Unexpected results - higher flow = higher temps

[virtualrain]

I have an interesting observation and I would like your thoughts...

My temps are higher at high flow rates and slightly lower (but certainly not worse) at lower flow rates.

Loop:

DDC Ultra w/mod /res > DDC Ultra (stock) > BI GTS 240 > BI GTS 240 > Apogee > MCW30 > Alphacool watercooled PSU

Note: I'm not yet watercooling my GPU.

I use a pair of 140mm SL Yate Loon fans on EACH rad (4 fans total). They run at 800 RPM.

I have an Aquaero and use that plus a couple of these circuits to vary the voltage to the pumps from 7V to 12V...
http://home.nc.rr.com/katyandrandy/pwm.htm

The following temps were noted after running dual Orthos Prime for 1hr. on separate days.

At 12V the flow is about 6lpm and temps are:
Room Temp: 21
Water in: 29.7
Water out: 29.5
IHS: 27.8 (thermal probe on side of IHS)
Core: 43 (as reported by Speedfan)
CPU: 44 (as reported by Spedfan)

At 7V the flow is about 3.5lpm and temps are:
Room Temp: 22
Water In: 29
Water Out: 29
IHS: 27.1
Core: 40
CPU: 43

Why would my temps be the same or lower with lower flow? I would have expected worse temps with lower flow.

I assume this means that the cooling effectiveness of my rad/fan combo cannot benefit from higher flow. Is it that the water is moving to quickly through the rads to be effectively cooled?

I assume more airflow and/or better rads would change this so that temps would decrease with higher flow?

I guess one other observation from this is given my system design, dual DDC's is really not helping.

When I get my graphics card in the loop, I suspect I should also look at upgrading my fans from 140mm SL's to ML's to move more air?

I'm interested in your thoughts.

EDIT/UPDATE: After input from a lot of people, I think the link in Post #30 provided by fairydust answers the question as to why more pump is not always better! Here's a review that anyone considering dual pumps should read!

Update: Added some photos.

Here's a few notes that will make the photos make a bit more sense...

- I originally fitted everything using a wooden MDF mockup of my mobo and expansion cards. At the time I was hoping to cool the GPU as well as the chipset but given the layout of the Mobo and other constraints of my design (namely the rads on the side door) could not cool both the GPU and chipset. Knowing I would soon be updating to a Conroe and G80, I decided to wait on the GPU cooling. At any rate, in one of the pics you can still see the wooden computer mockup with GPU block on it.

- All tubing is 7/16" Masterkleer with coolsleeves. The tubing connecting the rad to pump 2 at the top and the CPU block at the bottom obviously had to be given a lot of thought so the door could be opened and closed without kinking the tubing.

- Flow basically starts at the top of the chassis, flows through the rads and then works its way back to the top through the blocks.

- In the top of the chassis at the rear is the water cooled PSU which is the last water block in the chain. That feeds a small DD fillport res which then feeds the inline Remag flow meter before Pump 1 which has the Alphacool res which is mounted in the top front bays. Pump 1 then feeds Pump 2 which is mounted under the blowhole near the res. This then feeds the rad inlet.

- The DD Fillport res made bleeding a lot easier than useing a T. Trying to bleed using a T with dual DDC pumps is not practical. In fact, I had to glue some fish tank sponge to the top of the res to trap air so it would not get sucked back into the loop. I found that with dual pumps any air in the loop or the T will get sucked into the loop.

[mnewxcv]

how long did you run the system with each setting?

[virtualrain]

how long did you run the system with each setting?

The temps were recorded after 1hr of running Orthos Prime on both cores... temps had stabilized long before.

Incidentally, idle temps also exhibit the same differences between running the pumps at 7V vs. 12v... that is, at 7V idle temps are a degree or so lower.

[Reinvented]

Gee, I wonder why...

7v means less heat dump into the water, as opposed to 12v... get the picture?

Doesn't take a damn rocket scientist to figure that one out..

[Vikodemous]

The lower Flow rate is giving the water more time to cool through the rad before it goes back to the CPU.

[L'enFer]

I suspect I should also look at upgrading my fans

may be u should to upgrade wb's?

[MaxxxRacer]

The lower Flow rate is giving the water more time to cool through the rad before it goes back to the CPU.

Please go read the guides before you say such nonesense.

[Digz]

Gee, I wonder why...

7v means less heat dump into the water, as opposed to 12v... get the picture?

Doesn't take a damn rocket scientist to figure that one out..

OOOOOOOOOoooooo (handbags at the ready)

[Mord]

Yep. Slower water = more heat absorbed from the WB's AND more heat dissipated by the RAD. Pretty simple. That's why I run my setup at about 400l/h. The max of the pump is 800l/h. It's SICCE NOVA.

[Herr Brun]

Yep. Slower water = more heat absorbed from the WB's AND more heat dissipated by the RAD. Pretty simple. That's why I run my setup at about 400l/h. The max of the pump is 800l/h. It's SICCE NOVA.

You are wrong. Read the stickies please.

[exhausted mule]

You are wrong. Read the stickies please.

i'm sure that the performance curve of fan/flow/temp delta is not the same for both wb and rad.

everybodies setup will differ and its good to experiment to find your optimum setup.

[Mord]

Could you please explain what's wrong. I think it's obvious.

[Hound53]

Just out of curiosity how did you get your flowrates? Unless you took return line off and actualy measured the flow returning or have a small accurate inline flow meter I would bet you only flowing 3.5 LPM (or less) at the pumps designed head pressure at both voltages.

These are non positive discharge pumps. You could have the designed head with zero flow once it reaches it's designed head pressure, then the water just sit's there and spins (slippage) and builds up heat energy. I once saw a pump hooked up to PVC build up enough heat in the water to soften the PVC to the point where it swelled up and burst because someone turned off #1 pump and shut #2's discharge valve when switching pumps. It was a lot larger pump granted but still the same theory.

If you were to put a ampmeter, a flowmeter, a pressure gauge and a thermometer on a pump then slowly shut down on a discharge valve the pressure would rise to x with y flow and z amps. Keep shutting down on the valve and x and z would reach a point where they remained constant but y would continue to decrease until it hit zero. Study Newton's law of energy conservation look at the thermometer in that discharge line and you will see where the energy( x timesz) went.

[Mord]

Don't know about the others. I measure it the dummy way - start the pump and see how much water gets out the other way in the bucket for 5mins. Measure it and divide by 5. This way I know its flowrate with a good approximation.

[v0dka]

The pump dumps more heat?

Hound53 seems to be on to somthing but I dont understand a word of it lol

[Radical_53]

Strange behaviour. Still, the loop itself is somewhat strange to me (performance in total is not that good I'd say). If you are willing to, try a loop with just one rad and just the modded DDC. To me, the second pump and rad doesn't make too much sense on this kind of setup.

[Hound53]

Don't know about the others. I measure it the dummy way - start the pump and see how much water gets out the other way in the bucket for 5mins. Measure it and divide by 5. This way I know its flowrate with a good approximation.

Hey that works, and is a heck of a lot cheaper than a flowmeter that you would use once in a blue moon.

Hound53 seems to be on to somthing but I dont understand a word of it lol

Centrifugal pump theory and physics 101. In every day terms. Bigger is not necessarily better. Once one of these pumps reaches it's designed maximum head pressure that is the maximum it will flow. It could be rated at 10000 gpm but if the loop only flows 1 gpm at whatever the pumps max head is 9999 gpm 's worth of energy is wasted even though it weill be doiing it's best to move that 10000 gpm somewhere.

[ChongL]

I think we need Marci in here to school everyone

[L'enFer]

I think we need Marci in here to school everyone

or creidiki...

[Hound53]

An alternative way of setting the pump to max efficiency would be use a speed control with a flow meter, once the flow peaks that pump is moving as much water as it can through the loop.

A sensitive pressure gage and a bypass valve back to the pumps suction would acomplish the same thing. Put the gage in the branch that leads to the water blocks, open the bypass back to the pump suction until pressure started to drop. That would give you max flow to the block(s).

[voigts]

I wonder if running those rads in parallel instead of in series would make a difference, but I must say that dual pumps and dual 240 rads is total overkill.

[Modzilla]

Have we figured this one out yet? Im not the sharpest knife in the drawer but I think I have the answer. Inline pumps produce heat during operation and discharge this heat into the water. This heat comes from the power source and friction (friction of impeller against the impeller housing and the water). More volts equals more heat, and faster impeller speed equals more friction which equals more heat. At the higher flow, the pump is fed by 12v and the impeller speed is higher, thus more heat dump than at 7v and slower impeller speed. The speed of the water through the waterblocks and radiators does not affect cooling (not counting friction). All systems will, with time, reach a steady state of heat absorbtion and discharge, thats why temps must be read after the system has been running for an hour or more.
And we are assuming that speedfan and your temp probe are accurate and precise, and that .7 degree is a significant difference for this system. Anyway, I saw the same with my system: i switched out the danner mag 350 gph model for the danner mag 250 gph model and got lower temps. The 250 model uses less power to run and the impeller spins slower. The pumps are identical in all other respects.

[virtualrain]

Wow, lots of input here.

To answer some comments:

- I agree the system is overkill as it stands. As noted, I intend to add the GPU to the loop when I get an 8800 in the next couple of months and perhaps switch to a Storm when I upgrade to a Conroe however, I want to understand this first before I go to the Storm.

- I could definitely try using only one rad and one pump instead of dual pumps and rads at some point, but I'm too busy to rip the loop apart right now to try that. Alternatively, when I redo the loop, I could try two separate loops but the added tubing complexity in my rig may not be feasible.

- As for flow, I have a Remag flow meter in-line... http://www.remag.ch/vision.php?language=4&sub=1&tab=2 I am having some problems with this however... more on that in the next message.

I also tested the flow of the loop using buckets before I started using it and it flowed 6lpm.

- Someone said the performance wasn't all that great... I'm not sure what that's based on, but my temps are now 15 degrees lower than they were on air and the system is a hell of a lot quieter so in my mind: mission accomplished.

- Is there simply not enough restriction in the loop? When both pumps are running at full, the second pump does "clack" a bit... as Marci and others have reported when running Dual DDC's without enough restriction in the loop.

- I'm curious about the merits of the heat-dump theory?

[virtualrain]

An alternative way of setting the pump to max efficiency would be use a speed control with a flow meter, once the flow peaks that pump is moving as much water as it can through the loop.

A sensitive pressure gage and a bypass valve back to the pumps suction would acomplish the same thing. Put the gage in the branch that leads to the water blocks, open the bypass back to the pump suction until pressure started to drop. That would give you max flow to the block(s).

So here is the problem with my flow meter that may or may not be related to this issue... Keep in mind this could be a red-herring as it could be a faulty flow meter or the Aquaero mis-behaving.

According to spec, the Remag flow meter generates 700 imp/l. I have noticed that if I set the Aquaero to this value and then run the pumps at full speed, the flow reported is way off... it's 0.6lpm. If I run the pumps at lower voltage, the flow meter seems more accurate...

However, if i set the imp/l on the Aquaero to 70 imp/l, I get what seem like more accurate flow ratings of 6lpm at full speed and 3lpm at 7V.

Here's the flow meter output at different pump speeds for the rated 700imp/l. Note that 0% = 7V and 100% = 12V due to the voltage divider I'm using...

10% = 1.5lpm
20% = 1.9lpm
30% = 2.2lpm
40% = 2.6lpm
50% = 2.9lpm
60% = 0.1lpm (very erratic)
70% = 0.4lpm (now it's off by about an order of magnitude!)
80% = 0.6lpm
90% = 0.6lpm (no change)
100% = 0.6lpm

Note that I tested the loop with buckets and both pumps running at max and the flow was 6lpm! So again, I say this may be a red-herring and it could be a faulty flow meter or Aquaero.

Update: I confirmed on the Aquacomputer forums that the Aquaero can only manage flow readings up to 3lpm which explains the findings above. So this had absolutely nothing to do with the temperature problem at higher flow rates. Unfortunately, I can only extrapolate my maximum flow which according the to rate of increase of flow with pump voltage would put the max flow at 100% at about 4.5lpm.

[thunderstruck!]

I think we need Marci in here to school everyone

....waits patiently....