Doubling D5 for flow?

Joined
Apr 5, 2016
Messages
2,286
I'm pretty happy with my system, but it's been too long since I messed with it and I'm wondering if I'm leaving any performance on the table.

My system is such:

D5 pump ->
Flow meter ->
GPU block ->
CPU block ->
480mm rad ->
420mm rad ->
Reservoir ->
(D5 pump)

Before anyone says anything about rerouting my tubing: no.

At 100%, my system manages just a hair over that 1gpm mark. I'd like to add another D5 but I'm not versed in dual pumps.

To increase flow rate, would I want to install the pumps in series or parallel? Is another D5 my best choice or should I consider dual DDCs?
 
I have no actual experience here, only research from my own scheming. If you are looking at a drop in replacement with the same routing, it is probably easiest to buy one of the dual pump tops on the market, use your existing pump in one of the slots and buy another for the other slot.

EK has one which looks really nice, but it is a bit expensive at $129 just for the top.

XSPC as a dual D5 plexi top which is quite a bit cheaper. There are probably others too.
 
Series for increased pressure at the same flow, parallel for increased flow at the same pressure. Either will help you increase your actual flow, but there's probably not a lot of extra performance to be had if you're already around 1 GPM. I would honestly go with dual D5s over dual DDCs, both for the reduced noise and the increased reliability, but I don't think you need to do either unless you're just itching to play with something.
 
You will want the pumps in series. If a pump fails, the other one will pump back through the first one if you have it in parallel. Also, the flow gains in practical usage are much higher in series than in parallel due to the way pressure/flow curves work.

I am generally against EK stuff, and XSPC has historically made some good dual D5 tops.
 
You will want the pumps in series. If a pump fails, the other one will pump back through the first one if you have it in parallel.

You have that backwards. In series one pump follows the other in the loop, and in parallel they both occupy the same point in the loop in terms of flow.
 
You have that backwards. In series one pump follows the other in the loop, and in parallel they both occupy the same point in the loop in terms of flow.

If they are in parallel, and only one pump is running, you will push water back through the other pump. The system will either perform poorly or not at all.

You definitely want them in series, one after the other.

You should probably double check just to make sure, but every dual top I've seen is routed in series.
 
Can I safely assume that any dual top is going to be running the pumps in series? Or are they also made in parallel? I've yet to see a product description specify.
 
Can I safely assume that any dual top is going to be running the pumps in series? Or are they also made in parallel? I've yet to see a product description specify.

I feel like this should be a given, but I could be wrong, which is why I would verify.
 
the EK D5 XTOP Revo Dual D5 is confirmed to be serial. I'd have to google more to figure out the others.
 
I am generally against EK stuff, and XSPC has historically made some good dual D5 tops.

I know there are still hard feelings from the nickel plating issue, but that was like 2011, almost a decade ago. They have made some very good products since then.

The exception would be the first version of their Threadripper block which was a fail, but even decent companies fail once in a while. I think they presumed Threadripper would remain a very niche product and that it wasn't worth re-designing a bigger block for the platform, and then were surprised when ti became more popular and other block designers designed larger blocks for it. They replaced that TR block with a v2 that performs much better since.

In general EK stuff is great. My biggest issue is that it tends to be a little expensive. The $129 dual pump top illustrates what I am talking about.

XSPC has manufactured the only components I have ever had leak (swivel bends) and the first gen of the Photon D5 reservoir I have is a pain in the ass to fill because of a stupid design flaw, and the 5 step setting pump that came with the resrvoir was noisy and annoying from day one. (only reason I used it as long as I did was because I had no prior D5 experience, and just assumed that was what D5 pumps were like)

So, I guess my take is, all companies have issues, and I've been very happy with the EK parts I own (Fullcover block for my Pascal Titan X, and D5 G2 PWM pump) and am not opposed to using thier parts in th efuture, even though I opted for Watercool's Heatkiller IV Pro for my Threadripper CPU block this time around.
 
I ran mine in series, 2 separate pumps, but it's 4 radiators and I just wanted to try 2 pumps. It also let me run a bit slower speed, but pump speed had a pretty big effect on temps - like a 10 degree instant drop on loaded CPU temp by going from a low speed to wide open. I settled somewhere less than wide open, just fiddled with it until I got speed/temp I was happy with.

RVIEO-1.jpg
 
I ran mine in series, 2 separate pumps, but it's 4 radiators and I just wanted to try 2 pumps. It also let me run a bit slower speed, but pump speed had a pretty big effect on temps - like a 10 degree instant drop on loaded CPU temp by going from a low speed to wide open. I settled somewhere less than wide open, just fiddled with it until I got speed/temp I was happy with.

View attachment 220799


Wow, that's a surprisingly large difference.

Did you measure flow rates at all at each setting?

Personally I use a target temp and change pump speed based based on temp these days.

Has the added bonus of working the air out of the system due to flow rate changing.

I usually set the target temp for the pump a degree C below the target temp for the fans to make sure it is up to speed before the heavy lifting is needed.

You just need to be able to set a.minimum.pump speed or the thing will turn off and on constantly at idle which is annoying and probably causes premature wear.

This way it maxes out at heavy load, and sits at low speed at idle and doesn't unecessarily contribute to loop temp.
 
Wow, that's a surprisingly large difference.

Did you measure flow rates at all at each setting?

Personally I use a target temp and change pump speed based based on temp these days.

Has the added bonus of working the air out of the system due to flow rate changing.

I usually set the target temp for the pump a degree C below the target temp for the fans to make sure it is up to speed before the heavy lifting is needed.

You just need to be able to set a.minimum.pump speed or the thing will turn off and on constantly at idle which is annoying and probably causes premature wear.

This way it maxes out at heavy load, and sits at low speed at idle and doesn't unecessarily contribute to loop temp.

I don't have a flow meter. All my fans and pumps are set on temp curves in the BIOS, during normal operation everything just idles along, but if the CPU reaches 75C then the whole thing ramps up a lot. About the only thing I use that'll make it do that is benchmarks and I haven't ran any of those since dialing it in about a year ago. Day to day, it's pretty quiet for a rig with 2 pumps and 4 rads (2 are in the basement of the case, under the midplate), P/P fans all around.
 
I know there are still hard feelings from the nickel plating issue, but that was like 2011, almost a decade ago. They have made some very good products since then.

The exception would be the first version of their Threadripper block which was a fail, but even decent companies fail once in a while. I think they presumed Threadripper would remain a very niche product and that it wasn't worth re-designing a bigger block for the platform, and then were surprised when ti became more popular and other block designers designed larger blocks for it. They replaced that TR block with a v2 that performs much better since.

In general EK stuff is great. My biggest issue is that it tends to be a little expensive. The $129 dual pump top illustrates what I am talking about.

XSPC has manufactured the only components I have ever had leak (swivel bends) and the first gen of the Photon D5 reservoir I have is a pain in the ass to fill because of a stupid design flaw, and the 5 step setting pump that came with the resrvoir was noisy and annoying from day one. (only reason I used it as long as I did was because I had no prior D5 experience, and just assumed that was what D5 pumps were like)

So, I guess my take is, all companies have issues, and I've been very happy with the EK parts I own (Fullcover block for my Pascal Titan X, and D5 G2 PWM pump) and am not opposed to using thier parts in th efuture, even though I opted for Watercool's Heatkiller IV Pro for my Threadripper CPU block this time around.

Fair points. My issue with EK comes from the arrogance they appear to have (never admitted fault for their mistakes), which makes them like the nVidia of the watercooling world. And unlike nVidia, EK has competitors that come close, if not surpassing, their performance.
 
Fair points. My issue with EK comes from the arrogance they appear to have (never admitted fault for their mistakes), which makes them like the nVidia of the watercooling world. And unlike nVidia, EK has competitors that come close, if not surpassing, their performance.

I see that.

I wasn't using EK products back in the day, but I wish they had taken more responsibility for the failed nickel parts.

To be fair their memo on the subject did make sense. People were using a lot of stuff in their loops that could have been problematic.

That's why I only use glycol based coolants and never mix metals in my loop, including silver coils. Glycols are perfectly safe for all metals and are both corrosion and growth inhibitor a as well as surfactants all in one.

They were a small company at the time. My best guess is that of they had compensated everyone for the nickel disaster it would have driven them out of business.

They definitely could have handled it better though.
 
Just get bigger rad or reservoir, or faster pump...

2nd pump won't do anything, and won't speed things up - may even cause problems, or damage them.
 
To increase flow rate, would I want to install the pumps in series or parallel? Is another D5 my best choice or should I consider dual DDCs?

Technically its not flow that we need in our loops, its head pressure. Flow greater than 1.5gpm is wasted energy and D5 are somewhat head pressure limited so doubling your head pressure will definitely give gains if you are running multiple blocks and rads.

2nd pump won't do anything, and won't speed things up - may even cause problems, or damage them.

Dude, seriously that's misinformation.
 
Just get bigger rad or reservoir, or faster pump...

2nd pump won't do anything, and won't speed things up - may even cause problems, or damage them.
"Get a bigger reservoir" :LOL:

Technically its not flow that we need in our loops, its head pressure. Flow greater than 1.5gpm is wasted energy and D5 are somewhat head pressure limited so doubling your head pressure will definitely give gains if you are running multiple blocks and rads.
All the same though, right? Increasing head pressure for a set amount of loop restriction will increase flow. Is the gain linear? If, for example, I can maintain 1.09gpm with my current pump at 100%, should I expect close to 2.18gpm with a second identical pump in series?

Being a bit of a nitpick while I figure out how I'm even going to shoehorn in a second pump... :confused:
 
2nd pump won't do anything, and won't speed things up - may even cause problems, or damage them.
"Get a bigger reservoir" :LOL:


All the same though, right? Increasing head pressure for a set amount of loop restriction will increase flow. Is the gain linear? If, for example, I can maintain 1.09gpm with my current pump at 100%, should I expect close to 2.18gpm with a second identical pump in series?

Being a bit of a nitpick while I figure out how I'm even going to shoehorn in a second pump... :confused:

No, flow is not the same as head pressure. Head pressure is needed to overcome pressure drops due to resistance. These are low speed flows, ie. incompressiple flows (think Mach number). Adding pumps together in series will double their head pressure at the same flow rate.

http://www.xtremesystems.org/forums...e-Comparison-(Petra-s-Radiical-Alphacool-etc-
 
No, flow is not the same as head pressure. Head pressure is needed to overcome pressure drops due to resistance. These are low speed flows, ie. incompressiple flows (think Mach number). Adding pumps together in series will double their head pressure at the same flow rate.

http://www.xtremesystems.org/forums...e-Comparison-(Petra-s-Radiical-Alphacool-etc-


It is not the same, no, but a higher head-pressure will help you overcome the restrictions of the loop, and should result in a higher flow rate, up until a point.

"Get a bigger reservoir" :LOL:


All the same though, right? Increasing head pressure for a set amount of loop restriction will increase flow. Is the gain linear? If, for example, I can maintain 1.09gpm with my current pump at 100%, should I expect close to 2.18gpm with a second identical pump in series?

Being a bit of a nitpick while I figure out how I'm even going to shoehorn in a second pump... :confused:

I would expect an increase, but I wouldn't expect it to scale linearly.

The way I understand how it works is this:

Putting two pumps in series will NOT increase flow in a theoretical open system with no resistance. It would be wasted in that scenario. You can add as many pumps of the same deaign as you want, but the max flow without any resistance is not going to change.

The added head pressure does - however - help to defeat any resistance in your loop due to bends, blocks, radiators, etc.

So, assuming you are using your first pump in a real world loop that has back-pressure from various restrictions, adding a second pump should increase your flow, but no matter how many pumps you add you will never exceed the flow of that same single pump operating unrestricted.

In fact, it winds up being asymptotic, sortof like the curelve below:

RN33N.png


So as you add more and more head pressure, you will get closer and closer to the theoretical max flow rate of a single unrestricted pump.

As for how much you will gain by adding a pump, that depends on way too many factors for me to predict. How much is your current head pressure? How much back-pressure does your loop add, how much head pressure are you adding, etc. etc.

All I can say is, it will increase, and the new flow rate will be somewhere in between your current flow rate, and the flow rate your pump produces if completely unobstructed and operating on a plane surface.
 
You didn't read the link which showed, proved that two pumps in series DOUBLES head pressure at the same flow rate.

Theoretically, two pumps in parallel would instead double flow rate.

However, the crux is flow rate doesn't matter since regardless of pump, yall only need 1.5gpm max flow rate to achieve ideal heat transfer. What yall really need is head pressure.
 
You didn't read the link which showed, proved that two pumps in series DOUBLES head pressure at the same flow rate.

Theoretically, two pumps in parallel would instead double flow rate.

However, the crux is flow rate doesn't matter since regardless of pump, yall only need 1.5gpm max flow rate to achieve ideal heat transfer. What yall really need is head pressure.

Well,

The opposite is also true. If you have sufficient flow rate you don't need any more head pressure.

The sole purpose of the head pessure is to sustain the flow rate through the loop which has some back-pressure.

I accepted your statement of doubling of the head pressure as fact. The interesting part is what impact that head pressure will have on the flow rate.

In the case of OP he is already at 1.09. Doubling the head pressure will allow him to better counter the resistance in the loop and increase his flow rate. As for how much the flow rate will increase, that is tough to know. Hopefully enough to hit the 1.5GPM level where further increases are irrelevant, but even if it does, that will likely only have a marginal impact on actual temps.

Sure, we stop seeing any measurable improvement in temps above 1.5 gpm, but the 1-1.5 gpm range is already well into the realmmof diminishing returns.
 
Technically its not flow that we need in our loops, its head pressure. Flow greater than 1.5gpm is wasted energy and D5 are somewhat head pressure limited so doubling your head pressure will definitely give gains if you are running multiple blocks and rads.
Dude, seriously that's misinformation.

If they are placed in series you increase pressure - not flow rate. The volumetric flow rate must be constant - if its not you will starve the latter pump - and could wear out the bearing.
Ideally both pumps need to run at same speed, and have same flow rate.

He can only benefit from 2 pumps with variable speeds in parallel config. i.e. 2 pumps separately connected to rez, then pumps connected to the loop with a single T junction connection to the rest of the system.
 
Last edited:
If they are placed in series you increase pressure - not flow rate. The volumetric flow rate must be constant - if its not you will starve the latter pump - and could wear out the bearing.
Ideally both pumps need to run at same speed, and have same flow rate.

He can only benefit from 2 pumps with variable speeds in parallel config. i.e. 2 pumps separately connected to rez, then pumps connected to the loop with a single T junction connection to the rest of the system.

Yes, pumps in series increase pressure.

But increasing pressure results in increased flow if and only if your flow rate is limited by the restrictiveness of your loop. That pressure helps counteract the restrictiveness of the loop, and allows for higher flow.

It can never increase flow above the unrestricted max flow rate of the pump.

Example:

The theoretical max flow of a single D5 pump is about 2.2 gpm if you keep it level to your reservoir and have no restriction at all on out outlet, so it can just blast freely.

The OP's D5 is giving him 1.09 gpm in his loop, due to the various components and bends restricting his flow.

By adding a second (or more) pump in series he doubles the head pressure which in turn can help increase the flow above that 1.09gpm, (but how much isn't entirely clear to me. If I had to make an educated guess I'd say about 1.6 gpm)

No matter how many pumps he puts in series, the flow rate will never exceed the 2.2 gpm of the unrestricted D5.
 
Yes, pumps in series increase pressure.

But increasing pressure results in increased flow if and only if your flow rate is limited by the restrictiveness of your loop. That pressure helps counteract the restrictiveness of the loop, and allows for higher flow.

It can never increase flow above the unrestricted max flow rate of the pump.

Example:

The theoretical max flow of a single D5 pump is about 2.2 gpm if you keep it level to your reservoir and have no restriction at all on out outlet, so it can just blast freely.

The OP's D5 is giving him 1.09 gpm in his loop, due to the various components and bends restricting his flow.

By adding a second (or more) pump in series he doubles the head pressure which in turn can help increase the flow above that 1.09gpm, (but how much isn't entirely clear to me. If I had to make an educated guess I'd say about 1.6 gpm)

No matter how many pumps he puts in series, the flow rate will never exceed the 2.2 gpm of the unrestricted D5.

1580953944751.png


image from
https://www.sciencedirect.com/book/9780123876935/pipeline-rules-of-thumb-handbook
 
Maybe I'm remembering my fluid dynamics wrong (it's been almost 20 years after all) but something about that doesn't seem right to me.


Ok. Here we go.


You are correct that adding a second pump in series only adds head pressure, but you also have to consider the impact of the system curve.

pump_serie.png


The system curve shows that while adding another pump in series in isolation, you would expect to move from point 1 to point 2, with increased head, but the same flow, but when factoring in the system curve with both pumps running you actually move to point 3, where the added head actually helps you pump more flow through the restrictive loop.

I thought I was losing my mind there for a moment, but it is coming back to me now.

Link for reference.
 
Last edited:
Here is a much more relavant picture:

View attachment 221250

As you can see, in the real world, there is no benefit to running two pumps in parallel unless they are running two independent loops (shared reservoir is okay).

Would be more interesting if they included a line for a single D5 for comparison.

Looks like the EK Xtop Dual is killing it in that Chart.

What I don't understand is how they are registering such high flow rates on the right hand side of the chart. Pumps in series should max out their flow at the same flow rate as a single pump, which according to the specs I found when googling is 500L/h, which converts to ~2.2 Gpm. Those extreme above 2.2GPM flow rates thus should not be possible.

I guess the specs could be wrong. Maybe they underrate them in order to ahve a safety margin, and to deal with production variability.

That said, it is irrelevant anyway as their system curves for even the low restriction loop doesn't have flow going that crazy high.
 
Would be more interesting if they included a line for a single D5 for comparison.

Looks like the EK Xtop Dual is killing it in that Chart.

What I don't understand is how they are registering such high flow rates on the right hand side of the chart. Pumps in series should max out their flow at the same flow rate as a single pump, which according to the specs I found when googling is 500L/h, which converts to ~2.2 Gpm. Those extreme above 2.2GPM flow rates thus should not be possible.

I guess the specs could be wrong. Maybe they underrate them in order to ahve a safety margin, and to deal with production variability.

That said, it is irrelevant anyway as their system curves for even the low restriction loop doesn't have flow going that crazy high.

The article where that graph came from.

They're using D5 Strong pumps. Those run at 24 volts and are capable of 5.9 gpm. Interestingly, they're specced at 4.6 gpm at 12 volts.
 
Ok. Here we go.


You are correct that adding a second out p only adds head pressure, but you also have to consider the impact of the system curve.

View attachment 221253

The system curve shows that while adding another pump in series in isolation, you would expect to move from point 1 to point 2, with increased head, but the same flow, but when factoring in the system curve with both pumps running you actually move to point 3, where the added head actually helps you pump more flow through the restrictive loop.

I thought I was losing my.mind there for a moment, but it is coming back to me now.

yeah you are right.

Though in terms of actual temps it makes little sense anyware, (we aren't hitting boiling points within the loop), not sure if that little bit higher flow will increase the heat rejection from the rad substantially to matter either.
(while noting setup will be harder, will take more space, its very unlikely pumps will match 1:1 the speed always through mobo - will cause issue at some point, unless he uses some external power to provide static dc | and gets 2 new pumps - ?older/more used up bearings might bring the pump perf down?)

btw in that link they run it in parallel (or it looks like they do)
1580957754763.png
 
yeah you are right.

Though in terms of actual temps it makes little sense anyware, (we aren't hitting boiling points within the loop), not sure if that little bit higher flow will increase the heat rejection from the rad substantially to matter either.
(while noting setup will be harder, will take more space, its very unlikely pumps will match 1:1 the speed always through mobo - will cause issue at some point, unless he uses some external power to provide static dc | and gets 2 new pumps - ?older/more used up bearings might bring the pump perf down?)

btw in that link they run it in parallel (or it looks like they do)
View attachment 221267

I mostly agree with that.

Temps do improve all the way up to about 1.5 gpm (~340LPH).

Most of that improvement happens up until 1 gpm though. There is some significant limiting returns above that.
 
yeah you are right.

Though in terms of actual temps it makes little sense anyware, (we aren't hitting boiling points within the loop), not sure if that little bit higher flow will increase the heat rejection from the rad substantially to matter either.
(while noting setup will be harder, will take more space, its very unlikely pumps will match 1:1 the speed always through mobo - will cause issue at some point, unless he uses some external power to provide static dc | and gets 2 new pumps - ?older/more used up bearings might bring the pump perf down?)

btw in that link they run it in parallel (or it looks like they do)
View attachment 221267

That setup is specific to the Alphacool D5 Repack, which is incapable of running two D5s in series. It is a parallel only reservoir. I use one because it's aesthetically pleasing to me, and figure that the shared reservoir would offset some of the disadvantages of running dual loops.

In terms of actual benefits, it's not that hard to figure out. It's all about maintaining an equilibrium temperature in your loop, and the closer you have the equilibrium (difference between inlet/outlet temps), the more efficient your loop will be.

Let's start with 1 GPM with a CPU putting out 200 watts. 200 watts is 200 joules per second of energy that needs to be absorbed by the water (it's actually less as some goes into the motherboard, but we'll say all heat transfer occurs within the waterblock for this example). The specific heat of water is 4.186 joules per gram per * C, which means it takes 4.186 joules of energy to raise the temperature of 1 gram of water 1 degree C. 1 gpm equates to 3785 grams of water per minute, or 63 grams of water traveling through the block per second. 200 watts therefore increases the temperature of the water by approximately 0.76 degrees as it travels through the block (200 / 4.186 / 63 = 0.76). At half the flow rate, you're looking at 1.5 degrees, which results in higher CPU temps.

The biggest issue comes with the radiator, which is far more complicated. The main thing is that heat transfer decreases as the delta T decreases. You get the most heat transfer to air right at the inlet of the radiator, and it drops off as the coolant flows through the radiator. A radiator in equilibrium is going to transfer the same amount of heat the CPU produces, so your delta Ts at the inlet and outlet remain the same. The difference is that a high flow rate system will efficiently use the outlet side of the radiator, while a low flow rate system won't. The high flow rate system will therefore maintain a lower average temperature, while the low flow rate system has a higher average coolant temperature, once again leading to higher system temps.
 
Back
Top