A question about flow rate.

[Luggage]

You’d want to lock your fan speed and ideally your cpu/gpu boost while testing flow rate impact.

With the heaviest load you can muster, like furmark and linx on all cores but one.

And the difference you are looking for is the CPU/GPU temps not the water temp.

 

The saying is not “you need 1gpm”, it’s “you will see improved cooling performance, non-linear, up to about 1gpm. Over that is usually not worth it even for hcoc.”

 

[MedievalMatt]

23 minutes ago, Luggage said:

You’d want to lock your fan speed and ideally your cpu/gpu boost while testing flow rate impact.

With the heaviest load you can muster, like furmark and linx on all cores but one.

And the difference you are looking for is the CPU/GPU temps not the water temp.

 

The saying is not “you need 1gpm”, it’s “you will see improved cooling performance, non-linear, up to about 1gpm. Over that is usually not worth it even for hcoc.”

 

What im saying is that I dont see a difference.

 

Also why emphasize CPU/GPU temps.  In my experience water temps are a far better way to run a water loop.  My GPU at idle on the hotspot is 50C (once the water is up to temp) and under load gets to 65C.  It does that at low flow, or high flow.  It does that at 40C water temp, or 30C water temp.  Same with the CPU.  chills between 40 and 50C at idle and ramps up to 70-80C depending on the load (again, hotspot)  and doesnt really change except for the differences in fluid temp.  So if my GPU under load is 65C, and the fluid is at 40C, and I set the fluid setpoint to 45C.  the GPU goes up 5C, but so does the fluid.

 

Its far more effective to setup a water loop using fluid temps for fans etc. because thats what you are actively trying to cool at that point.  You are taking heat from the components and dumping it into the water, but now you have hot water.  you are cooling the water when you water cool.  the water is cooling your components.

[Luggage]

Because now you are testing your cooling, not flow rate. If you set fans with ha 40C target of course you're going to get just about that, otherwise the fan curve isn't working.

[MedievalMatt]

16 minutes ago, Luggage said:

Because now you are testing your cooling, not flow rate. If you set fans with ha 40C target of course you're going to get just about that, otherwise the fan curve isn't working.

But if what everyone claims about higher flow rates lowering temps it shouldnt matter about fan curves.  All that would happen in that event would be that turning the pump speed up (increasing flow) would mean the fans would need to spin slower.  If my fans spin at 80% under load with low flow.  When i raise the flow rate i would, allegedly, see a decrease in temps which would coorelate to a decrease in full load fan speed.  But thats not what I see.

 

The way the data works out the only difference in high vs. low flow I can see in testing is that the various fluid temp sensors I have all read more or less the same temp at high flow, and they read differing temps to varying degrees at low flow.  Everything else remains the same, CPU/GPU temps, fan speed, water temp (as a whole) all the same.

 

What you can infer from that is that even if I locked off my fan speed at 100%, there would likely not be a difference.

 

I can hit my system with a full load at low flow, wait for it to settle, and then ramp the pump speed up and the only change after 20 minutes is that all the temp sensors read the same (ish) now.  As opposed to having fluid jump 3-5 degrees while transiting a CPU/GPU block.

 

If raising flow rate lowered temps, id see my fan speed go down to maintain 40C, and it doesnt.

[Luggage]

166 lpm down to 44 lpm

~ 5C diff on GPU, instantly.

https://imgur.com/a/PLZlO1b

 

[MedievalMatt]

57 minutes ago, Luggage said:

166 lpm down to 44 lpm

~ 5C diff on GPU, instantly.

https://imgur.com/a/PLZlO1b

 

To make you a little happier I capps the fans at 100% and waited for things to settle.  One of the difference between high and low flow is the GPU temps are lower at high flow.  But as you can very obviously see this is because the water temp dropped by the same amount.  Aside from that the fluid temp overall is more consistent in the loop on high flow.

 

But as you can see there is no clear benefit to high flow on my system.  Can you explain why? 

[Luggage]

What gpu and cpu load? Did you monitor boost? Gaining GPU temp is good. Depending on workload cpu temp might not change because it's boosting higher or lower instead of changing temps, with PBO that is, unless you run manual oc.

[Sir Beregond]

On 8/10/2022 at 6:09 AM, For Science! said:

where with adequate flow, where the radiator is not important.

This statement is misleading. Radiator is always important. Regardless of flow rate, radiators are still the medium of thermal exchange. That's like saying you don't need the aluminum fin stack on an air cooler if airflow is adequate.

 

As for the OP, what is your delta vs ambient at idle and load? Personally coolant equalizing at 40-43C sounds terrible, but maybe you have a higher ambient than I do. My ambient is usually 21-23C. At idle I am usually 25-26C on coolant temp and at load that depends on my fan speed, but no more than a 10-15C delta over ambient. Typically no more than 12C delta. Its not the best loop in the world, but certainly does the job. Its generally considered extremely good if that delta is kept under 10C, but 10-15C is A-OK too. More than that and there are definitely parts of the loop whether flow-rate, design, etc. to be improved on.

 

 

[For Science!]

21 minutes ago, Sir Beregond said:

This statement is misleading. Radiator is always important. Regardless of flow rate, radiators are still the medium of thermal exchange. That's like saying you don't need the aluminum fin stack on an air cooler if airflow is adequate.

Please have a re-read of the sentence, what doesn't matter is where the radiator is, its location. That is not like saying you don't need a radiator.

[MedievalMatt]

7 hours ago, Luggage said:

What gpu and cpu load? Did you monitor boost? Gaining GPU temp is good. Depending on workload cpu temp might not change because it's boosting higher or lower instead of changing temps, with PBO that is, unless you run manual oc.

All core load is 4GHz, and you can see in the screenshots the CPU and GPU are both at 100%.  I can promise you core clocks and things did not change.

[MedievalMatt]

1 hour ago, Sir Beregond said:

This statement is misleading. Radiator is always important. Regardless of flow rate, radiators are still the medium of thermal exchange. That's like saying you don't need the aluminum fin stack on an air cooler if airflow is adequate.

 

As for the OP, what is your delta vs ambient at idle and load? Personally coolant equalizing at 40-43C sounds terrible, but maybe you have a higher ambient than I do. My ambient is usually 21-23C. At idle I am usually 25-26C on coolant temp and at load that depends on my fan speed, but no more than a 10-15C delta over ambient. Typically no more than 12C delta. Its not the best loop in the world, but certainly does the job. Its generally considered extremely good if that delta is kept under 10C, but 10-15C is A-OK too. More than that and there are definitely parts of the loop whether flow-rate, design, etc. to be improved on.

 

 

You can see in my screenshots, but ambient is 25-30C (its in a corner of the room and a side of the case gets crap airflow).  the coolant temp I use for the setpoint (of 40C) is at the pump, and I only see 43C after the GPU, and again after the CPU.

 

What the setpoints means is that my fans are off until the coolant reaches 40C, and then the goal is for that to just be the fluid temp until the end of time.  Based on my testing that gives me a good balance of silence (especially if I only need my computer for a few minutes, warmup takes 30 minutes) and and ideal delta for heat transfer in my room.  I can drop my idle fluid temps to 30 or below C if i run my fans harder, but i like the peace and quiet.  It becomes barely perceptible while gaming, which was my goal.

[Sir Beregond]

2 hours ago, For Science! said:

Please have a re-read of the sentence, what doesn't matter is where the radiator is, its location. That is not like saying you don't need a radiator.

Gotcha. Agree with that wholeheartedly. With adequate flow, loop order does not matter.

[MedievalMatt]

16 minutes ago, Sir Beregond said:

Gotcha. Agree with that wholeheartedly. With adequate flow, loop order does not matter.

I wonder if that perhaps is the idea behind the 1gpm (227ish lph) mentality.  Where if your flow is that high rad placement location is irrelevant at that point.

 

I can say that with my flow rate it matters where the rads are in the loop, but I have so many and with so much area, that it effectively doesnt.

[Sir Beregond]

13 minutes ago, MedievalMatt said:

I wonder if that perhaps is the idea behind the 1gpm (227ish lph) mentality.  Where if your flow is that high rad placement location is irrelevant at that point.

 

I can say that with my flow rate it matters where the rads are in the loop, but I have so many and with so much area, that it effectively doesnt.

Yeah its typically once you hit 3gpm that it truly is diminished returns. 1gpm isn't that much, but is a good spot to aim for.

[AnonymousGuy]

1 hour ago, MedievalMatt said:

I wonder if that perhaps is the idea behind the 1gpm (227ish lph) mentality.  Where if your flow is that high rad placement location is irrelevant at that point.

 

I can say that with my flow rate it matters where the rads are in the loop, but I have so many and with so much area, that it effectively doesnt.

1gpm also comes from the "what is realistic" category.  It's a nice round number and flow vs. resistance is exponential...so 1gpm is sorta before it starts really ramping up where you need tons more pressure to generate marginally more flow.  Tons of pressure like you get with running 3x D5 pumps in series is how you force leaks.

 

IMO anything above .5gpm is really OK even at higher power loads like 700W.  Nowadays I generalize that unless you're doing something like 7 gamers 1 cpu, there is no build that requires more than a single D5 pump.

[thrasher_565]

more flow = lower temps its been proven but its not worth it. like i said every 1 gpm is 1c drop... but that was at a time when flow was king and we didn't have things like jet plates slow things down. so the hard core water coolers used duel d5 pumps or more... to get like 7gpm. water cooling testing has been largely put on the back burner as its too expensive for more people and since you dont gane value oc anymore just buy a faster gpu/cpu if you want more power... but back then people did anything to get the flow rate up...

 

the cpu jet plate and i think gpu have jet plats too now? would be a bottle neck and there probably a limit on how much you can force thow it...

[AnonymousGuy]

13 hours ago, thrasher_565 said:

more flow = lower temps its been proven but its not worth it. like i said every 1 gpm is 1c drop... but that was at a time when flow was king and we didn't have things like jet plates slow things down. so the hard core water coolers used duel d5 pumps or more... to get like 7gpm. water cooling testing has been largely put on the back burner as its too expensive for more people and since you dont gane value oc anymore just buy a faster gpu/cpu if you want more power... but back then people did anything to get the flow rate up...

 

the cpu jet plate and i think gpu have jet plats too now? would be a bottle neck and there probably a limit on how much you can force thow it...

Yeah, every gpu block for the last few years uses a jet plate as well.

 

Fun fact: if you get crazy flowrates it can have the opposite effect where you create a marginally higher temperature from the friction of overchurning the water.  But we're talking like 1-2C within-the-margin changes.

 

IMO water cooling testing has gone down because everything has plateaued.  Even the chinese clone shit basically does as well as the overpriced EK stuff, and it got real boring reading reviews where the entire "results" table was every product within 2C of each other.

 

This spread on radiators for example:

 

[MedievalMatt]

5 hours ago, AnonymousGuy said:

Yeah, every gpu block for the last few years uses a jet plate as well.

 

Fun fact: if you get crazy flowrates it can have the opposite effect where you create a marginally higher temperature from the friction of overchurning the water.  But we're talking like 1-2C within-the-margin changes.

 

IMO water cooling testing has gone down because everything has plateaued.  Even the chinese clone shit basically does as well as the overpriced EK stuff, and it got real boring reading reviews where the entire "results" table was every product within 2C of each other.

 

 

JayzTwoCents does some coverage still.  But I do agree the market has stagnated.

 

The advent of Turbo and other "boost" things from factory are likely the cause combined with relatively low power draw components.

 

I suspect the newer stuff in the next few years will be a boon for watercooling in general.  a 450Watt 4070 is going to push a lot of people to watercooling.  Personally i dont miss the noise of my air cooled 3080ti FTW3.

[thrasher_565]

11 hours ago, AnonymousGuy said:

Yeah, every gpu block for the last few years uses a jet plate as well.

 

Fun fact: if you get crazy flowrates it can have the opposite effect where you create a marginally higher temperature from the friction of overchurning the water.  But we're talking like 1-2C within-the-margin changes.

 

IMO water cooling testing has gone down because everything has plateaued.  Even the chinese clone shit basically does as well as the overpriced EK stuff, and it got real boring reading reviews where the entire "results" table was every product within 2C of each other.

 

This spread on radiators for example:

 

ya there alot of factors that killed it. aio, higher cost, ocing not as good of value, dont need 4x 360 rads to cool any more (except the 8800k...) cases got better air cooling, rgb, sli dead, kinda funny how cases got more rad space... at the end of wc...

 

now a days its a minim of $1000 to get a loop on top of a new build cost of like $1800... instead people will just buy lower tier and upgrade say 5 years.

 

i never got to that point of doing test as i broke my koolanc controller... but some day i plan to have a full aquacomputer setup.

[fonzz1e]

IMO gpu benefits the most being watercooled

my 3070 on reaches 84c on air with 33c ambient (imma in tropical weather)

 

now at most it only peaks 61c with bykski block

i do plan to keep it for longterm (my last one was 1070)

 

i do expect more benefits for hot running gddr6x

[Sir Beregond]

Well if you take the time to tweak, and dial in settings, watercooling still has the advantage of lower temps, often allowing for higher boosts, etc. Take PBO/CO with Ryzen for example.

[MedievalMatt]

15 hours ago, Sir Beregond said:

Well if you take the time to tweak, and dial in settings, watercooling still has the advantage of lower temps, often allowing for higher boosts, etc. Take PBO/CO with Ryzen for example.

I can boost the core clocks on my 3080ti to 2100MHz in some cases, and I have an all core load speed of 4 GHz on my 5900X.  which is significantly more than i was getting on air.

 

It may also be worth noting that some of us live in climates where watercooling makes more sense.

[Sir Beregond]

7 hours ago, MedievalMatt said:

I can boost the core clocks on my 3080ti to 2100MHz in some cases, and I have an all core load speed of 4 GHz on my 5900X.  which is significantly more than i was getting on air.

 

It may also be worth noting that some of us live in climates where watercooling makes more sense.

I go PBO/CO for my 5900X, but also get 2100MHz boosts on the 3080 Ti in some cases. Can run stable above 2GHz. Haven't even tried undervolting yet.