How necessary is dual pumps in series?

Dutt1113

[H]ard|Gawd
Joined
Jun 30, 2005
Messages
1,601
I have a bitspower dual d5 top running through 2 large rads, 2 gpu blocks, and cpu block. The dual d5 top is clear and doesn't match my other parts which are black acetal. Its also big, bulky, and kinda ugly. I was considering going to a single ek-xres revo acetal d5 pump to run everything. I'm obviously not going to have the higher head pressure, but does this matter with what I have to run it through?
 
Well, I can't speak to your components, and what their flow is like, but I have a single D5 as part of an XSPC Photon 270 Pump/Res pushing through a 3x140mm Alphacool XT45 rad and a 2x140 Monsta rad as well as an EK fullcover for my Titan and an EK Supremacy EVO CPU block, and it provides more than enough flow through my loop.

You have one more block in there than I do, but I can't imagine that makes a huge difference.

I run my pump at full blast (setting 5) in my loop, but I could probably run it at 2-3 and still get enough flow.

I just run it at 5 because I find the noise level is about the same at all speeds, but the pitch changes, and I find the pitch to be the least disruptive at 5.

My point is, if I could get away with 2-3 in my loop, with one more block in yours, I can't imagine that a single D5 at full speed wouldn't be enough.

With these things it is pretty tough to tell though. Sometimes you just need to build them and test it.
 
I ran a single D5 bitspower res setup through 2x360 rads + 1x480 rad + cpu + dual 980ti blocks and still do the same except it's a single 1080ti block. Without something crazy I can't see why you would have problems.
 
I ran a single D5 bitspower res setup through 2x360 rads + 1x480 rad + cpu + dual 980ti blocks and still do the same except it's a single 1080ti block. Without something crazy I can't see why you would have problems.

Yeah. I was always under the impression that dual pumps was something for weird exotic loops like Linus's whole room loop, not something you need in even a high end multi-GPU/CPU gaming system.

I've been toying with the idea of running a long loop down from my office to the basement with a water chiller down there out of ear shot. For something like this I was considering a dual EK D5 top.
 
In short: not. The D5s and DDCs most common to custom loops already have excessive flow rates, compared to what's needed for effective cooling in a typical custom loop.

You could make the case for redundancy, but the failure rate of those pumps is such a statistical outlier that it's not really relevant for most users.
 
So, even If I had 3 large rads, a cpu, and 2 gpu's, a single d5 pump would be adequate?
 
Actually, there is a couple of things you need to consider. The D5 has a lower head pressure (3.9 meters) with a high flow rate, the DDC has a higher head pressure (5.2 meters) with a low flow rate. If you are pumping water up and down, depending on the size of your case and the configuration of your loop, the D5's flow rate may suffer. Adding a second pump can improve your flow rate significantly without having to run your pumps at full speed, resulting in less noise.

I have a massive Phanteks Enthoo Elite with large radiators, dual GPU blocks and there is a pretty dramatic difference in flow rate between single and dual pump. Flow rate for me has the biggest impact on CPU temps. On the other hand, I have a D5 in my HTPC which lays flat and has a single GPU, chipset and cpu cooled and because the D5 is pumping horizontally only the flow rates are amazing even at low speed.
 
Actually, there is a couple of things you need to consider. The D5 has a lower head pressure (3.9 meters) with a high flow rate, the DDC has a higher head pressure (5.2 meters) with a low flow rate. If you are pumping water up and down, depending on the size of your case and the configuration of your loop, the D5's flow rate may suffer. Adding a second pump can improve your flow rate significantly without having to run your pumps at full speed, resulting in less noise.

I have a massive Phanteks Enthoo Elite with large radiators, dual GPU blocks and there is a pretty dramatic difference in flow rate between single and dual pump. Flow rate for me has the biggest impact on CPU temps. On the other hand, I have a D5 in my HTPC which lays flat and has a single GPU, chipset and cpu cooled and because the D5 is pumping horizontally only the flow rates are amazing even at low speed.

Well, but in such a case you could argue if it wouldnt be better to run dual separated loops.
 
Actually, there is a couple of things you need to consider. The D5 has a lower head pressure (3.9 meters) with a high flow rate, the DDC has a higher head pressure (5.2 meters) with a low flow rate. If you are pumping water up and down, depending on the size of your case and the configuration of your loop, the D5's flow rate may suffer. Adding a second pump can improve your flow rate significantly without having to run your pumps at full speed, resulting in less noise.

I have a massive Phanteks Enthoo Elite with large radiators, dual GPU blocks and there is a pretty dramatic difference in flow rate between single and dual pump. Flow rate for me has the biggest impact on CPU temps. On the other hand, I have a D5 in my HTPC which lays flat and has a single GPU, chipset and cpu cooled and because the D5 is pumping horizontally only the flow rates are amazing even at low speed.
Head only matters until your whole loop is full. Water doesn't compress, and horror vacuui, thus once your loop is full, every inch of rise in your loop is counteracted by an inch of fall.
 
Last edited:
I'm running a basement cooling system with a pair of D5 pumps, travelling 12 feet up to my study, though GPU/CPU/VRM blocks then back down to the basement. I run both pumps at about 2/3rd speed and it's still overkill - one pump will still keep the coolant flowing.
 
So, even If I had 3 large rads, a cpu, and 2 gpu's, a single d5 pump would be adequate?

My D5 has a manual speed setting on the back. 1 thru 5.. I run it on 2.. I dont under stand why anyone would have these pumps turned all the way up. Pushing the water threw the rad at that speed leaves it no time to be cooled.
 
My D5 has a manual speed setting on the back. 1 thru 5.. I run it on 2.. I dont under stand why anyone would have these pumps turned all the way up. Pushing the water threw the rad at that speed leaves it no time to be cooled.

You apparently dont understand how thermodynamics works.


You want the fluid to move as fast as possible.

If a water cooling loop is working optimally there is no measaureable difference in coolant temperature before and after radiator or before and after the water blocks.

When this is the case, the system operates at it's most efficient, as it keeps a steady state between adding heat in the blocks and removing it in the radiators.

If you can measure a increase in temperatures across your water blocks and a decrease in temperature across your radiators you have an inefficient system which will result in higher CPU/GPU temps.


The reason this is true is because heat moves from warm to cold regardless of whether this results in an increase in temperature of the coolant. Heat ALWAYS moves from hot to cold.

The heat transfer is dependent on the temperature delta between your CPU and your coolant, and if you keep fresh coolant moving quickly through your block you maintain a lower average coolant temperature across the block, and thus have faster thermal transfer resulting in lower temperatures.

That being said, D5 pumps are pretty powerful, and in most systems you don't need to run them at full blast to get sufficient flow rates.

I find little to no difference in temps between flow settings on my D5. At 1 it is a little hotter than at 2, but then all the way up to 5 it makes almost no difference.

I still keepke at 5 though, as the noise the pump generates seems to be the same level across the speed range, it does change in pitch though, and I find the pitch to be less bothersome at 5 than at lower speeds.
 
Last edited:
I am NOT an expert, and I'm NOT an engineer/scientist. My son, however, is a senior at a university getting his degree in mechanical engineering, and just finished a couple of thermodynamic/heat transfer classes. We had a discussion on this a few weeks back.

His comments are that the optimal flow rate for cooling also has to do with reaching a flow rate that causes turbulence within the system. Once you hit that, a faster flowrate will only very marginally increase the heat transfer from the blocks to the coolant. Some people only turn their pumps up fast enough to overcome the resistance in the blocks and call it a day, others say they get more cooling if they speed the pumps up a bit past that but that it drops off quickly - my guess is they hit that magical turbulence level and see diminishing returns beyond that.

Again - I'm no expert so don't ask me to explain this further (and I'll leave it open that I misunderstood some fine points). I'm an IT guy, not an engineer.
 
You apparently dont understand how thermodynamics works.


You want the fluid to move as fast as possible.

If a water cooling loop is working optimally there is no measaureable difference in coolant temperature before and after radiator or before and after the water blocks.

When this is the case, the system operates at it's most efficient, as it keeps a steady state between adding heat in the blocks and removing it in the radiators.

If you can measure a increase in temperatures across your water blocks and a decrease in temperature across your radiators you have an inefficient system which will result in higher CPU/GPU temps.


The reason this is true is because heat moves from warm to cold regardless of whether this results in an increase in temperature of the coolant. Heat ALWAYS moves from hot to cold.

The heat transfer is dependent on the temperature delta between your CPU and your coolant, and if you keep fresh coolant moving quickly through your block you maintain a lower average coolant temperature across the block, and thus have faster thermal transfer resulting in lower temperatures.

That being said, D5 pumps are pretty powerful, and in most systems you don't need to run them at full blast to get sufficient flow rates.

I find little to no difference in temps between flow settings on my D5. At 1 it is a little hotter than at 2, but then all the way up to 5 it makes almost no difference.

I still keepke at 5 though, as the noise the pump generates seems to be the same level across the speed range, it does change in pitch though, and I find the pitch to be less bothersome at 5 than at lower speeds.


I do understand that mostly, its common sense really. The exchange of heat is not instantaneous, there for you need to have the coolant in the radiator long enough for the heat to be removed. The Radiators can only remove X amount of heat, and the block will only absorb X. Once that balance point is met, it doesnt matter really how fast you moving the water, as long as its moving. 2-3 degrees C is not that important, when the temps are between 45-55c on a summer day at full load.
 
I am NOT an expert, and I'm NOT an engineer/scientist. My son, however, is a senior at a university getting his degree in mechanical engineering, and just finished a couple of thermodynamic/heat transfer classes. We had a discussion on this a few weeks back.

His comments are that the optimal flow rate for cooling also has to do with reaching a flow rate that causes turbulence within the system. Once you hit that, a faster flowrate will only very marginally increase the heat transfer from the blocks to the coolant. Some people only turn their pumps up fast enough to overcome the resistance in the blocks and call it a day, others say they get more cooling if they speed the pumps up a bit past that but that it drops off quickly - my guess is they hit that magical turbulence level and see diminishing returns beyond that.

Again - I'm no expert so don't ask me to explain this further (and I'll leave it open that I misunderstood some fine points). I'm an IT guy, not an engineer.

The ideal flow rate is around 1-1.5gpm and that is really easy to achieve. This was born from testing data years back, so yea you guys are on to something. Anymore flow is wasted effort and its really about head pressure especially when we start adding in more and more blocks.


Well, but in such a case you could argue if it wouldnt be better to run dual separated loops.

Dual loops is a lot of effort for little gain. It only makes sense in specific cases when you purposely want to separate for example the cpu heat from the gpu heat.


You want the fluid to move as fast as possible.

Actually you don't. Higher flow than 1-1.5gpm will not improve the cooling much at all. Above 1.5gpm and we are talking fractions.


Some more info if anyone is interested...

http://martinsliquidlab.petrastech.com/MartinsFlowRateEstimator.html
 
I do understand that mostly, its common sense really. The exchange of heat is not instantaneous, there for you need to have the coolant in the radiator long enough for the heat to be removed. The Radiators can only remove X amount of heat, and the block will only absorb X. Once that balance point is met, it doesnt matter really how fast you moving the water, as long as its moving. 2-3 degrees C is not that important, when the temps are between 45-55c on a summer day at full load.


It becomes a differential equations problem with rates of change on each side, one across the radiator, removing heat, and one across the block adding heat.

It's a complicated calculation, but qualitatively it isn't that tough.

Yes, a shorter time in the radiator results in less heat removed from that particular volume of water passing through the radiator. The same however happens on the other end. The shorter time the water spends in the block, the less amount of heat that particular slug of water gains.

The important part here is that when we have lower flow, it results in greater heat transfer per unit volume of water in each end.

We don't care about heat transfer per unit volume of water. We care about heat transfer per unit time, and this measure is improved when the flow is higher to the point where you don't have a measurable heat gradient across your block or radiator.
 
Actually you don't. Higher flow than 1-1.5gpm will not improve the cooling much at all. Above 1.5gpm and we are talking fractions.

I feel like this is exactly what I have been saying, but qualitatively rather than quantitatively.

If you have a measaureable difference before or after your block/GPU your flow is on the low side, if you don't, your flow is good.

Exactly where that falls from a flow rate perspective I don't know for sure, and it is less relevant as it's going to differ wildly from block design to block design and temperature levels anyway.

You are suggesting that I am arguing higher is always better. I am not.

Note that I get adequate flow rated with my pump at the 2/5 setting.

I still run it at 5/5 as it results in a better noise profile in my application.

I would - however - err on the high side rather than the low side if I had to choose. There is no need to perfectly tube flow. Just turn it up until you see no further improvement in temps and then leave it there.

Other than a little extra power use, there is no down side to running higher flow than needed. Too low flow results in reduced performance though, so always err high.
 
Last edited:
It becomes a differential equations problem with rates of change on each side, one across the radiator, removing heat, and one across the block adding heat.

It's a complicated calculation, but qualitatively it isn't that tough.

Yes, a shorter time in the radiator results in less heat removed from that particular volume of water passing through the radiator. The same however happens on the other end. The shorter time the water spends in the block, the less amount of heat that particular slug of water gains.

The important part here is that when we have lower flow, it results in greater heat transfer per unit volume of water in each end.

We don't care about heat transfer per unit volume of water. We care about heat transfer per unit time, and this measure is improved when the flow is higher to the point where you don't have a measurable heat gradient across your block or radiator.


I guess that comes down to personal preference, but were both right.
 
I feel like this is exactly what I have been saying, but qualitatively rather than quantitatively.

If you have a measaureable difference before or after your block/GPU your flow is on the low side, if you don't, your flow is good.

Exactly where that falls from a flow rate perspective I don't know for sure, and it is less relevant as it's going to differ wildly from block design to block design and temperature levels anyway.

Note that I get adequate flow rated with my pump at the 2/5 setting.

I still run it at 5/5 as it results in a better noise progile in my my application.

You need whatever you need to hit 1gpm to 1.5gpm and that is all. Anything more is really wasted effort, thus you really don't need the fluid to move as fast as possible.
 
You need whatever you need to hit 1gpm to 1.5gpm and that is all. Anything more is really wasted effort, thus you really don't need the fluid to move as fast as possible.

I never said you need the fluid to move as fast as possible.

Please read my clarified edit above. (I tried to improve it before you read it but apparently failed)
 
D5 Pump Specifications..
 

Attachments

  • D5.png
    D5.png
    245.9 KB · Views: 0
It's tough to predict what your head pressure will be in any given loop as it depends on the restrictiveness of all blocks and radiators, fittings, tube diameter, bends, etc. Etc.

I find mine flows adequately at 2-3 of 5 though, so I assume it would be difficult to get enough head pressure to make a D5 inadequate.
 
Agreed...

Also, I am sure there is an absolute perfect condition setup for commercial purposes. But for our small system loops, it really just come down to what we want..
 
Well, but in such a case you could argue if it wouldnt be better to run dual separated loops.

Dual loops is a lot of effort for little gain. It only makes sense in specific cases when you purposely want to separate for example the cpu heat from the gpu heat.

Yeah, I'd argue that in 99.99% of cases, even in large loops with many blocks it does not make sense to run separate loops, if for no other reason than the law of averages.

Lets compare two simple loops with one block and one radiator each, to one larger loop with two blocks and two radiators.

The larger loop will have equal or lower temps 100% of the time.

The reason for this? The components cooled by both the blocks are rarely going to be at full load at the same time. When they are, performance will be roughly equivalent.

Most of the time, however, one of the components will be at full load while the other is at partial load or idle.

Example:
Typical game: ~100% GPU load, ~30% CPU load.

Encoding or rendering: 100% CPU load, ~0% GPU load.

In the gaming example your GPU will run cooler in the big loop than it would in separate loops, because it is not just using one radiator, but gaining from the capacity of both.

Vice versa for the rendering encoding loop.

General rule of thumb. Unless you know better (and even then, you probably think you do, but don't) it is better to run everything in one large loop.

Dual loops mostly became popular for show builds because they wanted to achieve cool looks with different color fluids, not for performance.
 
You need whatever you need to hit 1gpm to 1.5gpm and that is all. Anything more is really wasted effort, thus you really don't need the fluid to move as fast as possible.

I never said you need the fluid to move as fast as possible.

Please read my clarified edit above. (I tried to improve it before you read it but apparently failed)


Actually I take that back. In reviewing what I wrote, I did in fact type that (whoops) It's not what I INTENDED to type :p

I was typing a quick message from my phone during a training break and messed up. What I meant was, in general higher flow is better than lower flow. There is a threshold beyond which more flow won't really help, but it won't do much harm either, so the easiest way to maximize performance is just to pin the flow.
 
Last edited:
I am NOT an expert, and I'm NOT an engineer/scientist. My son, however, is a senior at a university getting his degree in mechanical engineering, and just finished a couple of thermodynamic/heat transfer classes. We had a discussion on this a few weeks back.

His comments are that the optimal flow rate for cooling also has to do with reaching a flow rate that causes turbulence within the system. Once you hit that, a faster flowrate will only very marginally increase the heat transfer from the blocks to the coolant. Some people only turn their pumps up fast enough to overcome the resistance in the blocks and call it a day, others say they get more cooling if they speed the pumps up a bit past that but that it drops off quickly - my guess is they hit that magical turbulence level and see diminishing returns beyond that.

Again - I'm no expert so don't ask me to explain this further (and I'll leave it open that I misunderstood some fine points). I'm an IT guy, not an engineer.


Yeah, I AM an engineer, and I have taken thermodynamics classes, but they were never my specialty, and I took them years ago, so I am rusty as all hell.

I am familiar with the cavitation problem that can result from turbulence in fluids. What happens is small pockets or bubbles form in the liquid due to too rapid movement, and then you wind up with less contact between the fluid and your heat transfer surface, theoretically reducing thermal transfer.

I'm used to reading about effects of these in boat propellers and things like that though, not in enclosed relatively low flow coolant loops. I would have thought we - even at full blast - would be far away from that level in a PC water cooling loop, especially after you have removed all air from the system.

Some simple testing should settle that though.

the non PWM D5 pumps only have 5 settings, so it should take too long.

Load up Heaven benchmark and prime95 simultaneously, with pump at each of the settings, keeping room temp constant, and wait for equilibrium. Measure system temp. Check temperature at each pump setting.

From my recollection from when I first set my system up, with the pump at speed settings 3-5 there was no difference at all. At 2 there was a difference but it was marginal, and may just have been measurement/testing error due to room temp shifts. At speed setting 1 there was a measurable temperature difference, but I can't recall what it was.
 
Last edited:
Yeah, I AM an engineer, and I have taken thermodynamics classes, but they were never my specialty, and I took them years ago, so I am rusty as all hell.

I am familiar with the cavitation problem that can result from turbulence in fluids. What happens is small pockets or bubbles form in the liquid due to too rapid movement, and then you wind up with less contact between the fluid and your heat transfer surface, theoretically reducing thermal transfer.

I'm used to reading about effects of these in boar propellers and things like that though, not in enclosed relatively low flow coolant loops. I would have thought we - even at full blast - would be far away from that level in a PC water cooling loop, especially after you have removed all air from the system.

Turbulence in propulsion isn't desirable, you want smooth flow when you're riding in a plane or a boat; but you do want turbulence in heat transfer (although not to the point of cavitation); in a turbulent flow, water that contacts the block will be rapidly mixing with the rest of the water, so you'll get heat transfer from convection in addition to conduction.
 
Head only matters until your whole loop is full. Water doesn't compress, and horror vacuui, thus once your loop is full, every inch of rise in your loop is counteracted by an inch of fall.

Static head - yes your absolutely correct. What goes up must come
down in a closed loop and that cancels out.

But head still matters (hah). You have head loss for pipe/tubing arrangements and for each piece of equipment you add to the loop.

For most WC setups and pumps like the D5 and DDC that isn’t hugely significant. But if you have a lot of 90’ bends or are trying to run piping across your house or drive 8 rads in a row, it will start to add up.
 
Yeah, I AM an engineer, and I have taken thermodynamics classes, but they were never my specialty, and I took them years ago, so I am rusty as all hell.

I am familiar with the cavitation problem that can result from turbulence in fluids. What happens is small pockets or bubbles form in the liquid due to too rapid movement, and then you wind up with less contact between the fluid and your heat transfer surface, theoretically reducing thermal transfer.


From my industrial maintenance background and fluid power classes in college, cavitation is generally only found in the pump or suction tube. The reason for this is due to pump suction exceeding the possible flow rate from the tank. Cavitation happens on the suction side of a loop, not the pressure side.
In addition, calling them bubbles isn't quite accurate without an explanation to the ignorant. "Bubbles", in common terms, notes a gas pocket in a fluid. The bubbles we get during cavitation are not gas, rather an actual void, or vacuum. You can't have a vacuum on the pressure side of a series hydraulic loop that has been properly primed and bled.

Different pump designs are also more or less susceptible to excess suction. The impeller style used in the pumps of PC loops are very resistant to cavitation due to the fact that the impeller is frictionless. Vane pumps are the worst offenders.
 
I will only add two things that i haven't seen mentioned (with my apologies in advance if i just missed them, i semi-skimmed through it):

- Quick disconnects; you want to use these and in some respectable quantity, yes, you will probably need a second pump; and even then as you may have noticed, it's 'probably'; it's something to keep in mind however. Perhaps a flow measuring unit could assist with that, if/when that time arrives.
- redundancy; one fails you, no probs, it's like nothing ever happened, you don't even need to do anything. Sounds overkill and i'd agree, until it actually happens to you in the worst possible moment, like while you're working, or that one darned weekend you've actually managed to get some time off.. and you need 2,3,6, whatever hours total to have everything back in place, plus all the fuss doing it, and that's assuming you just happened to have a second pump handy! How common! Things don't always fail early Saturday mornings :)

A note here that the above is in regard to a serial setup. Forget parallel.
 
From my industrial maintenance background and fluid power classes in college, cavitation is generally only found in the pump or suction tube. The reason for this is due to pump suction exceeding the possible flow rate from the tank. Cavitation happens on the suction side of a loop, not the pressure side.
In addition, calling them bubbles isn't quite accurate without an explanation to the ignorant. "Bubbles", in common terms, notes a gas pocket in a fluid. The bubbles we get during cavitation are not gas, rather an actual void, or vacuum. You can't have a vacuum on the pressure side of a series hydraulic loop that has been properly primed and bled.

Different pump designs are also more or less susceptible to excess suction. The impeller style used in the pumps of PC loops are very resistant to cavitation due to the fact that the impeller is frictionless. Vane pumps are the worst offenders.
I don't think cavitation is an issue even with the most powerful PC watercooling pumps. I'm an instrumentation engineer, so my experience with pumps is kinda sideline, but the pumps I deal with at work can cavitate, and it destroys the steel impellers of those pumps if left unchecked. A collapsing cavitation bubble is like a hammer blow. It would trash the plastic impellers we deal with in a heartbeat.
 
I don't think cavitation is an issue even with the most powerful PC watercooling pumps. I'm an instrumentation engineer, so my experience with pumps is kinda sideline, but the pumps I deal with at work can cavitate, and it destroys the steel impellers of those pumps if left unchecked. A collapsing cavitation bubble is like a hammer blow. It would trash the plastic impellers we deal with in a heartbeat.


Exactly. This is what I was saying.
Impeller pumps are very resistant to cavitation. It's high speed vane pumps that have the biggest issue.
When I was still in automotive diesel, the Ford 6.0l was terrible for it. The water pump(impeller) would chew through the front covers that were cast aluminum. Chew complete holes through it! That however had more to do with restricted tank-to-pump rather than pump rpm
 
Back
Top