Flow rate point of Diminishing Returns

[VanGoghComplex]

I've read somewhere that 1GPM is about the highest it's "worth" tuning your loop to. Is this a reliable figure? Is it that cut and dry, or does it vary based on other factors?

I ask because I'm about to install my first flow meter, and would like to set up my Aquaero to maintain a "good" number. Hoping that the D5 I'm using with two blocks and one rad can be allowed to spin relatively slow for silence.

 

[Zarathustra[H]]

I've read somewhere that 1GPM is about the highest it's "worth" tuning your loop to. Is this a reliable figure? Is it that cut and dry, or does it vary based on other factors?

I ask because I'm about to install my first flow meter, and would like to set up my Aquaero to maintain a "good" number. Hoping that the D5 I'm using with two blocks and one rad can be allowed to spin relatively slow for silence.

I've never bothered measuring flow, as I viewed it as just another thing that could break.

(That, and putting a flow meter in the loop makes the loop more constrictive and thus reduces flow.)


In my application (I have an older non-PWM pump, with 5 settable speed settings on the bottom by turning a dial) I found that when I changed the dial the lower speeds had a lower pitch, and the higher speeds had a higher pitch, but that the sound level remained just about identical regardless of pump speed.

Based on this, I just pinned it, and left it there. I determined there was no reason to mess with it. If max flow isn't increasing noise, why not just run it there all the time?


That said, I feel like while "1GPM" might be a nice round number, that every loop is different. Some rads and blocks may do better at higher or lower flow rates. Best way is probably to just test it.

I'd install it with manual control first, run prime95 and unigine heaven (or something else like this to load up the GPU) at the same time, to emulate a max load condition, start out with max pump speed and let the system reach thermal equilibrium. Then bit by bit lower the pump speed, letting it sit for a bit between each increment to reach equilibrium, and then when you see the temp start rising, stop and go back to the previous setting.

This way you know you have the best setting for your loop, and don't have to rely on questionable rules of thumb.

 

[cyberguyz]

1.0gpm net when all of the components are in place is a pretty good target to hit. Getting a rate faster than that will only give back incremental performance increases the components in your loop kinda the 90%-10& rule. If you add more components, try to get to that number but don't stress about getting more flow. Of course if you are getting better than 1.0gpm, it's all good. It certainly wouldn't hurt.

On the other side of that coin if your pump is giving 1.0gpm with no restrictions in place, it is junk. Get a better one

 

[Zarathustra[H]]

Of course if you are getting better than 1.0gpm, it's all good. It certainly wouldn't hurt.

You know, I was just thinking about this after I posted my post above.

I run my pump at full blast, because I figured, why not? It doesn't make a difference noise wise, so why not just run full blast all the time?

But I'm wondering, are there any downsides? I remember from when I leak tested my system without fans running that just between the friction in the loop and the heat the pump gave off, the loop was getting pretty damned warm. I wonder if my flow rates are well above whats needed, if I can actually get better temps by turning the pump down a little.

I might have to experiment with this.

 

[cyberguyz]

Some pumps do run warm and can transfer some of that heat to your coolant. I know there are a lot of D5 "armor" kits out there which supposedly helps them run cooler (I doubt it though). I've never found pump heating a problem though so since the rest of the system can easily handle the heat from the pump as well as from the cpu and viddy. I know I have certainly gotten really good results with lower flow rates and is one of the reason why I always 3/8 - 10mm tubing rather than the 1/2" / 16mm ID stuff. It is way easier to handle and for me works every bit as well.

Running the pump at full tilt won't wear it out faster either since they all pretty much run friction-free.

 

[VanGoghComplex]

Some pumps do run warm and can transfer some of that heat to your coolant. I know there are a lot of D5 "armor" kits out there which supposedly helps them run cooler (I doubt it though). I've never found pump heating a problem though so since the rest of the system can easily handle the heat from the pump as well as from the cpu and viddy. I know I have certainly gotten really good results with lower flow rates and is one of the reason why I always 3/8 - 10mm tubing rather than the 1/2" / 16mm ID stuff. It is way easier to handle and for me works every bit as well.

Running the pump at full tilt won't wear it out faster either since they all pretty much run friction-free.

I've always read that D5s dump their heat into the loop (being essentially liquid-cooled themselves - how's that for meta?) and that DDC pumps radiate heat outside - which is why housings covered in fins are popular for them. It makes sense for the DDC, the ones I've seen taken apart don't have much contact between the stator coils and the spherical wet end - I assume D5s are made a bit differently.

In either case, I'm honestly just nerding a bit on the idea of setting my pump on a PID loop using the flowmeter as the PV and the desired flow rate as the CV. I do this kind of stuff at work for industrial applications, as an instrumentation and controls engineer. =D

 

[Brian_B]

The energy from your pump (mostly measured as head loss, or the difference in pressure from the inlet to the outlet across the closed system) will get transferred into your coolant as heat. That isn't the same thing as the energy that is used in spinning the motor that is connected to the pump- some motors are air cooled, and some are liquid cooled. It pretty much boils down to if your pump is submerged or not - if it's in a tank, it's obviously going to get cooled by the liquid. If it's externally mounted, it could swing either way (but is probably air cooled).

That being said, your talking an extremely small number unless your doing something ludicrous like running a 1hp pump through 1/4" line.

The thing that kills pumps more than anything are starts/stops. Running it faster is harsher on bearings, but again, they run at pretty low speeds even at full speed, so the difference in wear would be pretty small.

Ideally, you want the slowest flow you can find that will give you an acceptable hot side loop water temperature. That minimizes the energy draw of your pump, and the higher temperature across your radiator will increase it's efficiency. More flow than that won't hurt anything, you give yourself some margin against corrosion/kinks/other pressure drops that creep up over time. Your just wasting a few watts on more pump than you need, and you lose a bit out of efficiency out of your radiator, but you gain a bit of efficiency on your blocks (more turbulence = better heat transfer), so it's a net ~meh~. Less flow than minimum will be a problem though, as stuff will start overheating... which is why most everyone errs on the side of more flow rather than less.

You could work out the math yourself if you really wanted to, it takes an amazingly small amount of flow to actually make an efficient, stable loop work well. You ~could~ control flow via a PID, but you probably want to use a core temp or something as your PV... if you have pump speed set as your CV and flow rate as your PV, it's going to be pretty constant once you have the loop installed.

 

[cyberguyz]

I've always read that D5s dump their heat into the loop (being essentially liquid-cooled themselves - how's that for meta?) and that DDC pumps radiate heat outside - which is why housings covered in fins are popular for them. It makes sense for the DDC, the ones I've seen taken apart don't have much contact between the stator coils and the spherical wet end - I assume D5s are made a bit differently.

In either case, I'm honestly just nerding a bit on the idea of setting my pump on a PID loop using the flowmeter as the PV and the desired flow rate as the CV. I do this kind of stuff at work for industrial applications, as an instrumentation and controls engineer. =D

Pretty much all my 12v pump exp is with the D5-style ones. I was using 120v eheim ones prior to that & used a relay card to turn them on/off with the system. Oh the joy when the Swiftech MCP650 pump came into my life!! I do remember it dumping its heat into the coolant. I always ran it at full tilt and the rest of the system managed its heat. My latest D5 acquisition hasn't made it into a loop yet. I do remember seeing the DDC pumps having the heatsink though. Never did work with those - they just seemed so.... teensy =D

 

[VanGoghComplex]

Pretty much all my 12v pump exp is with the D5-style ones. I was using 120v eheim ones prior to that & used a relay card to turn them on/off with the system. Oh the joy when the Swiftech MCP650 pump came into my life!! I do remember it dumping its heat into the coolant. I always ran it at full tilt and the rest of the system managed its heat. My latest D5 acquisition hasn't made it into a loop yet. I do remember seeing the DDC pumps having the heatsink though. Never did work with those - they just seemed so.... teensy =D

They're little, but they're scrappy. Picture a 5'3" marine with a penchant for bar fights. I'm actually using my first D5 currently, was all DDC prior, as I was working inside a case as opposed my my current open-air Core P5 and size was a selling point. Never had any complaints - as Brian B alludes to above, the required flow rate for a functional loop is pretty low (such that even those anemic Asetek pumps found in AIOs manage it) and a DDC is plenty burly enough! =)