM to the socket and measure power/voltage. Additionally methods like these come at a high derating strike, think +/- 20% at times. Where those measurements sustained or peak readings? I'd imagine that they were peak because a DMM or hand reader does not have the granularity to see sustained. I'm not 100% buying this just yet.


https://linustechtips.com/main/topic/839760-intel-core-i9-7980xe-hits-845w-at-49ghz/

Wall measurement, you can lop off a bunch with these readers as they bundle the total power draw from the PSU, and not what is being actively consumed by the CPU. The PSU is going to burn more to source the 12V into the CPU VRMs, of which those will burn off a lot because of the high heat they are running at. Again, no more than 300W max at the CPU.

 

[Woomack]

7900X at 4.7GHz/1.2V ( what is about max for wc at full load ) = ~300W from the wall ( measured by me ). Considering how 7900X is acting ( and some other users experience at some more cores ), I don't think there will be any special difference between something like 1x 480 and any larger rads if we are talking about CPU OC. Issue will be that waterblock won't be able to transfer so much heat to water in loop, in given time. Difference between 1 and 2 rads in this case will be like lowering loop temps by maybe 5°C what won't really help in overclocking while may lower fan speed a bit so I count it will only affect the noise. Something like that I saw while using 1x 360 rad vs 2x 360. 7980XE will be maybe 400W tops at similar voltage but I doubt you will keep it at the same voltage as lower cores because of mentioned heat in small area what is hard to transfer by waterblock so at lower voltage also max wattage will be lower.
2nd pump is useless and depends on how loop will go, may cause issues.

Just some thoughts about this topic as 600W is not possible and building huge loop with many rads is only a myth and is not really helping as many users think.

 

[JLambeth87]

Thanks for the info. I'll keep the two 480s I have now. Not planning on adding anything more.

I did purchase 2 more ekwb d5 pumps and 2 more 250ml radiators. I'll add them to the other side of the desk to push water back to the radiators. I think I'll have each loop setup so the pump on the radiator side pushes water to the waterblock side then goes through the blocks and then hits the next pump and gets pushed back to the radiator before going around again. Think this would work or would you recommend the pump being right before the block and again right before the radiator?

Also I'm wanting to add one of these style of flow meters to each loop. Any suggestions on ones you think I should use?

https://www.amazon.com/gp/aw/d/B01A...flow+meter&dpPl=1&dpID=31ltnE0iusL&ref=plSrch

Thanks!
Josh

 

[GTXJackBauer]

600w is possible if you're running a X99 Classified being pumped by TWO CPU Power headers and you're OCing on LN2. No way a custom liquid cooled loop can handle that kind of load.

As for pumps, if you're running two pumps per loop or one single big loop which I would recommend, run them in serial with a serial top. Just make sure to the run them in sync.

 

[Lochekey]

My concern with the two loops was not the cooling of the cpu but the cooling of the 2x 1080 tis.

 

[JLambeth87]

My concern with the two loops was not the cooling of the cpu but the cooling of the 2x 1080 tis.

Do you think i should add a second radiator or do you think I should really do a single loop system with just 2 pumps? If I have a res right before each pump (4 in total between the two loops) wouldn't that help the flow be consistent by not starving a pump? I will draw up a diagram of how I see this working for each in-case I am not making sense.

The two pumps on each loop will be controlled by the same header (IE both CPU loop pumps controlled off the CPU fan header and both pumps for the GPU's will be off another header). I do think that having almost 16ft of copper tubing exposed to the air between the GPU's and the radiators will help reduce the overall loop temp since copper will conduct heat so well. IDK I am open to suggestions.

Thanks in advance.

Josh

 

[GTXJackBauer]

This blueprint is starting to sound more confusing then it should be. Passive cooling doesn't remove that much heat load then say a rad with its fans does. You won't get better cooling or more flow with more reservoirs or more tubing. The shorter the loop, the better. The loop will reach a equalized state or what we call, equilibrium. It will even out and hit its thermal wall.

I think more reading should be done to clear some confusion.