1 - 20 of 114 Posts

#### charliehorse55

·
Joined
·
5,945 Posts
Discussion Starter · ·
Choosing the Correct Water Cooling Pump
A guide by charliehorse55

What flow rate do I need in my loop for the best performance?
Most loops achieve maximal efficiency around 1.0 GPM. Increasing the flow rate beyond this point as little to no effect on temperature for most modern blocks, and the extra heat dump from the added pumps required to reach higher flow rates can actually have a larger impact than the increase in flow. This is because the heat capacity of water is incredibly high. At 1.0 GPM of flow, it takes over 250W of heat to raise the water a single degree.
Heat capacity of water = 4187j per kg
1.0 GPM of flow in litters per second = 0.063 kg/second
4187j*0.063 kg/second

A joule is 1 watt for 1 second, so the units of time cancel out:

4187W * 0.063 kg = 264W

264W to raise a 1.0 GPM flow of water by a single degree

Let's assume you are running a CPU + dual GPU loop. With a heat dump of around 200W per component, the last component in the loop will have the water entering the block at 1.5C higher than the first. If you increased your flow rate to 2.0 GPM, this would drop to 0.75C, lowering the GPU temperature by 0.75C. This is not a huge gain, but it is appreciable, until you factor in the added heat dump from the pump. To get to 2.0 GPM on a dual GPU + CPU loop you would need to add another 30W of pump at a minimum. Even with a quad radiator, this is going to increase your water-air delta by about 0.2-0.5C, effectively raising the temperature of all of your blocks by that much. All of a sudden these small gains become absolutely meaningless.

If the heat capacity of water is so high, why can't I have less than 1.0 GPM?
At low flow rates, the performance of your water blocks drops dramatically as the flow inside of them switches from turbulent to laminar. Turbulent flow occurs when a lot of water is going through a small pipe. When water is flowing this way, it creates random vortices that ensure the water is constantly mixing, and that fresh, cold water is constantly touching the waterblock. If the flow rate is low enough, the water switches to flowing in a laminar way. When water flows in this way, it does not mix as much, and the water on the edges may be significantly warmer than the water around it. This is bad for performance, as heat is not transferred away from the block as quickly.

Older water-cooling equipment used very open blocks with large openings for water passing through, making it very easy for a laminar flow to develop. This is why so many older blocks are incredibly flow dependent, requiring high flow rates to get good results. Modern blocks use very small channels, so that turbulent flow occurs at much lower flow. Almost all waterblocks today are designed to have turbulent flow at low flow rates, and as such the temperature gains from higher flow rates have largely disappeared. Most blocks will continue to perform down to about 0.6-0.7 GPM, but it's a good idea to have around 1.0 GPM to have some headroom.

How do I know which pump I need to achieve 1 GPM?
Pumps are rated on a pressure/flow curve. At lower flow rates, they can produce more pressure, and at higher flow rates, less pressure. The blocks in your loop provide restriction that scales in the opposite direction, as higher flow rates lead to a greater pressure difference being required. The flow rate of your loop is determined by the intersection of these two lines:

A more detailed explanation of this graph can be found on xtremerigs.

Can't you simplify this?
Yes. You don't have to find the exact intersection of those two lines because you don't need to know the exact flow rate of your loop. You just have to make sure that is greater than 1 GPM. You can add up the pressure drop created by our components at 1 GPM, and then make sure you have a pump capable of that pressure at 1 GPM. Finding the right pump becomes as simple as adding up a few numbers!

I've made a rough guide to get a general idea of how much pumping power you need. The data on the pumps is deadly accurate, however as there are many, many different types of blocks it would be mundane to list all of the pressure drops (and there isn't data available for everything!). Feel free to substitute any value in the first chart with ones specific to your parts. (You can find this data on review websites).

First, add up the restriction of your loop (these values are safe averages)

Pressure Drop figures for components
Radiators: 0.25 PSI each (the size of the radiator has no real effect)
CPU Block: 0.9 PSI
GPU block: 0.9 PSI
Motherboard block: 2.0 PSI (fullcover)
Fittings: Negligible (An extreme setup with 10+ 90 deg fittings might add 0.5 PSI in total)
Tubing: 0.5 PSI per 10 feet for 3/8" ID, negligible for 1/2" ID.
Reservoir: Negligible

If you run your GPU blocks in parallel, divide the PSI drop by the number of blocks, so if you have 3 blocks in parallel you would only add 0.3 PSI to the total.

*Exceptions*
HW Labs GTX Radiators are around 0.7 PSI each
Anything by aquacomputer is about double the normal pressure drop

Now that you have the total pressure drop for your loop, find a pump that can handle at least that much pressure at 1 GPM of flow.

Maximum Head Pressure at 1.0 GPM

Iwaki RD-30: 13.72 PSI
MCP50X Stock: 6.54 PSI
MCP35X Stock: 6.52 PSI
MCP355 + Top: 6.14 PSI
MCP655 + Top: 4.95 PSI
MCP655 Stock: 4.72 PSI
MCP355 Stock: 4.63 PSI
MCP350 + Top: 4.48 PSI
DCP 4.0 Stock: 3.83 PSI
XSPC 750: 3.59 PSI
655-B + EK Top: 3.38 PSI
MCP350: 3.25 PSI
655-B Stock: 3.23 PSI
DCP 2.6: 2.55 PSI
Eheim 1250: 2.41 PSI
DCP 2.2: 2.08 PSI
Eheim 1048: 1.78
Eheim 1046: 1.02 PSI

Double pump setups are usually almost exactly double the performance of a single pump. Keep in mind, you should also choose your pump based on how loud it is, how much power it uses and how it will fit into your case. This is merely a guide to help you choose a pump that is powerful enough to handle your loop.

Feel free to ask anything, I will clear up any questions you may have and potentially add it to this post.

·
Joined
·
829 Posts
sub'd

#### BloodyRory

·
##### Registered
Joined
·
926 Posts
Can't wait to read this, sub'd also.

·
Joined
·
326 Posts

#### saer

·
##### Registered
Joined
·
829 Posts
This is exactly what everyone just getting into watercooling needs to read(me). I didn't feel confident with my setup up until this very moment.

Great work! this needs to be stickied

+rep

#### Chunky_Chimp

·
##### Super Moderator
Joined
·
19,303 Posts
Super. You should add the Aquacomputer/Eheim and Iwaki pumps to the list, since some enthusiasts do use those.
I have also asked Juggalo23451 to determine sticky eligibility; he will be responsible for making it so in that case.

#### Nexus6

·
##### Registered
Joined
·
2,182 Posts
What are the restrictions on RAM and Mobo blocks?
This is very informative. Awesome work

#### tsm106

·
##### Banned
Joined
·
21,557 Posts
Look at that lil 35x go with that stout pressure.

#### charliehorse55

·
Joined
·
5,945 Posts
Discussion Starter · ·
Quote:
Originally Posted by Chunky_Chimp;14815486
Super. You should add the Aquacomputer/Eheim and Iwaki pumps to the list, since some enthusiasts do use those.
I have also asked Juggalo23451 to determine sticky eligibility; he will be responsible for making it so in that case.

Iwaki RD-30 = 13.72 PSI
Eheim 1250 = 2.41 PSI
Eheim 1048 = 1.78
Eheim 1046 = 1.02 PSI

The Eheim pumps weren't really designed for watercooling, more for aquarium pumping so they have a very low pressure but high flow.

Quote:
Originally Posted by Nexus6;14815506
What are the restrictions on RAM and Mobo blocks?
This is very informative. Awesome work
A full cover motherboard block is usually around 1.5 PSI or more, with some of EK's offerings having over 3 PSI of drop!

Here is testing done by Skinnee for the ASUS Rampage III blocks:

I can't find any data on RAM blocks but I would imagine blocks that cover the top of the multiple ram slots (think EK Dominator block) would be very unrestrictive wheras per stick solutions such as Koolance's would be more restrictive (1+ PSI)..

#### Segovax

·
##### Banned
Joined
·
5,701 Posts
I'm curious as to what effect the MCP35X Reservoir Rev 2 has on the MCP35X pump if any.

#### charliehorse55

·
Joined
·
5,945 Posts
Discussion Starter · ·
It has little to no effect. From Martin's testing of the 35X:

As you can see there is little difference around 1 GPM.

#### Faster_is_better

·
##### Aren't this fancy!
Joined
·
7,226 Posts
Very nice, sticky would be good. I know If I ever were to get into WC, I'd want to consult this, and not have it buried away in OCN's threads

#### iamwardicus

·
##### Nerd / Geek
Joined
·
1,681 Posts
Sticky would be good. I love the comment about aquacomputers being about double the PSI My AM3 block theoretically has 2.6 PSI of pressure drop... Glad I'm not having any other blocks in my loop when I finish it for my planned 1gpm flow rate

#### charliehorse55

·
Joined
·
5,945 Posts
Discussion Starter · ·
Yeah, I don't know what aquacomputer block you have but the Kuplex Cryos HF is only about 2.0 PSI. I might change the recommend PSI drop for CPU blocks to 1.0, many blocks are below 1 PSI these days.

#### iamwardicus

·
##### Nerd / Geek
Joined
·
1,681 Posts
I have the Kryos HF for AM2/AM3. Still waiting for the money for a pump & res yet. Attempting to get a 2nd PA120.2 or a GTX 240 for future expansion. Regardless I still know that a 655 is more than enough for a cpu only loop and at >1gpm flow.

#### bmaverick

·
##### Registered
Joined
·
1,362 Posts
A sweet spot for pumps tends to be the old faithful MCP350 + Top giving the 4.48 PSI. The MCP355 tends to run hotter when worked harder vs. the MCP350 working harder. 4.48 PSI seems swell in a balanced loop for 1.0-1.5gpm of flow. Not too fast else the thermal transfer looses ground.

Loved the charts and work here.

1 - 20 of 114 Posts