Pump Planning Guide

Pump Planning Guide

So what if you prefer doing a little homework and planning of your pumping needs?  Rather than continue constantly updating my old estimator, I decided to write up this quick help guide that can be used with new data as it becomes available.  I’ve seen other people utilize the method and it makes it fairly simple. I also think it adds value in hopefully promoting pump curve and pressure drop curves as being standard performance specs in watercooling parts.  There is really no reason this sort of data can’t be included with the standard performance specs.

First you need to understand the basics of how pump curves and pressure drop curves interact and how the flow rate is determined.  The pressure head you see in pump specs is typically the maximum pressure, but unfortunately the pressure down the curve between these two points is really what you need which might be linear or along a wavy curve.

PumpSpecsYou also need the same for restriction in the form of a pressure drop curve which is also a curve as restriction increase rapidly as flow rate increases.

This is what you end up with when you add up all the pumps in series and all the restriction to create the two pump vs pressure drop curves:Pump101-01

Where the two curves intersect is what determines the flow rate (vertical green line).  It can be calculated mathematically, but it’s a bit detailed and more than what we really need for general pump planning.

A rule of thumb on flow rate is to try to achieve 1GPM although you can operate down to .7GPM without too much trouble, it really is just a general rule to help ensure you have enough flow rate to bleed the system easily.

Using that 1GPM target and to simplify the concept further, you can think about pressure gains and losses at 1GPM.  If your pump produces 5 PSI at 1GPM, you can then loose up to 5 PSI @ 1GPM in restriction and still maintain that 1GPM  flow rate or better.


Pump = Swiftech MCP 655-PWM @100%.


Pressure head of the pump at 1GPM = 4.6PSI.

Pressure drop of 10′ 1/2″ tubing and one reservoir = .3PSI @1 GPM

This means you still have 4.3PSI worth of pressure head to loose for the rest of the block and radiators.

Let’s take a typical 360 radiator like the XSPC RS360


Pressure drop of the radiator at 1GPM is 0.28 PSI

Now let’s look at the CPU block, let’s assume we have a Swiftech Apogee HD


Pressure drop of the CPU block at 1GPM is 1.70 PSI

So, the resulting pressure difference at 1GPM is +4.6 PSI (Pump) -0.3 PSI (Tubing & Reservoir) -0.28 PSI (Radiator) -1.7 PSI (CPU Block) = +2.32 PSI.   This means you will have more than 1GPM in flow rate and good to go!

That’s all there is to it, add up the pump pressure at 1GPM, subtract the sum of the pressure drops.


This is just a simple extraction of curves I had in my estimator specifically tabulating what the pressure rise (pumps) is or what the pressure drop (blocks/rads/etc) is at the 1 GPM mark in PSI.  The intent is to make this easily adaptable to new data and other internet data.


Here is a compilation of pump info I have collected over time.  I’m not going to guarantee it is correct, but it is what I had extracted from various sources and should get you in the ballpark.  This is also based on more than one test bed, so it’s only approximate.



This is the other stuff…again no guarantees here and a mixture of test rigs so values are only approximate:



What about two pumps in series?

Add up the pump pressures. A pump with 5PSI and a pump with 4PSI together in series will produce 9PSI.

What about two or three GPU blocks in parallel instead of series?

Take the pressure drop for one and divide it by the number of blocks.  A GPU block with a pressure drop of 1PSI at 1GPM will together in parallel with the same block results in 1/2PSI or 0.5PSI.  Of coarse the flow rate through the GPU blocks is also similarly split so if you get 1GPM in the system, two parallel blocks will each get 0.5 GPM so take that into consideration as well.  You can start seeing some GPU cooling degredation if you really bog down flow in parallel since it is possible to get it down in the .3GPM range.

What if I want the absolute best performance?

You can gain some performance with more than one pump, but it’s not as much as you might hope for.  Typically the CPU block and radiator performance curves relative to flow rate are not much more than 1-3 degrees or so beyond the 1GPM rule and many times much less than that.  It is a factor of flow rate gains and pump heat dump where more pump does increase flow and performance however it negatively impacts the performance by adding heat into the loop.  Generally you can see gains upwards of about 40W or so and much more than that can actually cause more harm than good.  10-25W worth of pumping power is typically plenty good.  More can be better, but more is not always better.  Two smaller 10-25W pumps in series are a good combo of providing near optimal pumping performance if you’re aiming to get every last degree out of your parts. Some have pushed to even more, but diminishing returns present themselves fairly quickly past about 40-60W worth of pumping power.

What if I have a different unit of measure than GPM and PSI?

There are many good conversion tools, here are a couple I found that may help you.

Flow Converter:


Pressure Converter:


Happy Pump Planning!

  1. Kristian says:

    GPU blocks in parallell.. a correction

    First, thank you for a fantastic work with this web site !!! 😀

    I found an error in following:
    “What about two or three GPU blocks in parallel instead of series?

    Take the pressure drop for one and divide it by the number of blocks. A GPU block with a pressure drop of 1PSI at 1GPM will together in parallel with the same block results in 1/2PSI or 0.5PSI.”

    The correct answer is 1/4PSI or 0.25PSI. for two blocks
    Since the flow is down 50% of each block+
    pressure drop grows by second order gives: dP = 1PSI*(1/2)^2
    for two blocks in parallell.
    in general case dP = 1PSI*(1/nof_blocks)^2

    Here is a comprehensive guide for pressure loss calulation and pumps:

    Click to access The_Centrifugal_Pump.pdf

    see page 82-88

    or google: centrifugal pumps grundfos

    Best Regards
    Kristian A.
    WC enthusiast and Process Engineer

  2. […] source: Water Cooling Estimator Spreadsheet looking through the pump guides, this one is good still good guide. [mods: can we upload files to the […]

  3. Thomas says:

    Hmm, can’t figure out how to get the PSI (at any flow rate), the stats given for the pump is Q-max 800L/h which would be 3.52gpm (US) (Are your gpm values US or Imp?), the other value given is H-max which is 4m, nothing about pressure (This is EK-DCP 4.0).

  4. Thomas says:

    Think I figured it out, 4m ~ 13ft ~ 5.63 PSI (I assume that is at the 3.52 gpm?). Is that how you do it?

    • Martinm210 says:

      Stats are usually max head at zero flow which is the left most point on the pump curve and max flow which is zero pressure. To get an approximate pressure at 1GPM you can divide the max pressure is PSI by the max flow in GPM. This isn’t perfect since many pump curves are curved, but it’s better than nothing.

  5. Thomas Archer says:

    I have a D5 Vario and a D5 Pwm. I have the Vario suspended from tubing and it’s very quiet. I want to add the Pwm to my loop for redundancy/increased pressure. I would have to mount it to the case. Would it be OK to have these 2 pumps in series? D5 Vario at full speed suspended, D5 Pwm with speed curve mounted to case. My thinking is the pump mounted to the case will be more noisy so it would be great if it could ramp up only when needed via pwm.

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s