A question about flow rate.

[MedievalMatt]

I get this has probably been asked and discussed a bajillion times at this point.  But im curious about flow rate.  Specifically how do I determine if I need more and should add a second pump.

 

So I have 5 total rads, 2 480s, 2 360s, and a single 120.  I also have a flow meter and the GPU (3080ti Hydro copper) and CPU (r9 5900x) are in the loop.

 

With the pump at 0% (0% PWM defaults to 2000ish RPM speed) the flow rate is reported as 77ish lph.  With the pump at 100% (5000ish RPM) flow rate reports at 212ish lph.

So far, with a temp setpoint of the fluid at 40C, the only difference I can measure between 0% and 100% pump speed aside from noise is that the temperature of the fluid in the loop as a whole is more consistent.  That is to say at 0% my pump temp will be 40C and the GPU and CPU fluid outlets will be in the 43-45C range.  With the pump at 100% the whole loop will be roughly 40c.  This is at full load.  The CPU reports its temp tops out at 70c, and the GPU is chillin at 65C under furmark/cinebench.

 

From what Ive read in other places people target 1gpm (227lph) or they say to look at temps.  personally I dont see the benefit of a homogenous fluid temp as a plus.  In either case the fluid entering the "hot" components ends up being the same temp, so its extra noise for no reason in my mind.  There is also the fact that I cant approach 1gpm unless I add a second pump.

Am I missing some nuance of flow rate, or does it not matter as much at all?  Granted if my pump was somehow loud at idle I would up the speed to find a quiet spot, but my monitor fan is louder than my entire system at idle.... so I just let it idle.

Looking at the graphs from the latest video from LTT about waterblocks got me curious hence the research and question.

[thrasher_565]

1 hour ago, MedievalMatt said:

I get this has probably been asked and discussed a bajillion times at this point.  But im curious about flow rate.  Specifically how do I determine if I need more and should add a second pump.

 

So I have 5 total rads, 2 480s, 2 360s, and a single 120.  I also have a flow meter and the GPU (3080ti Hydro copper) and CPU (r9 5900x) are in the loop.

 

With the pump at 0% (0% PWM defaults to 2000ish RPM speed) the flow rate is reported as 77ish lph.  With the pump at 100% (5000ish RPM) flow rate reports at 212ish lph.

So far, with a temp setpoint of the fluid at 40C, the only difference I can measure between 0% and 100% pump speed aside from noise is that the temperature of the fluid in the loop as a whole is more consistent.  That is to say at 0% my pump temp will be 40C and the GPU and CPU fluid outlets will be in the 43-45C range.  With the pump at 100% the whole loop will be roughly 40c.  This is at full load.  The CPU reports its temp tops out at 70c, and the GPU is chillin at 65C under furmark/cinebench.

 

From what Ive read in other places people target 1gpm (227lph) or they say to look at temps.  personally I dont see the benefit of a homogenous fluid temp as a plus.  In either case the fluid entering the "hot" components ends up being the same temp, so its extra noise for no reason in my mind.  There is also the fact that I cant approach 1gpm unless I add a second pump.

Am I missing some nuance of flow rate, or does it not matter as much at all?  Granted if my pump was somehow loud at idle I would up the speed to find a quiet spot, but my monitor fan is louder than my entire system at idle.... so I just let it idle.

Looking at the graphs from the latest video from LTT about waterblocks got me curious hence the research and question.

so the faster the flow rate the lower the temp. to get that you will need a more power full pump or more pumps. but if your oc cant take advantage of the added pump and lower temps there no added benefit other then looking cool.

 

but with new stuff oc ing the cpu is some what pointless over there default oc and same with the gpu form what i heard, thow under voting can have a benefit when it come to cost vs nose vs power.  if you want a faster cpu / gpu you jsut buy the nexed tier up simple as that.

 

back then oc gave you alot of performance for cheap in some case but with so many gpus skews there no point.

 

thing to note is there are hi flow rate parts and there are less flow rate parts but selecting and build for each could have some in pack to the temps.

 

for every 1gpm is about 1c temp drop.

 

 

 

Edited August 9, 2022 by thrasher_565

[LIGISTX]

1 hour ago, MedievalMatt said:

I get this has probably been asked and discussed a bajillion times at this point.  But im curious about flow rate.  Specifically how do I determine if I need more and should add a second pump.

 

So I have 5 total rads, 2 480s, 2 360s, and a single 120.  I also have a flow meter and the GPU (3080ti Hydro copper) and CPU (r9 5900x) are in the loop.

 

With the pump at 0% (0% PWM defaults to 2000ish RPM speed) the flow rate is reported as 77ish lph.  With the pump at 100% (5000ish RPM) flow rate reports at 212ish lph.

So far, with a temp setpoint of the fluid at 40C, the only difference I can measure between 0% and 100% pump speed aside from noise is that the temperature of the fluid in the loop as a whole is more consistent.  That is to say at 0% my pump temp will be 40C and the GPU and CPU fluid outlets will be in the 43-45C range.  With the pump at 100% the whole loop will be roughly 40c.  This is at full load.  The CPU reports its temp tops out at 70c, and the GPU is chillin at 65C under furmark/cinebench.

 

From what Ive read in other places people target 1gpm (227lph) or they say to look at temps.  personally I dont see the benefit of a homogenous fluid temp as a plus.  In either case the fluid entering the "hot" components ends up being the same temp, so its extra noise for no reason in my mind.  There is also the fact that I cant approach 1gpm unless I add a second pump.

Am I missing some nuance of flow rate, or does it not matter as much at all?  Granted if my pump was somehow loud at idle I would up the speed to find a quiet spot, but my monitor fan is louder than my entire system at idle.... so I just let it idle.

Looking at the graphs from the latest video from LTT about waterblocks got me curious hence the research and question.

I am a bit surprised by how warm the GPU is, my 2080 never goes over ~55 even when my room is warm, and I have a lot less rad space then you.

 

BUT, that said, more flow does not necessarily mean more cool. I typically speed my pump up a bit when under load and it reacts about how you describe, but no real change. I don't think you are missing anything, thats just how it works. Turns out.... enough flow is enough

[MedievalMatt]

28 minutes ago, thrasher_565 said:

so the faster the flow rate the lower the temp. to get that you will need a more power full pump or more pumps. but if your oc cant take advantage of the added pump and lower temps there no added benefit other then looking cool.

 

but with new stuff oc ing the cpu is some what pointless over there default oc and same with the gpu form what i heard, thow under voting can have a benefit when it come to cost vs nose vs power.  if you want a faster cpu / gpu you jsut buy the nexed tier up simple as that.

 

back then oc gave you alot of performance for cheap in some case but with so many gpus skews there no point.

 

thing to note is there are hi flow rate parts and there are less flow rate parts but selecting and build for each could have some in pack to the temps.

 

 

 

Perhaps you misunderstood, or im not understanding.  You say the difference is temps.  Which temps?  Lets take my GPU fluid outlet as an example in a vacuum.  This is the temp of the coolant after passing through the GPU block but before the rads.  if I run my pump at idle with full load on the GPU this will be 43C.  if I do the same but with my pump at 100% it will be 40C.  Everything else about the setup is exactly the same except flow rate.  The reported GPU temp in HWInfo, fan speed, everything is the same.  Even them temp of the coolant entering the pump, so right before getting to the GPU block, is 40c.  so im running the pump faster for a more uniform fluid temp?  it seems unecessary.

[thrasher_565]

2 minutes ago, MedievalMatt said:

Perhaps you misunderstood, or im not understanding.  You say the difference is temps.  Which temps?  Lets take my GPU fluid outlet as an example in a vacuum.  This is the temp of the coolant after passing through the GPU block but before the rads.  if I run my pump at idle with full load on the GPU this will be 43C.  if I do the same but with my pump at 100% it will be 40C.  Everything else about the setup is exactly the same except flow rate.  The reported GPU temp in HWInfo, fan speed, everything is the same.  Even them temp of the coolant entering the pump, so right before getting to the GPU block, is 40c.  so im running the pump faster for a more uniform fluid temp?  it seems unecessary.

the slower flow will mean the fluid can heat up more and take more heat with it but if your rads can take it back down to 40c then there's no point.  regardless of pump speed.  but say you add in another pump then it can be like 39c. if that enough for then nexed bit of stable oc or not i dont no.  at least that's how i under stand it...

[MedievalMatt]

2 minutes ago, LIGISTX said:

I am a bit surprised by how warm the GPU is, my 2080 never goes over ~55 even when my room is warm, and I have a lot less rad space then you.

 

BUT, that said, more flow does not necessarily mean more cool. I typically speed my pump up a bit when under load and it reacts about how you describe, but no real change. I don't think you are missing anything, thats just how it works. Turns out.... enough flow is enough

3080ti also sucks down more power than the 2080, by a lot.  The 3080ti I have is hardware limited to 450Watts, and basically will always use that under load.  IIRC the 2080 is somewhere in the 300 range.  I know I can shunt mod etc to increase that.  but im not ready for all that yet.  I also have the card overclocked so that could account for higher temps.

 

Good to know it seems Im not missing something.  Seems to me the "benefit" to running a faster pump is more uniform coolant temps, as in it will be the same temp more or less everywhere, and a slightly longer "warm up" period.  both seem like they are beneficial in very edge cases.

 

70lph would be my lower limit.  anything below about 180lph doesn't dislodge bubbles.  so for the first week or two I run the system at 100% to get all the stubborn bubbles out.  Also worth noting that i dont have any way to calibrate the flow sensor, so it could be off by some amount.  The relative performance I think is what is more important here though.

[MedievalMatt]

1 minute ago, thrasher_565 said:

the slower flow will mean the fluid can heat up more and take more heat with it but if your rads can take it back down to 40c then there's no point.  regardless of pump speed.  but say you add in another pump then it can be like 39c. if that enough for then nexed bit of stable oc or not i dont no.  at least that's how i under stand it...

ah ok I see.  I use a temp setpoint for the fans, so their "goal" is to keep the water at 40c, which under full load is 80% fan speed.  I could run the water at 30C if I really wanted to sit next to an airplane, lol.

[LIGISTX]

Just now, MedievalMatt said:

3080ti also sucks down more power than the 2080, by a lot.  The 3080ti I have is hardware limited to 450Watts, and basically will always use that under load.  IIRC the 2080 is somewhere in the 300 range.  I know I can shunt mod etc to increase that.  but im not ready for all that yet.  I also have the card overclocked so that could account for higher temps.

Mine is also OCed, and while yours does draw more wattage, you have much, much more rad space. My GPU is usually 49-50c while I’m game load, with my fans at 800 rpm. I’m surprised your care gets 20c warmer, that just seems crazy. 
 

But, it’s working, so ¯\_(ツ)_/¯. I wonder if the GPU isn’t making perfect contact with the cold plate. Are those temps normal for a waterblocked 3080ti? 

[MedievalMatt]

1 minute ago, LIGISTX said:

Mine is also OCed, and while yours does draw more wattage, you have much, much more rad space. My GPU is usually 49-50c while I’m game load, with my fans at 800 rpm. I’m surprised your care gets 20c warmer, that just seems crazy. 
 

But, it’s working, so ¯\_(ツ)_/¯. I wonder if the GPU isn’t making perfect contact with the cold plate. Are those temps normal for a waterblocked 3080ti? 

to be fair, i just checked and what i reported is the hotspot temp.  the "average" is like 10-15C lower on other parts of the silicon.  I had a 3080ti with an EK block on it which actually was way worse for temps than the hydro copper.  perhaps I botched the install, who knows.

 

I also dont really look at CPU/ GPU temps really because what happens is the card will idle at 40-50 hotspot.  then spike up to 65 immediately when there is a load and stay there.  The CPU is worse with turbo and all that going gangbusters the temp swings between 72 and 45c at idle willy nilly lol.  Its not really helpful to setup a loop like that, water temp is a much better way to setup the thermal situation.

[thrasher_565]

i run my d5 100% all the time and it should be quiet if done right. i dont under stand why people slow it down.... the non variable d5 is better then the variable one not only that it should crate less heat for the pump running 100% just like slowing down dc fans vs pwm it kills the controller.  but alot of gamers ware head phones and have it quiet at idle and 100% gaming.

[Sir Beregond]

9 hours ago, thrasher_565 said:

i run my d5 100% all the time and it should be quiet if done right. i dont under stand why people slow it down.... the non variable d5 is better then the variable one not only that it should crate less heat for the pump running 100% just like slowing down dc fans vs pwm it kills the controller.  but alot of gamers ware head phones and have it quiet at idle and 100% gaming.

D5's dump some heat into the loop, it's not much, but with multiple pumps, something to consider as you approach the realm of deminishing returns with flow rates vs temp drops. But just running one, I am likewise these days running mine at 100%. I don't hear it.

 

Yes, 1GPM is often recommended because it's not too slow where you will have a detriment to performance of the loop, and its a good spot before you hit diminishing returns on what you can get out of the loop performance wise with increased flow rate.

 

I'd be curious what your ambient to coolant temp under load delta is.

[Mick Naughty]

I keep d5 at 30% and goes to 60% high cpu temp. My 3090 sits around 40c under load if my room door is open.
I'd rather hear my fans over the pump. But do prefer more flow when I can. Id run 2 if I were you.

[For Science!]

15 hours ago, MedievalMatt said:

I get this has probably been asked and discussed a bajillion times at this point.  But im curious about flow rate.  Specifically how do I determine if I need more and should add a second pump.

 

So I have 5 total rads, 2 480s, 2 360s, and a single 120.  I also have a flow meter and the GPU (3080ti Hydro copper) and CPU (r9 5900x) are in the loop.

 

With the pump at 0% (0% PWM defaults to 2000ish RPM speed) the flow rate is reported as 77ish lph.  With the pump at 100% (5000ish RPM) flow rate reports at 212ish lph.

So far, with a temp setpoint of the fluid at 40C, the only difference I can measure between 0% and 100% pump speed aside from noise is that the temperature of the fluid in the loop as a whole is more consistent.  That is to say at 0% my pump temp will be 40C and the GPU and CPU fluid outlets will be in the 43-45C range.  With the pump at 100% the whole loop will be roughly 40c.  This is at full load.  The CPU reports its temp tops out at 70c, and the GPU is chillin at 65C under furmark/cinebench.

 

From what Ive read in other places people target 1gpm (227lph) or they say to look at temps.  personally I dont see the benefit of a homogenous fluid temp as a plus.  In either case the fluid entering the "hot" components ends up being the same temp, so its extra noise for no reason in my mind.  There is also the fact that I cant approach 1gpm unless I add a second pump.

Am I missing some nuance of flow rate, or does it not matter as much at all?  Granted if my pump was somehow loud at idle I would up the speed to find a quiet spot, but my monitor fan is louder than my entire system at idle.... so I just let it idle.

Looking at the graphs from the latest video from LTT about waterblocks got me curious hence the research and question.

No, you are not missing some nuance, flow rate beyond molasses speed simply does not have a large impact on thermals. 

[AnonymousGuy]

2 hours ago, For Science! said:

No, you are not missing some nuance, flow rate beyond molasses speed simply does not have a large impact on thermals. 

This.

 

Anything above a trickle is adequate in PC watercooling because the heatloads are low enough where you can treat the loop as a system instead of individual heat sources.  1GPM is way beyond what you really need to hit to see a difference.

 

Now...if you had a GPU that consume like 2kW by itself then you would probably want to step up your flow rate so the fluid isn't getting significantly heated within the block.  And then by increasing the flowrate the loop becomes a system again.

[MedievalMatt]

22 minutes ago, AnonymousGuy said:

This.

 

Anything above a trickle is adequate in PC watercooling because the heatloads are low enough where you can treat the loop as a system instead of individual heat sources.  1GPM is way beyond what you really need to hit to see a difference.

 

Now...if you had a GPU that consume like 2kW by itself then you would probably want to step up your flow rate so the fluid isn't getting significantly heated within the block.  And then by increasing the flowrate the loop becomes a system again.

Ok, that was my logic.  Seeing the Fluid leaving the GPU has picked up 3 degrees C of waste energy seems like a more efficient coolant flow rate than a difference of .3C.

 

Im gonna stick to my 70-90 lph (0.3 gpm) since there is literally no benefit to a higher flow.  Even under full system load the fans dont have to approach 100% speed to deal with cooling the fluid, and component temps are well within the boost threshold.

 

For reference for others who may be wondering.

 

Room temp is 25-30C (because its poorly ventilated and AC is expensive) and the fluid setpoint is 40C.  Which means my fans will spin up to keep the liquid at 40C.  I take the temp of the fluid after leaving the pump for the setpoint.  Which then immediately goes to the GPU and is heated.  So im effectively heating the water to 43C and then the rads are dissipating all of that so im back to 40C at the pump.  Fans approach 80% speed but never exceed, idle fan speed is 30%.

 

I wonder where everyone is getting this idea that flow rate has a huge effect on temps.  based on my, admittedly small sample size, testing it really doesnt.  Rad surface area and fans make the temp difference.

[Mick Naughty]

Wouldnt say its huge, just a factor. Unless we are gonna use some blanket statement and say it doesnt matter. If that were the case my temps wouldnt change, but they do. Or else I'd run my pumps at the bare minimum.

[AnonymousGuy]

10 minutes ago, MedievalMatt said:

I wonder where everyone is getting this idea that flow rate has a huge effect on temps.  based on my, admittedly small sample size, testing it really doesnt.  Rad surface area and fans make the temp difference.

Correct.  Flow rate matters a small amount to create turbulence which helps transfer heat at the very edge of the contact surface between the waterblock and the fluid.  I think it also matters a little to get better "penetration" into the full surface of the radiator vs. the water only going through a single tube. 

 

Beyond that it only matters if the water from say the GPU is 5C hotter than the inlet, which means a downstream component is now seeing water+5C inlet temperature.  By turning up the flow rate the loop becomes an average where the inlet is always close to the outlet temperatures.

 

(again though with PC watercooling the evidence seems to indicate that if you're getting basically anything above a trickle it's good enough.  People have had really clogged up blocks where it's like "how the hell is any water flowing at all?" and their temperatures aren't *that* bad)

[MedievalMatt]

14 minutes ago, AnonymousGuy said:

Correct.  Flow rate matters a small amount to create turbulence which helps transfer heat at the very edge of the contact surface between the waterblock and the fluid.  I think it also matters a little to get better "penetration" into the full surface of the radiator vs. the water only going through a single tube. 

 

Beyond that it only matters if the water from say the GPU is 5C hotter than the inlet, which means a downstream component is now seeing water+5C inlet temperature.  By turning up the flow rate the loop becomes an average where the inlet is always close to the outlet temperatures.

 

(again though with PC watercooling the evidence seems to indicate that if you're getting basically anything above a trickle it's good enough.  People have had really clogged up blocks where it's like "how the hell is any water flowing at all?" and their temperatures aren't *that* bad)

I can tell you for sure that with a low flow rate the rad surface temps are pretty consistent.  I dont have any "cold" spots to be felt.  Perhaps if the system wasnt bled properly and there is trapped air.

 

I also also see the argument for heating up the fluid to then take that significantly warmer fluid into another component, but you can solve this problem adding a rad in between.

 

Higher flow rates seem to homogenize the fluid temps.  Meaning the loop heats up and cools down more as a whole unit as opposed to having a hot side and a cold side.

 

Perhaps there is a mindset where having the fluid be the same temp throughout the entire loop at all times is the target?  but this seems like it lowers cooling efficiency.  Thermal transfer necessitates a gradient and the bigger the gradient the more efficient (to a point).  Cooling 45C water to 40C is taking out a lot more energy than cooling 45C water to 44.5C.

 

The hot side of my loop gets to 43-45C.  but by the time the water gets back to the pump its at 40C.  I feel like thats more efficient that the hot side being 40.5C and the cold side being 40C.  but maybe thats a product of having excess radiator area.

[AnonymousGuy]

42 minutes ago, MedievalMatt said:

Perhaps there is a mindset where having the fluid be the same temp throughout the entire loop at all times is the target?  but this seems like it lowers cooling efficiency.  Thermal transfer necessitates a gradient and the bigger the gradient the more efficient (to a point).  Cooling 45C water to 40C is taking out a lot more energy than cooling 45C water to 44.5C.

Yes but no.  When you have a really high flowrate the water gets back to the radiator faster.  So maybe 45 to 44.5 five times vs. 45 to 40 one time.

 

Does it really work this way in real life?  I don't know but it's easier to just think of a loop as practically being homogenous.  "how much wattage is going into this loop, how much wattage can my radiators shed at a certain water delta T above air temperature" (function of surface area and fan speed) and somewhere there's an equilibrium point that doesn't really include flow rate.

[For Science!]

12 hours ago, MedievalMatt said:

I can tell you for sure that with a low flow rate the rad surface temps are pretty consistent.  I dont have any "cold" spots to be felt.  Perhaps if the system wasnt bled properly and there is trapped air.

 

I also also see the argument for heating up the fluid to then take that significantly warmer fluid into another component, but you can solve this problem adding a rad in between.

 

Higher flow rates seem to homogenize the fluid temps.  Meaning the loop heats up and cools down more as a whole unit as opposed to having a hot side and a cold side.

 

Perhaps there is a mindset where having the fluid be the same temp throughout the entire loop at all times is the target?  but this seems like it lowers cooling efficiency.  Thermal transfer necessitates a gradient and the bigger the gradient the more efficient (to a point).  Cooling 45C water to 40C is taking out a lot more energy than cooling 45C water to 44.5C.

 

The hot side of my loop gets to 43-45C.  but by the time the water gets back to the pump its at 40C.  I feel like thats more efficient that the hot side being 40.5C and the cold side being 40C.  but maybe thats a product of having excess radiator area.

efficiency and effectiveness are two different things. Sure, the radiator will be more efficient if the water temperature is 90 degrees over ambient, but as a cooling solution, it will not be as effective if the water temperature was 20 degrees over ambient. 

 

"The loop order does not matter" is a child of having enough flow rate, where with adequate flow, where the radiator is not important. As the flow rate goes to molasses rates, then it becomes a factor once again.

[Sir Beregond]

17 hours ago, AnonymousGuy said:

This.

 

Anything above a trickle is adequate in PC watercooling because the heatloads are low enough where you can treat the loop as a system instead of individual heat sources.  1GPM is way beyond what you really need to hit to see a difference.

What? I'd say maybe 0.5gpm - 1gpm is good, yes. But below that? There is definitely lost performance to flow rates below 0.5gpm.

 

If @MedievalMatt is indeed measuring as much as a 3C difference from one part of the loop to another, that tells me the flow is woefully inadequate. With proper flow rate, the temp should be largely the same throughout the loop, or at most 1C difference.

[MedievalMatt]

On 8/10/2022 at 12:20 PM, Sir Beregond said:

What? I'd say maybe 0.5gpm - 1gpm is good, yes. But below that? There is definitely lost performance to flow rates below 0.5gpm.

 

If @MedievalMatt is indeed measuring as much as a 3C difference from one part of the loop to another, that tells me the flow is woefully inadequate. With proper flow rate, the temp should be largely the same throughout the loop, or at most 1C difference.

one part to another in this case is from the pump outlet to the GPU outlet.  so effectively the temp change of the water from before and after the GPU.  GPU gets 40c water, and 43C water comes out.  My loop order then has a couple rads before the water goes to the CPU, and I dont have a sensor before the CPU, but the flow meter does sense temp and is on the CPU outlet and is also reading 42-43C under load.  So we can safely assert that the rads between the CPU and GPU are doing some cooling work.  the bulk of my cooling though occurs after the fluid is through the hot components, my 2 480mms and a single 360mm, and then goes to the pump/res.

Im curious to know what the benefit is to a homogenous water temp.  I havent noticed any real world performance differences between .8(ish) gpm and .3gpm and surely thats what should matter.

 

At the end of the day the water going into my GPU is 40C no matter the flow rate.

[MedievalMatt]

On 8/10/2022 at 8:09 AM, For Science! said:

efficiency and effectiveness are two different things. Sure, the radiator will be more efficient if the water temperature is 90 degrees over ambient, but as a cooling solution, it will not be as effective if the water temperature was 20 degrees over ambient. 

 

"The loop order does not matter" is a child of having enough flow rate, where with adequate flow, where the radiator is not important. As the flow rate goes to molasses rates, then it becomes a factor once again.

I hear you.  But if my GPU hotspot averages 65C with my flow rate at .3gm and only goes down to 63.5C with my flow rate at .8gpm.  Either way there is no thermal throttling taking place.  and that's a water temp difference of 40.5C GPU outlet or 43C outlet.

 

Logically, at least on my setup, getting a 2 degree drop in temps when temps aren't a critical performance limitation for almost triple the pump speed doesn't seem like a worthwhile trade off.  It doesnt bother me that the water is warmer after leaving the GPU by 3 degrees, thats how its supposed to work.

 

Im wondering if I have so much rad space im into the realm of "it doesnt matter anymore" with regard to flow rate.

[thrasher_565]

2 hours ago, MedievalMatt said:

I hear you.  But if my GPU hotspot averages 65C with my flow rate at .3gm and only goes down to 63.5C with my flow rate at .8gpm.  Either way there is no thermal throttling taking place.  and that's a water temp difference of 40.5C GPU outlet or 43C outlet.

 

Logically, at least on my setup, getting a 2 degree drop in temps when temps aren't a critical performance limitation for almost triple the pump speed doesn't seem like a worthwhile trade off.  It doesnt bother me that the water is warmer after leaving the GPU by 3 degrees, thats how its supposed to work.

 

Im wondering if I have so much rad space im into the realm of "it doesnt matter anymore" with regard to flow rate.

its just an balancing act each thing adds resistance to the loop, ya you can buy less resistant parts but you first would need to no what you have. if your close to the nexed gpm then having the pump a bit faster or a part less restrictive to get there but if not then no point. getting lower then 40c is going to be vary hard and costy... back in the days people did 13v there pumps... but we also did things because that's what we thought you had to do. a system had 2x 120mm thick rads at times with push pull gt fans... that's just what exstream cooling was then. it all comes down to cost and what temps you want at what nose level. short answer it dosent really mater. 

 

all you can do it test things add parts in take em out really. more rads dose = more resistance even thow it incress the cooling...

 

another thing is sensors might not be the most accrete thing so that's probably going to be the part that thows a ranch in things... gn has found this out over the years and has let off a bit on it.

 

back then a hi flow was king so having like 7gpm was ideal but then im guessing jet plates became a thing and slowed things down..

 

now adays 1gpm is good enough

[MedievalMatt]

1 hour ago, thrasher_565 said:

its just an balancing act each thing adds resistance to the loop, ya you can buy less resistant parts but you first would need to no what you have. if your close to the nexed gpm then having the pump a bit faster or a part less restrictive to get there but if not then no point. getting lower then 40c is going to be vary hard and costy... back in the days people did 13v there pumps... but we also did things because that's what we thought you had to do. a system had 2x 120mm thick rads at times with push pull gt fans... that's just what exstream cooling was then. it all comes down to cost and what temps you want at what nose level. short answer it dosent really mater. 

 

all you can do it test things add parts in take em out really. more rads dose = more resistance even thow it incress the cooling...

 

another thing is sensors might not be the most accrete thing so that's probably going to be the part that thows a ranch in things... gn has found this out over the years and has let off a bit on it.

 

back then a hi flow was king so having like 7gpm was ideal but then im guessing jet plates became a thing and slowed things down..

 

now adays 1gpm is good enough

Sensor accuracy only matters if you want to compare across multiple systems, or if there is a specific problem related to thermals.  the accuracy in the case only matters in the relative.  So im reading 40C and talking about increases of 3C.  If its 35C in reality and increasing 3C we are effectively talking about the same delta (ya I get that there is not a linear scale in terms of energy).  If I wanted to compare my temps to yours then yes, accuracy matters in terms of the real temperature.  Its the same thing with flow rate, my meter might be off.  But its going to be off by that same amount all the time and so isnt variable.  But the flow rate accuracy also doesnt matter much because the discussion is "does more flow rate mean more better?".

 

What im gathering from this thread is that it depends a lot on the other cooling/heating components in the loop.  There is a give and take to rad area and flow rate/pump speed and there is a spectrum with an ideal curve.  What im reporting is that I dont see a CPU/GPU temp difference, and as a whole a loop water temperature difference, with 0.3 gpm and 0.8gpm(0.8 is the most my pump can do as read by my meter).

 

So working the pump harder has no real world benefit.  If I want lower temps the way to do that in my system would be to increase fan speed.  Doing the math though I have 1800 Square milimeters of total radiator surface area.  I suppose if you had less area that it may matter more your flow be higher, but it would also be higher anyways because there is less restriction.

 

Id be interested to see what other folks are running radiator area size and what their temp deltas are between full load and idle (ish, windows defender is doing some stuff so the CPU is turboing to 4.5GHz on a couple cores which is why its 70C).  For reference Ive attached a pick of my monitoring suite at idle.  The meters with the orange segment are air temp, and the system is in a corner and there is some poor air circulation there so a couple of the intake temps get warm.  The ones labelled side and bottom intake are more or less actual ambient air.