AIO/Custom Loop Variable Pump Speed question

[GhostRoadieBL]

what is the effect to the cooling performance of an AIO or Custom Loop's cooling effectiveness by having a PWM pump curve vs full duty cycle all the time? 

 

This will obviously effect liquid dwell time in the blocks and radiators so does a lower pump speed shorten the loop time to thermal saturation or can it cause a lower loop saturation temperature due to the more efficient transfer from block to liquid (more dwell time for heat transfer) and liquid to radiator (wider temperature difference and longer time in contact with fins)?

 

My theory is assuming;

- a consistent heat load source (ie CPU during benchmarking, not looking for "less CPU load puts less heat into the loop" answers)

- a fixed air temp entering the radiators (eliminating atmosphere variables, assume open bench temperature controlled space)

- not pushing the loop flow rate so low as to overwhelm the thermal capacity of the liquid in the blocks

 

I believe, by having the pump at full flow all the time the loop would be less efficient as heat transfer rates require some dwell time and running something like a D5 into a CPU and 240mm radiator at maximum flow rate is going to cause a higher saturation temperature (entropy left over in the liquid after the radiator) than lowering the flowrate to allow more efficient heat transfer from fluid to air by tuning the pump speed to less than full speed. 

 

I don't have a way to test this so I'm looking for someone willing to do a test or maybe has had this thought as well and found the solution to cure my curiosity. 

 

[AngryBeaver]

30 minutes ago, GhostRoadieBL said:

what is the effect to the cooling performance of an AIO or Custom Loop's cooling effectiveness by having a PWM pump curve vs full duty cycle all the time? 

 

This will obviously effect liquid dwell time in the blocks and radiators so does a lower pump speed shorten the loop time to thermal saturation or can it cause a lower loop saturation temperature due to the more efficient transfer from block to liquid (more dwell time for heat transfer) and liquid to radiator (wider temperature difference and longer time in contact with fins)?

 

My theory is assuming;

- a consistent heat load source (ie CPU during benchmarking, not looking for "less CPU load puts less heat into the loop" answers)

- a fixed air temp entering the radiators (eliminating atmosphere variables, assume open bench temperature controlled space)

- not pushing the loop flow rate so low as to overwhelm the thermal capacity of the liquid in the blocks

 

I believe, by having the pump at full flow all the time the loop would be less efficient as heat transfer rates require some dwell time and running something like a D5 into a CPU and 240mm radiator at maximum flow rate is going to cause a higher saturation temperature (entropy left over in the liquid after the radiator) than lowering the flowrate to allow more efficient heat transfer from fluid to air by tuning the pump speed to less than full speed. 

 

I don't have a way to test this so I'm looking for someone willing to do a test or maybe has had this thought as well and found the solution to cure my curiosity. 

 

Higher flow reduces temp variance outside the loop. It does reduce the time water spends in the blocks/rads when you look at it over a time frame, but it does spend less time per trip.

 

Generally higher pumps speed (more flow) will bring better temps. Remember thermal transfer happens more efficiently as the sources get further apart in temps. So having a constant stream of cool water allows more heat to transfer as it is constantly replaced with fresh cool water.  When you slow down the fluid has more time to rise in the blocks which means it can absorb less total energy. So while the water leaving actually is hotter than a fast flow... the temps on the components goes up because of less thermal transfer.

[GhostRoadieBL]

1 hour ago, AngryBeaver said:

Generally higher pumps speed (more flow) will bring better temps.

So while the water leaving actually is hotter than a fast flow... the temps on the components goes up because of less thermal transfer.

 

This is what I'm looking to test, how low flow can a loop go before the component temps start to rise due to a lack of volume of fluid flowing through the system to take away the heat. 

giving the fluid more time in the block to take in more temperature (entropy transferred to the liquid) isn't a bad thing if there is still enough flow for the amount of watts of heat energy being exhausted by the components.

Faster fluid flow can provide a cooler volume of fluid into block until the thermal mass of the fluid in the loop is saturated, but eventually the loop will saturate regardless of how fast the fluid moves at which point the thermal transfer is at it's least efficient in the blocks but most efficient in the radiators. That's why the loop is saturated equal energy in and out causing no more change in temperature. 

 

so I think to be more clear; the fluid flow rate is the variable to test, volume of the loop is constant, after saturation the thermal transfer rate "should" be constant but is based on how fast the fluid transfers heat between the blocks and radiators which is based on flow rate of the fluid, thermal energy in and out are constants based on the power consumption of the CPU and the radiator's heat transfer to the air after loop saturation. 

 

I guess the filtered question becomes how over provisioned is the flow rate of a pump running at full vs running at lower rpms have the manufacturers made them? how low can you make a loop's flow rate before the block can't transfer heat fast enough and makes the components heat up and is max speed the best setting when it puts the most wear and tear on the pump?

this also effects air bubbles forming and moving in the system, as well as increased turbulence which causes erosion in components,

[NattyKathy]

I volunteer to run a test. Don't have the energy to tear down my system and run it on an open bench but I'll try to do things methodical as I can. Will post results later tonight

[AngryBeaver]

1 hour ago, GhostRoadieBL said:

 

This is what I'm looking to test, how low flow can a loop go before the component temps start to rise due to a lack of volume of fluid flowing through the system to take away the heat. 

giving the fluid more time in the block to take in more temperature (entropy transferred to the liquid) isn't a bad thing if there is still enough flow for the amount of watts of heat energy being exhausted by the components.

Faster fluid flow can provide a cooler volume of fluid into block until the thermal mass of the fluid in the loop is saturated, but eventually the loop will saturate regardless of how fast the fluid moves at which point the thermal transfer is at it's least efficient in the blocks but most efficient in the radiators. That's why the loop is saturated equal energy in and out causing no more change in temperature. 

 

so I think to be more clear; the fluid flow rate is the variable to test, volume of the loop is constant, after saturation the thermal transfer rate "should" be constant but is based on how fast the fluid transfers heat between the blocks and radiators which is based on flow rate of the fluid, thermal energy in and out are constants based on the power consumption of the CPU and the radiator's heat transfer to the air after loop saturation. 

 

I guess the filtered qustion becomes how over provisioned is the flow rate of a pump running at full vs running at lower rpms have the manufacturers made them? how low can you make a loop's flow rate before the block can't transfer heat fast enough and makes the components heat up and is max speed the best setting when it puts the most wear and tear on the pump?

this also effects air bubbles forming and moving in the system, as well as increased turbulence which causes erosion in components,

Yes the loop will eventually hit equilibrium. Though that comes down to water temp vs ambient, air flow, and surface area. 

[GhostRoadieBL]

1 hour ago, Natty Ice said:

I volunteer to run a test. Don't have the energy to tear down my system and run it on an open bench but I'll try to do things methodical as I can. Will post results later tonight

Thanks for doing that!

I don't think open test bench is really needed as the testing is really only after loop saturation. 

 

Specifically I'm looking for:

- time to loop saturation at max pump speed as the control

- after loop saturation temp is reached reducing the pump speed until the CPU package temp rises to find the minimum pump speed to maintain the cooling performance

- running the pump at that speed to repeat the time to loop saturation to see if flow rate makes a noticeable difference

 

29 minutes ago, AngryBeaver said:

Yes the loop will eventually hit equilibrium. Though that comes down to water temp vs ambient, air flow, and surface area. 

these are the fixed variables though, water temp vs ambient is a constant at saturation, airflow is constant assuming the fans are at a fixed speed for the saturation temperature, and surface area based on the parts. 

The loop will always reach equilibrium but the time it takes to get there "should" be based on flow rate assuming a perfectly efficient transfer of heat energy through the loop. No system is tuned to perfect efficiency and manufactures of pumps need to account for everything from a single 120mm rad to a multiple 480mm rad system with reduced flow rate from the components' resistance to flow. This causes higher pressure in the system to try to make up for the resistance leading to more erosion and wear on the components. If you can reduce the flow rate, thus reducing the kinetic friction, thus lowering the erosion and wear on the components while keeping the thermal exchange happening as efficiently as with higher flow rates your components will need less maintenance and keep their nickel plating longer.

[NattyKathy]

40 minutes ago, GhostRoadieBL said:

Thanks for doing that!

I don't think open test bench is really needed as the testing is really only after loop saturation. 

 

Specifically I'm looking for:

- time to loop saturation at max pump speed as the control

- after loop saturation temp is reached reducing the pump speed until the CPU package temp rises to find the minimum pump speed to maintain the cooling performance

- running the pump at that speed to repeat the time to loop saturation to see if flow rate makes a noticeable difference

 

these are the fixed variables though, water temp vs ambient is a constant at saturation, airflow is constant assuming the fans are at a fixed speed for the saturation temperature, and surface area based on the parts. 

The loop will always reach equilibrium but the time it takes to get there "should" be based on flow rate assuming a perfectly efficient transfer of heat energy through the loop. No system is tuned to perfect efficiency and manufactures of pumps need to account for everything from a single 120mm rad to a multiple 480mm rad system with reduced flow rate from the components' resistance to flow. This causes higher pressure in the system to try to make up for the resistance leading to more erosion and wear on the components. If you can reduce the flow rate, thus reducing the kinetic friction, thus lowering the erosion and wear on the components while keeping the thermal exchange happening as efficiently as with higher flow rates your components will need less maintenance and keep their nickel plating longer.

I'm deviating a bit from the test you laid out but I think the results will still provide insight. The AIO I'm testing with (Corsair H115i Pro RGB) only offers 3 distinct steps for pump speed with iCue so I'm running separate tests from idle/low equilibrium to load/high equilibrium at all 3 pump speeds. Additionally, since I'm testing using the AIO on my GPU (my CPU is also AIO cooled but accurately and precisely monitoring temperatures on this FX chip is... challenging) I'm also running all 3 pump speeds with the GPU power limit at 165W, and again at 330W to see how power dissipation affects things. Results coming soon!

[AngryBeaver]

I'll drop this here. De8auer did a good video on it. Jay has another one on it and I think Linus even did one. All of the results are pretty much the same.

 

 

[NattyKathy]

That DerBauer video answers the question pretty definitively. Nonetheless, I said I'd run a test so here are my results.

 

Testing was done on the system in my sig, using the Vega 64 and H115i Pro AIO. iCue offers three pump RPM settings for this unit: 1100 RPM, 2100 RPM, and 2800 RPM; I tested each with the GPU power limit set to 165W, and again with power limit at 330W. The AIO's Radiator is mounted to the side of the case as an exhaust so I was able to remove the side panel with rad attached and place it a small distance away to minimize effects of heat in the rest of the system. AIO fans were set to their maximum speed of 1200 RPM. The system was given ample time to cool down between tests, until all temperatures were within 1*C of the idle temps recorded before the first run. I couldn't find my thermometer to check ambient temps but it felt like about 65*F/18*C. For a load I used Furmark at 640x480 with no AA.

 

And, yeah, regarding OP's initial question, I did not notice any substantial difference in the length of time it took for the closed loop to reach equilibrium. In all my tests the temperature delta above idle reached about 90% of it's final value within about 2-3 minutes of the start of the run, and by the 10 minute mark it had stabilized to the point that no rise was recorded for 60+ seconds. Since the temp rise follows a more or less logarithmic curve I struggled to pinpoint an exact "ok it's done heating up" moment but honestly it all looked the same to me regardless of pump speed and power dissipation.

 

As for the temperature data, I'll let that speak for itself

 

GPU Core+HBM 165W
Pump Quiet (1100 RPM)
    starting temp- 24*C core / 25*C hotspot / 25.5*C liquid
    ending temp- 44*C core / 59*C hotspot / 40.4*C liquid
Pump Balanced (2100 RPM)
    starting temp- 24*C core / 25*C hotspot / 25.9*C liquid
    ending temp- 39*C core / 53*C hotspot / 36.5*C liquid
Pump Extreme (2800 RPM)
    starting temp- 24*C core / 25*C hotspot / 26.5*C liquid
    ending temp- 39*C core / 52*C hotspot / 36.7*C liquid

GPU Core+HBM 330W
Pump Quiet (1100 RPM)
    starting temp- 24*C core / 25*C hotspot / 25.9*C liquid
    ending temp- 62*C core / 92*C hotspot / 51.2*C liquid
Pump Balanced (2100 RPM)
    starting temp- 24*C core / 25*C hotspot / 25.6*C liquid
    ending temp- 53*C core / 82*C hotspot / 46.0*C liquid
Pump Extreme (2800 RPM)
    starting temp- 24*C core / 25*C hotspot / 26.5*C liquid
    ending temp- 53*C core / 82*C hotspot / 46.4*C liquid

 

Keep those pumps a-spinnin, folks. Flow rate is important.

 

One thing that I am curious about, is what would happen if we tried to have it "both ways", ie have high flow rate through the waterblock and low flow rate through the radiators in the same loop. Theoretically such a scenario would be achievable by using multiple radiators and connecting them in parallel instead of series; the pump and CPU/GPU would still see the full flow but the rads would see 1/n flow and 1*n dwell time (unless my understanding of fluid dynamics is totally borked, which it probably is)  I bet there's a YouTube video for that too.

[For Science!]

Yes, for AIOs, I always recommend keeping them at constant max speed, they do not have much flow to begin with, and slowing them down from there really starts getting into molasses territory. The same cannot be said for a custom loop with a D5 or DDC, and so even very low flow rates would exceed what an AIO pump can achieve.

[AngryBeaver]

35 minutes ago, Natty Ice said:

That DerBauer video answers the question pretty definitively. Nonetheless, I said I'd run a test so here are my results.

 

Testing was done on the system in my sig, using the Vega 64 and H115i Pro AIO. iCue offers three pump RPM settings for this unit: 1100 RPM, 2100 RPM, and 2800 RPM; I tested each with the GPU power limit set to 165W, and again with power limit at 330W. The AIO's Radiator is mounted to the side of the case as an exhaust so I was able to remove the side panel with rad attached and place it a small distance away to minimize effects of heat in the rest of the system. AIO fans were set to their maximum speed of 1200 RPM. The system was given ample time to cool down between tests, until all temperatures were within 1*C of the idle temps recorded before the first run. I couldn't find my thermometer to check ambient temps but it felt like about 65*F/18*C. For a load I used Furmark at 640x480 with no AA.

 

And, yeah, regarding OP's initial question, I did not notice any substantial difference in the length of time it took for the closed loop to reach equilibrium. In all my tests the temperature delta above idle reached about 90% of it's final value within about 2-3 minutes of the start of the run, and by the 10 minute mark it had stabilized to the point that no rise was recorded for 60+ seconds. Since the temp rise follows a more or less logarithmic curve I struggled to pinpoint an exact "ok it's done heating up" moment but honestly it all looked the same to me regardless of pump speed and power dissipation.

 

As for the temperature data, I'll let that speak for itself

 

GPU Core+HBM 165W
Pump Quiet (1100 RPM)
    starting temp- 24*C core / 25*C hotspot / 25.5*C liquid
    ending temp- 44*C core / 59*C hotspot / 40.4*C liquid
Pump Balanced (2100 RPM)
    starting temp- 24*C core / 25*C hotspot / 25.9*C liquid
    ending temp- 39*C core / 53*C hotspot / 36.5*C liquid
Pump Extreme (2800 RPM)
    starting temp- 24*C core / 25*C hotspot / 26.5*C liquid
    ending temp- 39*C core / 52*C hotspot / 36.7*C liquid

GPU Core+HBM 330W
Pump Quiet (1100 RPM)
    starting temp- 24*C core / 25*C hotspot / 25.9*C liquid
    ending temp- 62*C core / 92*C hotspot / 51.2*C liquid
Pump Balanced (2100 RPM)
    starting temp- 24*C core / 25*C hotspot / 25.6*C liquid
    ending temp- 53*C core / 82*C hotspot / 46.0*C liquid
Pump Extreme (2800 RPM)
    starting temp- 24*C core / 25*C hotspot / 26.5*C liquid
    ending temp- 53*C core / 82*C hotspot / 46.4*C liquid

 

Keep those pumps a-spinnin, folks. Flow rate is important.

 

One thing that I am curious about, is what would happen if we tried to have it "both ways", ie have high flow rate through the waterblock and low flow rate through the radiators in the same loop. Theoretically such a scenario would be achievable by using multiple radiators and connecting them in parallel instead of series; the pump and CPU/GPU would still see the full flow but the rads would see 1/n flow and 1*n dwell time (unless my understanding of fluid dynamics is totally borked, which it probably is)  I bet there's a YouTube video for that too.

The loop flow rate will always be = to most constrained point unless you run a splitter, but that will mean water tends to take the easier path.

 

Also while your AIO test does basically match up with the video I posted... there is a big difference in aio flow rate which maxes out at around .2-.3 gpm and a custom loop where gpm can be over 2+.

 

 

[LogicalDrm]

-> Moved to Custom Loop and Exotic Cooling

[GhostRoadieBL]

6 hours ago, Natty Ice said:

Keep those pumps a-spinnin, folks. Flow rate is important

Thanks for testing that, 

It really looks like 2100 and 2800 would be within margin of error. So pump balanced would reduce the flowrate and reduce wear on the components. 

I'll dig into derbauer's video as well when I get home today. 

 

[Mick Naughty]

I know my temps are directly related to pump speed. If not, I’d just run it so I couldn’t hear it. But the lower it goes, the faster my fans need to spin. Which is why I find that comfortable medium. 
 

[GhostRoadieBL]

12 minutes ago, Mick Naughty said:

I know my temps are directly related to pump speed. If not, I’d just run it so I couldn’t hear it. But the lower it goes, the faster my fans need to spin. Which is why I find that comfortable medium. 
 

That's not really what was being tested, it's after the loop is saturated which means the fluid is carrying the most thermal energy it can while leaving the block and the radiator/s are exhausting the most heat to the air for the loop. 

Once at that situation if you were to reduce the pump speed you can find the lowest 'stable temperature' flow rate for the system. 

 

The theory stands that at that flow rate the time it takes a loop to reach saturation is the same or near the same as when the pump is at its highest flow rate causing the most wear and erosion on components. 

[Mick Naughty]

12 minutes ago, GhostRoadieBL said:

That's not really what was being tested, it's after the loop is saturated which means the fluid is carrying the most thermal energy it can while leaving the block and the radiator/s are exhausting the most heat to the air for the loop. 

Once at that situation if you were to reduce the pump speed you can find the lowest 'stable temperature' flow rate for the system. 

 

The theory stands that at that flow rate the time it takes a loop to reach saturation is the same or near the same as when the pump is at its highest flow rate causing the most wear and erosion on components. 

Not any different than what I just said. I can game for hours, raise my pump speed and temps will drop. No debate. 

[NattyKathy]

7 hours ago, For Science! said:

Yes, for AIOs, I always recommend keeping them at constant max speed, they do not have much flow to begin with, and slowing them down from there really starts getting into molasses territory. The same cannot be said for a custom loop with a D5 or DDC, and so even very low flow rates would exceed what an AIO pump can achieve.

 

7 hours ago, AngryBeaver said:

The loop flow rate will always be = to most constrained point unless you run a splitter, but that will mean water tends to take the easier path.

 

Also while your AIO test does basically match up with the video I posted... there is a big difference in aio flow rate which maxes out at around .2-.3 gpm and a custom loop where gpm can be over 2+.

 

 

 

Makes sense, with the pumps on AIOs being so tiny most of the time and downright miniscule sometimes.

 

Which is why I'm a little confused about the lack of improvement I saw going from 2100 RPM - 2800 RPM. I thought coldplate limits maybe (something I believe I've run into when cooling dense Ryzen chips with an AIO), but then I'd expect different results from different heat dissipation levels. I expected the gains from 2100-2800 RPM to be low but I didn't think they'd be zero

 

Anyways, clearly more dimensions to the AIO side of @GhostRoadieBL's question

[GhostRoadieBL]

39 minutes ago, Natty Ice said:

 

 

Makes sense, with the pumps on AIOs being so tiny most of the time and downright miniscule sometimes.

 

Which is why I'm a little confused about the lack of improvement I saw going from 2100 RPM - 2800 RPM. I thought coldplate limits maybe (something I believe I've run into when cooling dense Ryzen chips with an AIO), but then I'd expect different results from different heat dissipation levels. I expected the gains from 2100-2800 RPM to be low but I didn't think they'd be zero

I think there is a slight overbuild in the cooler's design, possibly to account for manufacturing flaws (ie cold plate thickness variations or pump impellor molding tolerances) 

the variable we can't test for is pump output pressure on the system since none of these pumps are positive displacement there will be some slippage in the fluid flow as well as pressure increase from the resistance of the cold plates (which should be well designed and have near zero resistance but that is not guaranteed) 

 

deep diving into this is complicated, I learned a lot from some of my thermodynamics and engineering courses but there's definitely smarter people than me designing these coolers and pumps. Ideally a user should be able to lower the flow rate of their system to a point where at max heat load the pump is at it's minimum speed to move the watts of heat and at lower loads and idle they should be able to almost stop the flow entirely if the thermal transfer of the volume of fluid to other molecules is good enough. considering there are D5 PWM pumps on the market which can easily slow to near trickles of fluid it would be interesting seeing a system running with no flowing liquid then ramp up as the heat load increases. 

 

[GhostRoadieBL]

DerBauer's video was extremely close to the test I was looking to run so I can extrapolate most of the conclusion from his data. 

He was specifically looking for "does a lower pump speed reduce the temps" which isn't what I was looking for at all, there is nothing in the physical sciences of closed system cooling which supports that myth. 

However his data does support the theory of not needing 100% pump speed to maintain equilibrium in the system, the "normal" setting at 56% flow rate and "high" setting at 100% flow rate maintains the CPU temperature while greatly reducing the volume and velocity of the water passing through the blocks. A flow rate around 60% seems to be that specific system's sweet spot when at maximum watt load while reducing the wear on the components. 

 

Thanks @Natty Ice and @AngryBeaver for helping me solve this one. After some more testing on block resistance to erosion (so many different fluids) I will have to put together a "how to tune your water cooling system" tutorial to help keep people's systems from unnecessary early wear and tear. 

[bowrilla]

39 minutes ago, GhostRoadieBL said:

DerBauer's video was extremely close to the test I was looking to run so I can extrapolate most of the conclusion from his data. 

He was specifically looking for "does a lower pump speed reduce the temps" which isn't what I was looking for at all, there is nothing in the physical sciences of closed system cooling which supports that myth. 

However his data does support the theory of not needing 100% pump speed to maintain equilibrium in the system, the "normal" setting at 56% flow rate and "high" setting at 100% flow rate maintains the CPU temperature while greatly reducing the volume and velocity of the water passing through the blocks. A flow rate around 60% seems to be that specific system's sweet spot when at maximum watt load while reducing the wear on the components. 

 

Thanks @Natty Ice and @AngryBeaver for helping me solve this one. After some more testing on block resistance to erosion (so many different fluids) I will have to put together a "how to tune your water cooling system" tutorial to help keep people's systems from unnecessary early wear and tear. 

I just finished a test run out of curiosity since there are some weird values in that video. The flow values are ridiculously low. Was a stupid thing to start this in the evening and turned out to be an all nighter. Will go through the data tomorrow but let's say I'm surprised by the result to the point that I again have doubts about the measurements. Will get a bit of sleep and get back later.

[For Science!]

1 hour ago, bowrilla said:

I just finished a test run out of curiosity since there are some weird values in that video. The flow values are ridiculously low. Was a stupid thing to start this in the evening and turned out to be an all nighter. Will go through the data tomorrow but let's say I'm surprised by the result to the point that I again have doubts about the measurements. Will get a bit of sleep and get back later.

My new theory is that the units should be gallon based instead of litres...?
 

2 hours ago, GhostRoadieBL said:

As for time to equilibrium, I think lower flow rates will achieve steady state faster than fast flow rates, although the steady state may be at a slightly higher temp (emphasis on slightly). Radiators increase efficiency as the coolant temperature increases with respect to the ambient air. So in an extreme scenario like a car where you do have notably hot coolant going into a cool radiator, the radiator is working most efficiently. When the coolant is not so much hotter than ambient (in essence, all the time for a pc) the radiator works inefficiently until the coolant temp starts to creep up (by absorbing heat), so just like how reducing the total volume of the loop would speed up the time to steady state, I think so would lowering the flow rate.
 

see original discussion of the flow rate here:

Quote

As much as I agree with der8auer's points and claims (that 4% pump speed performs similar to 100%), I am actually unsure about his quoted flow rates. An unobstructed D5 is rated for 1500 L/h, and my own loop with 3 waterblocks and 2 restrictive radiators can achieve 1 gpm (227 L/h) at about 70-80% pump speed on a D5. So for him to only get 37 L/h at 100% makes me think that either my instrument or his flow meters are way off chart (he didn't have nearly as much restriction as I did). I used a Koolance Flowmeter (https://koolance.com/ins-fm17n-coolant-flow-meter) coupled to  the suitable multiplier https://koolance.com/files/products/manuals/manual_ins-fm17,18_d130eng.pdf

 

[GhostRoadieBL]

7 hours ago, For Science! said:

units should be gallon based instead of litres

He may be using a very restricted block for testing. It does look more likely to be in gal/h and he didn't realize because noone was really looking to verify the actual flowrate as much as the focus was on max/min pump speed. 

 

7 hours ago, For Science! said:

so just like how reducing the total volume of the loop would speed up the time to steady state, I think so would lowering the flow rate

That is true with higher pressure capable pumps which can overcome the resistance in the system. The typical impeller design doesn't have the same characteristics as industrial or automotive pump designs so it will have slippage much sooner as back pressure increases halting the increase in flowrate. That's why an apparent 44% increase in flowrate comes from ~150% increase in pumpspeed in derbauer's testing. 

[bowrilla]

8 hours ago, For Science! said:

My new theory is that the units should be gallon based instead of litres...?

The usual flow rate would be then given in GPM (Gallons/minute) which would be ~8400l/h. If it was GPH it would have been ~140l/h. Both are pretty unlikely considering that my small SPC pump with 2 blocks and 2 radiators and quite a lot of bends makes ~169l/h max at just 2400RPM.

 

So, here are my preliminary data:

 

I have some doubts about it since the difference in K over ambient scaling down the line is surprisingly small. I just changed the fluid 2 days ago and am confident, that there is no air in the block (maybe in the top rad a bit though). Also mind you: both GPU and CPU blocks are in parallel and both radiators are in parallel cutting flow rate through those components roughly (!) in half. Maybe 30min Prime95 SmallFFT was not enough heat to saturate the loop allthough water temps did not change during the last 3-5mins and before that was climbing by a tiny 0,1K every 2-3mins. The internal temp sensor is built into the Aquacomputer MPS High Flow sensor, the other one (XSPC screw plug) is closer to the pump res. Those temps are taken after the blocks and before entering the reservoir.

FlowTests.xlsx

[For Science!]

49 minutes ago, bowrilla said:

The usual flow rate would be then given in GPM (Gallons/minute) which would be ~8400l/h. If it was GPH it would have been ~140l/h. Both are pretty unlikely considering that my small SPC pump with 2 blocks and 2 radiators and quite a lot of bends makes ~169l/h max at just 2400RPM.

 

So, here are my preliminary data:

 

I have some doubts about it since the difference in K over ambient scaling down the line is surprisingly small. I just changed the fluid 2 days ago and am confident, that there is no air in the block (maybe in the top rad a bit though). Also mind you: both GPU and CPU blocks are in parallel and both radiators are in parallel cutting flow rate through those components roughly (!) in half. Maybe 30min Prime95 SmallFFT was not enough heat to saturate the loop allthough water temps did not change during the last 3-5mins and before that was climbing by a tiny 0,1K every 2-3mins. The internal temp sensor is built into the Aquacomputer MPS High Flow sensor, the other one (XSPC screw plug) is closer to the pump res. Those temps are taken after the blocks and before entering the reservoir.

FlowTests.xlsx 35.11 kB · 0 downloads

Sooooo..... the one sentence conclusion is that between 170 l/h and <50 l/h you saw a couple degrees maximum increase in tDie and coolant?

 

[bowrilla]

11 minutes ago, For Science! said:

Sooooo..... the one sentence conclusion is that between 170 l/h and <50 l/h you saw a couple degrees maximum increase in tDie and coolant?

 

Yes, 6K on Tdie and about 2K coolant (100% fan speed). Which seems ridiculously low especially considering that the blocks are in parallel cutting flow rate roughly in half.

 

This is why I have still some doubts about the results. I might need to do it again and keep Prime95 on for longer and adding the GPU with FurMark for more heat.

 

Edit: I expected to have something more in the 10-15K range on Tdie.