XD3 Orientation

[Smokester]

Hello

2 Questions:

1) Can the XD3 be mounted with a 90 Degree offset, such that the OUTLET ports are facing DOWN.

I understand all the documentation states to be mounted upright, and I found several posts that clearly state not to lay horizontally, but nothing about outlet down.  Regardless of whether this is OK or not, could I potentially have some kind of explanation to any side effects of doing it?

2) If there are two pumps in series, do they both require liquid being filled a couple centimetres from top, or only one?

I also have a XD5 for this loop.  I have seen it explicitly stated to NOT fill the XD3 (possibly XD5 as well) completely to the top.  Is this just to leave room for expansion/contraction from heat and/or provide a 'safe' place for air to collect?  The XD3 will be higher in the loop if that makes any difference.

Thank in advance for any information provided 🙂

[LeDoyen]

A question that may solve your problem entirely : Why two pumps in a single loop? 🙂

Dual pumps are very rarely needed, usually for over the top builds with huge restrictive loops, like workstations with 4+ GPUs and a ton of radiators, or decked up super towers like the 1000D, or the LianLi desk PC.. 

Also, mixing a D5 with a DDC pump is going to be tricky to control to say the least. they have different speed specs and head pressures so you will have fun times balancing them out. 

As for having the XD3 sideways, you'll have more chances to suck air through the pump inlet. It will sit higher in the water. That can damage it with time.. and the XD3 doesn't need that given its tendency to cause early RMA.

If i were you (and if your build is the one in your hardware profile) i'd save the costs of the XD3 and go with a single XD5. 

[Smokester]

Hi @LeDoyen , thanks for response.

Curious, why would outlet at the bottom increase chances of air in system? You mention the inlet being the issue, could you clarify why?   This brings me to my other question I asked above:

2 hours ago, Smokester said:

2) If there are two pumps in series, do they both require liquid being filled a couple centimetres from top, or only one?

Would filling the system through the XD5 and leaving air space in that while removing all air from XD3 solve the issue of air from the inlet?  Would there be any other performance consequences or the more serious reliability issues outlined in laying the XD3 horizontal (creating a vortex I think I read somewhere)?

Manual specifically states mounting vertically  (but nothing about rotations) and the general rule of thumb is for outlet to be below reservoir, so this seemed reasonable. Totally understandable if you don't know for sure.  Was just checking here for some clarity from community or possibly Corsair themselves.

 

Why two pumps? Well, there's plenty of reasons, redundancy and noise are top of the list though. Reducing wear and tear and keeping the XD3 cooler by running slower is a plus, but not a reason. An unintended benefit, figured I could capitalize on higher flow rate of D5 and higher pressure of DDC, however you are implying that is not the case and that I need to "balance" the pumps.  To be clear, the plan was always to run at similar PWM curves.

I am no watercooling expert, and would love to know any theory if I am wrong, but this is my understanding:

So long as both pumps are working under their P-Q curve (or more specifically, their safe operating zone),  and within reasonable pressures (cavitation) of each other, everything is fine and no "balancing" is required when mixing and matching pumps. The pumps will simply add their pressures and the flow rate will adjust to the new P-Q curve

Since pumps in a series will not increase flow rate directly there is no risk of flow rate ever going beyond rated capacities (not to mention flow restrictions).  As for pressure, I have not found Corsair's exact rated maximum operating pressure, but have seen other DDC pumps rated for 1.5 bar (15 metre head)  and D5's for 10 Bar which would both be far beyond any doubling of pressure combining these pumps could do.  Outside of a stress test (which if there are any notable reasons there could  be an issue I would certainly modify it), these pumps would never be running anywhere near full power. 

I welcome any input here.  If there are some serious concerns I am not aware of I'll certainly look into it. Need an excuse to buy a good flow meter anyway 🙂

 

As for why this combination? XD3 came in kit at a pretty substantial discount, but due to noise and reliability complaints out there for the XD3, decided I would prefer an XD5, so grabbed an XD5, later on, also on sale. I am fully aware of the overkill in power but this system will be run nearly 24/7, downtime is not acceptable (hours are OK, days/weeks/months are not) and thus redundancy and noise are crucial. Also, "why not?"

But all off topic, "why" I want to do this isn't really important. Just want to know of any potential issues of mounting in that direction and how to manage air in the pumps.  The air from inlet is interesting and possibly question #2 answers #1?

Thanks for response 🙂

 

[LeDoyen]

There's no wear to speak of honestly since the impellers are water lubricated, and only spinning around a small ceramic ball. If you use a clear coolant there shouldn't be any significant wear, probably for the lifetime of your PC.

For head pressure, it's usualy just below 500mBar (between 3 and 5m head pressure for D5 and DDC respectively).

I don't know how the XD3 works, but often the Xylem DDC pumps go to max speed at 50% PWM, while the D5 is full speed at 100%. That may be another issue, unless Corsair makes their own controllers and change the PWM response.

If you are going to make a typical loop with one CPU and one GPU, you won't even need to accelerate the pumps. The waterblocks have a minimum required flow below which the temperatures will climb sharply. But once you meet that flow, the block temperature won't change drastically if you increase it.

For example, if you load your GPU with some furmark torture test or an OCCT test, with your pump at say 40% PWM, you may gain 2 or 3° by running the pump at full speed. Higher water flow doesn't mean more cooling. It just means... higher water flow 🙂

A single D5 or DDC is already more than enough for such build. If you need redundancy, it's usually better to go with a dual pump instead of 2 discrete pumps with reservoirs.

You could also just have a spare D5 pump (just the pump module) in case yours bites the dust, but D5 are very reliable, and you can find Xylem pumps in every watercooling store anyway and get it pretty quickly if you didn't want to invest in a spare to keep at home. Actually, DDC pumps are also very reliable, when they are not encased in a cheap plastic enclosure with no airflow 🙂 D5 is watercooled, and DDC is air cooled, which is why i'd rather advise to keep the XD5 if you go single pump.

Custom loop pumps are very dependable. If it doesn't die right away from electronics failures, you know it will last for years.

 

For your question 1, it's not the outlet at the bottom that is the problem but the inlet being raised higher up. The risk of ingesting air will be a lot higher and that's how you damage a watercooling pump. They are probably not powerful enough to cause cavitation but the strong disturbance of passing air will damage the bearing or the impeller.

The outlet has to be below the reservoir if the reservoir and the pump are aligned. With a sideways pump it changes, obviously. The only red lines are to not have the pumps head down, and have the inlet always covered with water. beyond that, it's up to the manufacturer to arrange the ports as they like. The problem with the XD3 is that the reservoir is extremely small so there's not much water above the inlet. by rotating the pump 90° you further decrease that gap. 

#2 i think you won't have a choice. it will be extremely hard to purge all air from the XD3, maybe even impossible. same deal if you wanted to fill the XD5 completely. 

 

[c-attack]

The XD3 has a very specific intake action and altering the angle is going to have a negative impact on performance and noise.  The actual pump inlet is in the center of the reservoir and it has specific guide channels designed to work with gravity.  By turning that side up, you move the vortex action into a place it shouldn't be.

 

 

 

I don't see any reason to use the XD3 at all in this build unless you don't have space for the XD5, which does not appear to be the case.  You are not gaining anything by running both pumps in tandem.  You are likely to add noise.  DDC pumps don't do quiet all that well.  They are geared like a truck and produce high pressure at low speeds, but it's at 3700 rpm at 50% and I assure you that is loud.  For what you are describing, a XD5 or any D5 pump is what you want.  It can set low and quiet when there is no load.  It is capable of higher throughput at maximum at it will make less noise at 100% than the XD3 does at 50%.  I use a dual DDC and D5 pump combination, but I also am running an open case with dual external 480mm radiators and a solid 1 meter of vertical climb.  Inside a normal case I would have no need for two.  It's the extremely tall nature of the build that makes two pumps worthwhile, as well as the ability to put the DDC below the deck with the other radiator where I can't hear it.  

[Smokester]

@LeDoyen Thanks again for the well written response.

8 hours ago, LeDoyen said:

For head pressure, it's usualy just below 500mBar (between 3 and 5m head pressure for D5 and DDC respectively).

I think this was into my response about max operating pressure.  What I was specifically referring to was the theoretical pressure it could withstand as opposed to what it can pump out... ie working within it's safe limits.  But ya, combining the pumps is well within the limits.

 

8 hours ago, LeDoyen said:

Xylem DDC pumps go to max speed at 50% PWM

You are pretty close in regards to Corsair.  According to Techpowerups review they showed 4711 - 4545 - 3645 RPM for 100%-60%-50% PWM respectively with flow rates and sound very close 50 and 100%.

 

8 hours ago, LeDoyen said:

If you need redundancy, it's usually better to go with a dual pump instead of 2 discrete pumps with reservoirs.

As mentioned, the pumps are purchased.   I have them, can't return them.  I could accept the XD3 as a sunk cost at this point and keep as a backup though.

 

8 hours ago, LeDoyen said:

DDC pumps are also very reliable, when they are not encased in a cheap plastic enclosure with no airflow

With cooling in mind, it is mounted  with an air INTAKE fan directly to it's backside and will be taking coolant intake from the radiators.  I doubt those are going to make huge differences since it's not water cooled and the plastic on the backside is a terrible heat sink but combined with running at 30-40% speed I figured heat shouldn't be an issue or at least minimized as much as possible.

8 hours ago, LeDoyen said:

#2 i think you won't have a choice. it will be extremely hard to purge all air from the XD3, maybe even impossible. same deal if you wanted to fill the XD5 completely. 

I assumed if only the XD5 was running during fill process it would be similar to removing air from rad.  Guess I'll find out if I do indeed move forward with build, even if it's in the normal vertical position.

7 hours ago, c-attack said:

The actual pump inlet is in the center of the reservoir and it has specific guide channels designed to work with gravity. 

@c-attack Thanks for response.  This makes sense and I think ultimately answers the question. 

 

7 hours ago, c-attack said:

I don't see any reason to use the XD3 at all in this build unless you don't have space for the XD5, which does not appear to be the case.  You are not gaining anything by running both pumps in tandem.  You are likely to add noise.  DDC pumps don't do quiet all that well.  They are geared like a truck and produce high pressure at low speeds, but it's at 3700 rpm at 50% and I assure you that is loud. 

The question of need, was never a question.  As  mentioned, I bought the kit including the XD3 at a significant discount and decided I would prefer an XD5,  Waited for a sale and bought that as well.  I now have both pumps and neither can be returned.  I have them and want to use them.  As stated to Ledoyen, I could accept the XD3 as a sunk cost for getting such a good deal on the  kit, but more so just want to make it work if I can.  The goal was to run XD3 at 30-40% speed at the significantly lower RPM's /noise levels of 50%+ with the same planned for the XD5 and seeing how temps compare to increasing RPM's  and introducing noise.

 

Question 2 is still unresolved here though:

On 11/22/2022 at 1:14 PM, Smokester said:

2) If there are two pumps in series, do they both require liquid being filled a couple centimetres from top, or only one?

Is this just to provide space for heat expansion/contraction and if so, is it only necessary in one pump?

 

Lastly I should note. I don't live in the States. Finding watercooling parts is not as easy as going to Micro Centre or Amazon.  My choices are far more limited and unless I want unreviewed Chinese brands, I have to either order directly from EK with significant increases in costs, shipping and shipping delays or something similar.  Corsair on the other hand services my market well and is  priced appropriately.  So the act of 'just get another if one fails' isn't something I can necessarily bank on, and thus redundancy is highly desirable.

[LeDoyen]

depending where you live, Aquatuning could be a retailer to consider for spare parts.

The thing is, with those two very different pumps, you may introduce more problems trying to solve a nonexisting issue. Keeping the XD3 as backup is pretty viable.

you may not even need to run the XD5 past 50% PWM to be honest. and that one could take runing at full speed all year long, they are sturdy AF.

The only failures i see on the forum related to XD5 are people disconnecting/reconnecting the PWM cable while the pump is powered up.. mostly user errors.

Either way, if any of the two pumps die, you'll have to purge the loop to take it out, so you might as well only use one at a time.

As for the air in the reservoir, you want to have some for expansion yes.  Now, it's not a car radiator. you may have 15 - 20° between coldest and hottest, but it's not a good idea to test how high the pressure can get before it leaks 😛

Besides, it's nice to see some water movement in the res 🙂

[LeDoyen]

28 minutes ago, Smokester said:

You are pretty close in regards to Corsair.  According to Techpowerups review they showed 4711 - 4545 - 3645 RPM for 100%-60%-50% PWM respectively with flow rates and sound very close 50 and 100%.

 

That was for the XD5?

DDC usually don't go above 2200 RPM (4400 on the reading) and D5 up to 2400 RPM (4800 on the reading).

What i saw was that D5 get to 4800 at 100% PWM, while DDCs go to 4400 at 50% PWM (the 50 to 100% range do nothing, the pump is already running at full blast). If the XD3 behaves like a Xylem, and you run both pumps at 50%, you'll have a XD3 at full speed and a XD5 close to half speed

[c-attack]

So on the fill you will disconnect either the XD5 or XD3 from it's power.  Fill the system normally like you have one pump.  I would choose the XD5 as the working fill pump in most all circumstances because of the larger reservoir and easier fill.  The liquid will pass through the XD3 and once it starts freely circulating the entire loop, flip the PSU back off and hook the XD3 back to molex.  Now power up with zero chance of the XD3 running dry.  The XD5 at maximum with the PWM/tach wire disconnected should have no trouble filling things up on its own.

 

On a side note, you might consider leaving the XD3 PWM/tach connector on any MB or Commander fan header so it will not run 100% at max during the fill process.  That thing hits pretty hard when it turns on and may make things a bit frothy.  Once you are up and in the OS with CUE, you can play around with both pump speeds including maxing both out to power out any air pockets.  This will be a more controlled procedure with exact speed settings available.  Over the years there have been a few funny user stories about terrified cats, wives, and thrown screwdrivers when the XD3 kicked on the first time.  

[Smokester]

3 hours ago, LeDoyen said:

That was for the XD5?

XD3.  Guess he going off the reading if reading is different than actual for some reason.  Regardless, was just saying his data for the XD3 compares with what you  were saying. Although in his sample, the drop off appears to be below 60%, I'm sure there's going to be variance between pumps/tops/testing methods.

https://www.techpowerup.com/review/corsair-hydro-x-series-xd3-pump-reservoir-combo/5.html

4 hours ago, LeDoyen said:

What i saw was that D5 get to 4800 at 100% PWM, while DDCs go to 4400 at 50% PWM (the 50 to 100% range do nothing, the pump is already running at full blast). If the XD3 behaves like a Xylem, and you run both pumps at 50%, you'll have a XD3 at full speed and a XD5 close to half speed

Ya I'll certainly have to fool around with it when I get there.  That was why plan was to run DDC @ 30-40%, because I had seen that behaviour from the review  I just linked above.  I can easily adjust both pumps to run at their maximum inaudible speeds for most uses and only increase more if I can see any discernible impact in temperature under load.  My very basic napkin math figured I would get about 1 gpm running both in that scenario, so I'm not expecting much benefit going louder, but that's the point of testing.

4 hours ago, LeDoyen said:

Either way, if any of the two pumps die, you'll have to purge the loop to take it out, so you might as well only use one at a time.

Yes but the key element here is the system will keep running (albeit hotter and would need to adjust speed of working pump to compensate) and could be done at leisure instead of shutting down the middle of doing something important.  But agreed, a failed pump would need to be removed.

4 hours ago, LeDoyen said:

As for the air in the reservoir, you want to have some for expansion yes.  Now, it's not a car radiator. you may have 15 - 20° between coldest and hottest, but it's not a good idea to test how high the pressure can get before it leaks

Absolutely agree! Just wanted to know/confirm the theory.  I'm now curious if I'll be able to keep air in both if I'm only filling with XD5. Guess I'll find out, but by the sounds of it, as long as one has air space, should be fine.  If I can leave some space in DDC (D5 WILL have air space, filling with that) I will though.

 

@c-attack Thanks, yes using the XD5 for filling is the plan 🙂 .  Will only hook up the molex and mobo jumper to psu while filling the loop, and not the commander pro or anything else related to the PC. My only concern was whether both pumps should be given air space at top or if only one needed air that space. Commander pro and everything else connected to PC will come after leak testing.

4 hours ago, c-attack said:

Once you are up and in the OS with CUE, you can play around with both pump speeds including maxing both out to power out any air pockets.  This will be a more controlled procedure with exact speed settings available. 

Yup that's the plan. Although I will fire up the XD3 at 100% during the leak testing.  I'll be sure to turn off the PSU before switching any cables.

4 hours ago, c-attack said:

Over the years there have been a few funny user stories about terrified cats, wives, and thrown screwdrivers when the XD3 kicked on the first time.  

Looking forward to that 🙂

 

All in all, I have both pumps installed so I will be using them.  Through the testing if I do find the DDC to be completely useless or increasing noise to levels higher than with the D5 alone, I do have option of removing it as you guys suggest.  Given the sunk costs, might as well see if lower noise can be achieved for similar performance by running both at lower speeds.  I'll post some temps of each pump at various RPM's both together and individually when I'm all done in case curious.

[Smokester]

Was still waiting on a couple fittings to come in and do not have a GPU block yet, but figured I would do a makeshift build for now to test my system, the pumps (given all the DoA claims) and just get things up an running in general.  But as promised wrote out findings in case anyone googling ever cared.

The findings were mostly expected with a couple oddities. As expected, 1 pump is more than capable handling 2 rads, a cpu block, 5x 90 degree fittings,  and a slightly abnormal amount of vertical (about .8m) for my build size. 

Loop order was DDC > CPU > D5 > 30mmx360mm rad > 58mmx360 rad > DDC .  DDC was mounted as instructed (vertical and NO 90 degree shift). Temperature sensors were in the D5 and DDC.  Because of this loop order, working under assumption that the sensors are both accurate, the temperatures give the exact temperature going into the CPU (DDC), and the AVERAGE temperature coming out of the CPU before the rads (D5)... Correct me if I'm wrong about that, but assumption is that the D5 temperature, given time for reservoir to adjust, will be a true representation of the average coolant temp coming from CPU.

Testing was done with about 30 min (or until steady state) of p95 small fft's. Testing 1 pump at 20, 40, 70 and 100%, both pumps at 20, 25, 50, 75 and 100% as well running them with one at lowest PWM  speed (10%  was the setting but likely closer to 20% was the true minimum) and the other at 10, 40, 70, 100%.  Intake fans were set to 60% and the Rads (exhaust) 50%.  The tests provided good insights to make educated guesses on flow rates by monitoring coolant temp (lower = lower flow rate) along with the knowledge of where CPU temps began to suffer from lower flow rates (I believe in practice anything over .75-1 gpm usually sees heavy diminishing returns depending on water block)

CPU temp ranged from 74-88 C and coolant from 24-32 degrees C.

 

TLDR: Flow rates scaled higher with 2 pumps (shown  by coolant temps increasing with =< CPU temps).  At low RPM's specifically, the pumps were superior together than alone.  XD3 performed well solo but better with XD5 at low rpms.  The XD5 did not perform well by itself at low RPMS, producing the highest temps by far at 20% and much more reasonable, but still the next worst temps @ 40%.

 

The Expected Results

1 pump even at 40% produced temperatures only a few degrees above the lowest (DDC was 1 degree lower than D5 here).  The DDC at 20% produced was not much worse than 40%. The D5 at 20% however struggled mightily with a 12 degree increase in temperature and by far the lowest post rad temps (indicating major flow issues). 

2 pumps at 50% were essentially equal to 1 pump at 100%  and both pumps worked very well with each other at 25% each, to produce temps only 1-2 degrees higher than the best temps (and outperforming a single pump at 40%).

DDC became inaudible around 50% while in the case and under the desk.  D5 was inaudible below 70% but even at 100%, was not bad.  The pitch of the DDC/XD3 more than anything what makes it unbearable at high speeds more so than the actual volume. That said, the XD3 volume at high speeds is definitely untolerable.

Overall impression of XD5 was that it was an excellent performer for a minimal amount of noise, specifically at high RPMS.

 

Unexpected Results

1) My XD3 actually has a pretty linear curve in terms of RPM/PWM%.  60% PWM was roughly 68% of max RPM and even the higher PWM%'s scaled quite linearly

2) XD3 performed very well at low rpms, matching or beating the XD5 at equivalent PWM settings.

3) The XD3 seemed to introduce heat into the loop, which I was of the understanding that DDC's do not behave this way (perhaps this is a pro of Corsair design?).  It was really strange, but the higher the XD3 speed, the higher the pre CPU temperature (to the point in some cases was higher than the XD5/pre rad temp).  There is clearly something wrong with these readings as pre cpu temp can't be lower than pre rad temps... Potentially the XD3's pump heat just bled into the temp sensor (mounted bottom of XD3) but without actually increasing the outlet temp.  This temperature scaling was consistent throughout all the tests (higher XD3 speed meant greater XD3 temp) and not an outlier or anomaly in my tests.

[LeDoyen]

your unexpected 2 and 3 are very expected ^^

2 : the DDC has more head pressure so it will have better flow at the same RPM as the D5. it overcomes restrictions with more ease, where the D5 needs to be ran faster for the same result.

3 : your asumption seems correct to me. The DDC warms up quickly without heatsink. it's very likely to be the motor heat you're sensing in the probe.

The D5 will have such an effect too since it's sinking heat in the coolant, but it's more subtle, and negligible when dealing with a CPU, let alone a CPU plus GPU in the loop.

I have an unused pump/res a decoration on my desk, doing water noises, with about half a liter of coolant and no radiator obviously. when i crank up the D5, it warms up the liquid from room temp to warm to the touch in 10 - 15 minutes. I never bothered to measure but it does get warm. but agaian.. negligible.

From your test, you can see that basically, past 40% on the D5, you'll get marginal benefits on temps. the temp increase when the flow is too low is pretty sharp!

now for a CPU only, that doesn't mean much yet. as you saw, it will perform about the same wether you're at 77 or 88°C. you'll have to watch water flow a bit more when having the GPU in. They are more sensitive to low flow given how easy GPUs are to cool compared to CPUs.

With a sensible pump curve, you'll rarely see over 3°C of delta T between the radiators input and output under heavy load, after reaching steady state. It's not really worth it to seek the lowest deltas for better cooling. the radiators need to warm up to be efficient, and in the end, slowing down the pump and the fans can give better results for comfortable use.

[c-attack]

Be careful of over-analyzing the temperature readings.  Somewhere in this forum there is a thread where I discussed having to go through 20 thermal probes to get 4 that were "precise" enough to use for measuring to 0.1C.  Those probes were from 6-7 different manufacturers and the results not consistent between the same model.  While I am a data hound and you may be as well, there is something to be said for sticking one temp probe in and not digging in too deep.  I am not sure if I could do it, but most users should probably aim for that.  Chances are your XD3 temp probe reads higher than the XD5.  You always can switch them so it's not so obvious.  I can't deal with it either.  

[Smokester]

3 hours ago, LeDoyen said:

your unexpected 2 and 3 are very expected

Suppose with #2 was more so the degree to which it outperformed.  Overall was very impressed with XD3 in the lower ranges when noise wasn't a factor.  The XD3 was serviceable at 20%, the XD5 was not.  And ya, for #3, have to use better judgment here.  The post rad temperature can't possibly be higher than pre rad, that was the 'surprise' there... So something weird was going on there. Was aware that D5's were liquid cooled and DDC's were not, so that also added to the 'wtf?' of it lol.  It actually was quite noticeable tho in my readings.  When XD5 was running faster than DDC the temperature differences became very pronounced (as much as 7-8 degrees at low speeds).

3 hours ago, LeDoyen said:

the temp increase when the flow is too low is pretty sharp!

Yes definitely.  This was expected and essentially the overlying purpose of the tests...  To find minimum running speeds.  Planning to run DDC at 30% and D5 at 40%.  With absolutely no certainty here, I am estimating there is a P-Q intersection around this area, and thus achieving flow 'balance', whisper quiet operation and high performance at very low speeds.

3 hours ago, LeDoyen said:

slowing down the pump and the fans can give better results for comfortable use.

^^ Agreed. See response above.  And ya, the Delta was less than 3 Degrees in most cases, in many of the high flow cases, less than .5.  The rad fans were only running at 50% also.  Speeding them up would reduce the delta even more.  Stuck with 50% to provide a good amount of active cooling while not completely nerfing the data.  I'll probably keep them around that speed in day to day useage anyway

3 hours ago, LeDoyen said:

you'll have to watch water flow a bit more when having the GPU

Back to the drawing board at that point

🙂

2 hours ago, c-attack said:

Chances are your XD3 temp probe reads higher than the XD5

Very possible, and thought crossed my mind.  But the trends were very consistent and outcomes mostly expected with exception of the heat from XD3, but even then, it followed the trend.   I did mention I was working on assumptions that they were indeed calibrated properly.   They both however were the supplied ones from Corsair, and while there certainly will be variance in manufacturing, there can be some level of confidence that they work similarly. And in case there was any confusion, the XD3 probe very rarely was hotter, only when  XD3 was running substantially faster the XD5.

2 hours ago, c-attack said:

While I am a data hound and you may be as well, there is something to be said for sticking one temp probe in and not digging in too deep.

I am as well 🙂

This wasn't meant to be anything official. But was nice to see the data. And overall, like I said, the data was very consistent.  Having the 2 probes configured like I did provided valuable insight into the flow rates which was one of the goals here.  Since there are many conflicting opinions as to whether the pumps need to be ran at equal power or not, it is some nice peace of mind that I can pick a speed for both pumps that appears to be an intersection of P-Q curves (as mentioned in response to LeDoyen, can't say with certainty, but it is a reasonable estimation).  Hopefully the Aquaflow Next will be available by the time I get my GPU block and can do some more in depth testing then.

 

 

Thanks again to both of you for the interesting discussion 🙂

[LeDoyen]

High flow next? 

It will be more helpful than rely on temperature probes to adjust the pump curves. At least, it will be quicker. Each waterblock has a sweet spot in terms of flow below which temps will spike up a bit (and it won't be visible much on the water temp probes, hence why a flowmeter helps there).

For my particular GPU one, under 80 L/h, temperatures start climbing. That's 0.3 gpm only. keeping the flow above 100l/h works perfectly even with demanding games.

So, you may find, depending on what block you use and how restrictive your loop will be that you don't need much pump speed to get going. That was basically why i was saying earlier that two pumps is much overkill 😛

Edited November 29, 2022 by LeDoyen

[Smokester]

9 hours ago, LeDoyen said:

High flow next? 

Yes I meant Aquacomputer's High Flow Next. Typo in previous post.

9 hours ago, LeDoyen said:

So, you may find, depending on what block you use and how restrictive your loop will be that you don't need much pump speed to get going. That was basically why i was saying earlier that two pumps is much overkill

Haha ya. TBF we were never in disagreement there. I always knew it would be overkill.  Having 2 pumps was a matter of circumstance and using both was a decision of redundancy mostly, with a touch of noise and a bonus of implied reduced wear and tear.... and also a question of "why not".... as in, you've come this far..... speaking for myself and not others, my feeling is custom water cooling, much like overclocking or buying a car with over 300hp is a hobby more than functional need.  While I do have a frugal, value seeking nature, the moment I bought the 303i (Corsairs Kit) I knew I was embarking in a new hobby and the only notion of value here was the substantial discount I got on the kit, and the value of the kit vs buying pieces individually, not cooling /$.  So ya, 'why not' take it another step

9 hours ago, LeDoyen said:

For my particular GPU one, under 80 L/h, temperatures start climbing. That's 0.3 gpm only. keeping the flow above 100l/h works perfectly even with demanding games.

Ya my understanding of modern GPU waterblocks is that they are less restrictive and in part due to the direct die contact are much easier to cool despite generally producing more heat.  I'll still have to test, but I doubt I'll have to adjust my pump(s) curve much.   Do you have a separate line for your GPU or is that 100l/h also cooling your CPU?  Seems a little low for a CPU but I could see how gaming wouldn't be a significant stress on CPU (as opposed to P95 small fft's).

[LeDoyen]

it's a single loop, GPU into CPU, and 3 slim rads in series (2x 360 + one 240) with a DDC. Not much effect on the CPU no matter what test it runs, cinebench, prime95.. i don't need much more flow. The CPU temp differences between 80 and the max i can reach at 220L/h under cinebench is about 7 or 8°C

I still vary pump speed but only between 120 - 130 L/h (~1300 RPM on the pump, 2600 on the reading). A hefty flow margin, no pump noise at all (on my build, the DDC is a lot quieter than the D5 i had before on that reservoir, even at full speed).

And yea, it's definitely a fun hobby, totally not necesary, hence indispensible 😛 If you're into data hoarding and stupid levels of optimization and automation, definitely have a go at Aquasuite since you'll have a licence for it with the HFNext 😉

[Smokester]

13 minutes ago, LeDoyen said:

it's a single loop, GPU into CPU, and 3 slim rads in series (2x 360 + one 240) with a DDC. Not much effect on the CPU no matter what test it runs, cinebench, prime95.. i don't need much more flow. The CPU temp differences between 80 and the max i can reach at 220L/h under cinebench is about 7 or 8°C

Ya that makes sense.  Wish I had a flow meter to test that myself, but realistically the only number that mattered in my case was that D5 at 20% was not serviceable, so as @c-attackeluded to, there is something to be said for not over analyzing.  But that's not me lol.  The whole hobby thing; I like deep diving.  Been doing overclocking in some form since core 2 duo's and more extensively since devils canyon (4790k) which included delidding and lapping.... not because I should've, but because I could lol.

31 minutes ago, LeDoyen said:

I still vary pump speed but only between 120 - 130 L/h

Based on which sensor? Coolant/CPU/GPU? Was thinking I would eventually set my rad fans to vary based on coolant temps and the pumps to vary based on GPU/CPU temps but with a very high hysteresis to prevent needless pump speed variations for short delta changes.  I have heard frequent changes in pump speeds can affect their life span, not sure if anyone outside of manufacturers could really confirm that though, but regardless, small and less frequent changes is probably still a good practice. Not to mention noise fluctuations if that was a factor (not in our cases... but could be for some).

20 minutes ago, LeDoyen said:

And yea, it's definitely a fun hobby, totally not necesary, hence indispensible 😛 If you're into data hoarding and stupid levels of optimization and automation, definitely have a go at Aquasuite since you'll have a licence for it with the HFNext 😉

I use Fan Control v137 currently, definitely love it. Jays2cents really did a number on popularizing it.  I've seen videos on aquasuite, and would need to brush up if I did indeed buy it, but do you have any experience with it? Better than something like Fan Control?  Fan Control sort of works with Commander Pro, although sensors are sometimes sketchy even without iCue....  which is nice, and is great without having a GPU waterblock where I can set the bottom intake to follow either GPU or CPU curve depending on what's needed. 

With Aquasuite I believe the license is strictly for software updates right? Ie when license expires it doesn't change any functionality, just removes ability to download updates to software?

[LeDoyen]

pump speed and fans change based on the delta between water and room temperature. this is the value that governs radiator efficiency, and obviously, no need to readjust the fan curve in summer or winter.

If it's 35°C in the house, no need to blast the fans as a curve based on water temp alone would do. if it's 35, and water is at 40, the fans will spin at the same speed as they would if my room was at 20 and the water at 25.

Aquasuite allows to do a lot more than fan control, but obviously only works with aquacomputer controllers. Describing all you can do with it would warrant a post that i'd still be writing tomorrow. Maybe the feature i like the most is the virtual sensors. you can use just about any input you want (temperature sensors, flowmeter, anything from HWinfo, power, voltages, speeds etc etc.. and combine them, make operations, logical, mathematical, to get any kind of output you desire.

could be something simple as getting that water / ambient temperature difference for fan control (two inputs, one substraction), or another one i use that takes the temperatures before and after radiators, combined with the water flow to calculate the dissipated thermal power in watts, taken off the loop in real time.

there's also a handy inputs section where you can create manual inputs for those sensors. that allowed me to make two modules : one to trigger a cooling cycle that will blast the fans until the water falls below 3°C over ambient, to cool off the PC quickly when i play late at night to cool off the PC before switching it off, or another one i made for the pump speed that will cycle it between slow and full speed for whatever duration i want, and a number of cycles i chose. I use that one to purge the air off the loop when i drain and refill it. It gets rid of trapped air pretty nicely.

Can also use virtual sensors just to alter slightly a sensor response. Like doing some averaging or adding hysteresis for a jumpy sensor, like CPU temperature that jumps up and down al the time, or the ambient temperature sensor that can move a lot when i get a fan going in the room in summer.

It can be daunting to the new user, but if you're curious, it's a rabbit hole worth going down into 🙂 you can go simple, or go to town on loop control and monitoring😛

Oh and you're right about the licence. it's just for update access, to support the software development. You always have access to the latest update your licence got you, unless there's an important security update or a crippling bug on your version, in which case they would offer the next stable one.

Edited November 29, 2022 by LeDoyen

[Smokester]

5 hours ago, LeDoyen said:

pump speed and fans change based on the delta between water and room temperature.

Excellent point. I think that's exactly how I'll setup my rad fans.  Curious though, my thoughts of increasing pump speed on coolant temperature felt like it might be a bit of a trap if increased flow leads to increased coolant temperatures.  Obviously not an issue for you though. 

 

Thanks for your "short" summary on aquasuite, can tell you have a lot to say about it, I probably would've read your essay in entirety lol.  I'll certainly go down that rabbit hole when time comes.  All in all sounds like you recommend HFN (don't think you specifically said you used it, but obviously at least one of their products)?  Would love to order it directly from them, but shipping alone is 60$.  Probably have to wait for Amazon to start selling again once they catch up with backorders.

7 hours ago, LeDoyen said:

one to trigger a cooling cycle that will blast the fans until the water falls below 3°C over ambient, to cool off the PC quickly when i play late at night to cool off the PC before switching it off,

Not sure if you're a car guy or not, but sounds a lot like a turbo timer lol.  What's the benefit of cooling before shutdown?

7 hours ago, LeDoyen said:

another one i made for the pump speed that will cycle it between slow and full speed for whatever duration i want, and a number of cycles i chose

Might have to pick your brain on that once I officially get my loop done properly (had to ditch my drain valve while I wait for rest of parts to come in).  Just did a purge today to replace distilled water was using with some Corsair coolant (didn't have separate additives for using straight water) now that I can confirm loop working as intended.  Next purge and redo of tubing will come once I get GPU block in a couple months and can properly hookup drain hose.

 

7 hours ago, LeDoyen said:

or adding hysteresis for a jumpy sensor, like CPU temperature that jumps up and down al the time,

Ya this was why I was saying I would create some sort of high hysteresis if I were to create a variable pump curve. Probably very low step/down also.

[LeDoyen]

Increasing pump flow doesn't lead to higher water temps (unless your sensor picks up pump heat obviously, and then you should relocate the sensor). You'd want to try to keep the water flow high enough that the temp difference between rad in and rad out don't creep up too much. That's one of the reason i use more flow than strictly necessary.

In idle that delta is about 0.5°, when gaming, it may climb up to 2,5°, so that's negligible in terms of perturbating fan control. you really have to have very low flow to start messing with fan response.

 

I don't recommend specifically the high flow next, i thought you had it ordered already 🙂

I personally use a high flow 2 because i didn't want a dedicated USB flowmeter. Mine plugs in an Octo controller's flowmeter port, and the controller manages the whole loop.

regarding cooling the loop before shutdown, it has no benefit for the computer, only to me since my bed head is right against the computer. Just personal comfort as i like to cool down the room before sleep.

and for fan control or pump, since it's always based on water temp, that alone will add a ton of hysteresis. All you really need is some smoothing on the temperature probes, and only if the readings are a bit jumpy.

 

[Smokester]

1 hour ago, LeDoyen said:

Increasing pump flow doesn't lead to higher water temps

My understanding is that lower flow equals more time in rad and thus more time to cool down. Delta obviously increases too since more time on waterblock  increases pre rad temps also but that added heat is easily negated by rads.  I 'backspaced' my caveat about that in earlier post that you also  were increasing fan speeds too which cancels that out. But ya, it's clearly not an issue in your situation.  Flow being higher than "strictly" necessary makes sense in this context.

1 hour ago, LeDoyen said:

I don't recommend specifically the high flow next, i thought you had it ordered already 🙂

I personally use a high flow 2 because i didn't want a dedicated USB flowmeter.

For the reasons you don't want it are the reasons I do lol.  Don't want to invest in controller also since I have a motherboard with Temp headers, along with commander pro (which I might scrap for several reasons) and a couple extra usb headers available. HFN just seems to fit the bill in terms of credibility/reputation/quality/accuracy, interfacing, aqua suite, just not for 200$ (CAD, with shipping and FX etc).... 

Just tried Aquatuning, which I think you suggested, to find a Aquacomputer MPS 400.  50$ part, 60$ shipping and then add  like 35% for  FX... ya it's a little rough getting parts here lol.  I just have to wait. I'll be in Florida  in March, can try and find a Micro Centre then maybe... stock up on some parts.

[LeDoyen]

importing is a b***, same deal for me when i want parts only available in the US. Here obviously AC parts are easy to get.

And yea definitely use the motherboard temperature sensor headers. the best i could get on a commander pro was temperature readings with +-0.5°C precision. getting 1°C swings isn't a big deal when air cooling but on watercooling where you coolest and hottest may be apart by 10°C, it's pretty annoying.

I try not to advertise competitor products here, but given your mindset regarding watercooling, you know what i'd say 😛

For the water flow, as you said, decreasing flow to get more residence time in the rads is negated by longer time in the blocks., it always equalizes so it's kind of an irrelevant parameter.

Your best bet is to keep the waterblocks fed with sufficient flow to cool well, and really just adjust the pump speed by ear after that, to get it silent, wether you go fixed speed or variable.

[c-attack]

3 hours ago, Smokester said:

My understanding is that lower flow equals more time in rad and thus more time to cool down. Delta obviously increases too since more time on waterblock  increases pre rad temps also but that added heat is easily negated by rads. 

Sort of, but the problem with lingering fluid in the block is the heat is transmitted back across to the CPU/GPU.  So regardless of what happens elsewhere in the loop, 32C liquid temp in CPU block = 32C CPU temp at 0v.  29C liquid temp in the block = 29C CPU temp at 0v.  Lower in block temp is better and this is the real advantage of faster flow rates vs "number of trips to radiator and back per minute vs fan speed heat dissipation" kind of thing.  This is where flow design in your blocks comes in and when you drop the flow rate/pump speed below the magic threshold, you'll see the CPU temps pop up several degrees immediately.  You won't have to wait for that unit of water to make it's way around the loop and return.  This is also why my favorite sensor is CPU+GPU out right after the heat pickup since it is the most telling about the load and heat added.

Edited November 30, 2022 by c-attack

[Smokester]

2 hours ago, LeDoyen said:

And yea definitely use the motherboard temperature sensor headers. the best i could get on a commander pro was temperature readings with +-0.5°C precision. getting 1°C swings isn't a big deal when air cooling but on watercooling where you coolest and hottest may be apart by 10°C, it's pretty annoying.

Can't speak to accuracy, but my commander pro is reading to 100th of degree in iCue.  Not sure if you were specifying accuracy or just readings. 

Wouldn't say you are recommending a competitor product since there really isn't anything competing with Corsair.  ie, Corsair hasn't entered that market. I think you off the hook there 🙂

1 hour ago, c-attack said:

This is where flow design in your blocks comes in and when you drop the flow rate/pump speed below the magic threshold, you'll see the CPU temps pop up several degrees immediately.

I never really thought of the "why" temperatures sky rocket at that magic threshold outside of low flow, but coolant stuck in the block makes a lot of sense.  We are in agreement for just about everything else. Might've been a little confusion but was questioning that in the context that I had wondered if setting a pump curve off of strictly the coolant temp could potentially drive the pump into an upward spiral with increased pump speeds reducing time in rad and thus triggering a higher pump speed.  Given you wouldn't be setting a 0-100% curve, but rather would be something more like 30-50% as an example, I realize it isn't worth worrying about.

 

2 hours ago, c-attack said:

This is also why my favorite sensor is CPU+GPU out right after the heat pickup since it is the most telling about the load and heat added.

The sensor being the "package" (or equivalent) or the coolant temperature immediately after the block? I think you meant the latter but just checking.