Discussion A Procedure for Defining the Correct Pump Speed and Radiator Fan Speed

Imacflier

Distinguished
Jan 19, 2014
558
27
19,090
Hi, Guys,

I am building a strictly gaming and rendering system on an ITX platform.
It will use a custom loop with a PWM pump and and 200mm x 200mm x 75mm(?) PWM fan and radiator.
I will be using a Gigabyte Z390 I AORUS PRO WIFI (rev. 1.0) motherboard.
The CPU will be a Core I9-9900ks.
The GPU will be an RTX 3080 Waterforce.
System storage will be a pair of 4TB m.2 nvme drives.
System RAM will be 64 GB of 4400MHz DDR4
Power Supply is TBD, so please recommend something?

The Gigabyte software provides an intuitive, graphical control of pump and fan speeds.
The issue is determining the best curve for the cooling system. Obviously slower is better to minimize system noise generation.

Ms. Google says that the idle temperature is about 40* and thermal throttling occurs somewhere around 90-100* for the I9-9900ks. My previous experience with space level hardware tells me that I should have enough cooling capacity to keep temperatures at no more than 80% of maximum.

So the following is the procedure I believe will satisfy the above requirements (NOTE: I have NOT tried this yet, so I earnestly solicit your opinions and comments!):

1. Using the software default speed values, run the system at idle.
2. If the cpu temp is less than 40* (and I expect it to be lower), manually reduce the pump and radiator fan speed until it equals 40*. This is the low rpm setting for the speed control software..
3. Manually set the pump and fan to maximum.
4. Stress test the cpu to ensure that at 100% turboboost utilization the cooling system keeps the cpu at 80* or less. If it does not, then additional radiator capacity is required!
5. If the cpu temp is less than 80*, then reduce the pump and radiator fan speeds until it does equal 80*. This sets the maximum point of the pump and radiator fan speed control to the lowest value acceptable.

So far as I am aware heat generation is linear with increasing cpu load, so the two points are all that is needed to completely define the pump and radiator speed curve.

I see I have neglected the cooling load of that RTX 3080 Waterforce, which means I will have to find a way to stress test the cpu and gpu concurrently at step 3 above.

What else have I missed or got wrong and is there a easier or more reliable way to define the speed curve?

Discussion, criticism, and advice are open!!

TIA

Larry
 
Hi, Guys,

I am building a strictly gaming and rendering system on an ITX platform.
It will use a custom loop with a PWM pump and and 200mm x 200mm x 75mm(?) PWM fan and radiator.
I will be using a Gigabyte Z390 I AORUS PRO WIFI (rev. 1.0) motherboard.
The CPU will be a Core I9-9900ks.
The GPU will be an RTX 3080 Waterforce.
System storage will be a pair of 4TB m.2 nvme drives.
System RAM will be 64 GB of 4400MHz DDR4
Power Supply is TBD, so please recommend something?

The Gigabyte software provides an intuitive, graphical control of pump and fan speeds.
The issue is determining the best curve for the cooling system. Obviously slower is better to minimize system noise generation.

Ms. Google says that the idle temperature is about 40* and thermal throttling occurs somewhere around 90-100* for the I9-9900ks. My previous experience with space level hardware tells me that I should have enough cooling capacity to keep temperatures at no more than 80% of maximum.

So the following is the procedure I believe will satisfy the above requirements (NOTE: I have NOT tried this yet, so I earnestly solicit your opinions and comments!):

1. Using the software default speed values, run the system at idle.
2. If the cpu temp is less than 40* (and I expect it to be lower), manually reduce the pump and radiator fan speed until it equals 40*. This is the low rpm setting for the speed control software..
3. Manually set the pump and fan to maximum.
4. Stress test the cpu to ensure that at 100% turboboost utilization the cooling system keeps the cpu at 80* or less. If it does not, then additional radiator capacity is required!
5. If the cpu temp is less than 80*, then reduce the pump and radiator fan speeds until it does equal 80*. This sets the maximum point of the pump and radiator fan speed control to the lowest value acceptable.

So far as I am aware heat generation is linear with increasing cpu load, so the two points are all that is needed to completely define the pump and radiator speed curve.

I see I have neglected the cooling load of that RTX 3080 Waterforce, which means I will have to find a way to stress test the cpu and gpu concurrently at step 3 above.

What else have I missed or got wrong and is there a easier or more reliable way to define the speed curve?

Discussion, criticism, and advice are open!!

TIA

Larry
Generally the pump is most efficient at 90 to 100% constant. IMO, use vary just the fan speed for quieting.
Idle temps can't be determined without the entire cooling system coming to a stable temp, which depending on the radiator, fans, etc could take a long time.
 
Kanewolf,

Good points, both. However, unless I run the pump at full speed at all times, the rate of heat removal from the cpu is inversely proportional to flow rate. I am not sure if the time scales for that removal are significant or not....looks as if an experiment will be necessary!

And an experiment on the time required forcooling system stabilation on cpu idle as well since the rate of heat removal becomes less as the temperature delta decreases. I have no experience on the times involved al though I suspect you are correct.

Thank you very much for the well considered reply!

Larry
 
Kanewolf,

Good points, both. However, unless I run the pump at full speed at all times, the rate of heat removal from the cpu is inversely proportional to flow rate. I am not sure if the time scales for that removal are significant or not....looks as if an experiment will be necessary!

And an experiment on the time required forcooling system stabilation on cpu idle as well since the rate of heat removal becomes less as the temperature delta decreases. I have no experience on the times involved al though I suspect you are correct.

Thank you very much for the well considered reply!

Larry
I forgot to ask: how would YOU go about determining the idle end of the cooling curve?

I suppose I could use an arbritrary temperature between 40* and 80* (say 60*)and extend the line from the 80* setting through the 60* setting to reach the desired 40* idle setting.....but that seems a bit inelegant!

Running the pump full speed at all times can be quite loud on many DDR pumps, btw!

Larry
 
I forgot to ask: how would YOU go about determining the idle end of the cooling curve?

I suppose I could use an arbritrary temperature between 40* and 80* (say 60*)and extend the line from the 80* setting through the 60* setting to reach the desired 40* idle setting.....but that seems a bit inelegant!

Running the pump full speed at all times can be quite loud on many DDR pumps, btw!

Larry
I would never use water cooling. Too much hassle for me. I don't push hardware to extremes. Big air all the way for me.
 
Hi, Guys,

I am building a strictly gaming and rendering system on an ITX platform.
It will use a custom loop with a PWM pump and and 200mm x 200mm x 75mm(?) PWM fan and radiator.
I will be using a Gigabyte Z390 I AORUS PRO WIFI (rev. 1.0) motherboard.
The CPU will be a Core I9-9900ks.
The GPU will be an RTX 3080 Waterforce.
System storage will be a pair of 4TB m.2 nvme drives.
System RAM will be 64 GB of 4400MHz DDR4
Power Supply is TBD, so please recommend something?

The Gigabyte software provides an intuitive, graphical control of pump and fan speeds.
The issue is determining the best curve for the cooling system. Obviously slower is better to minimize system noise generation.

Ms. Google says that the idle temperature is about 40* and thermal throttling occurs somewhere around 90-100* for the I9-9900ks. My previous experience with space level hardware tells me that I should have enough cooling capacity to keep temperatures at no more than 80% of maximum.

So the following is the procedure I believe will satisfy the above requirements (NOTE: I have NOT tried this yet, so I earnestly solicit your opinions and comments!):

1. Using the software default speed values, run the system at idle.
2. If the cpu temp is less than 40* (and I expect it to be lower), manually reduce the pump and radiator fan speed until it equals 40*. This is the low rpm setting for the speed control software..
3. Manually set the pump and fan to maximum.
4. Stress test the cpu to ensure that at 100% turboboost utilization the cooling system keeps the cpu at 80* or less. If it does not, then additional radiator capacity is required!
5. If the cpu temp is less than 80*, then reduce the pump and radiator fan speeds until it does equal 80*. This sets the maximum point of the pump and radiator fan speed control to the lowest value acceptable.

So far as I am aware heat generation is linear with increasing cpu load, so the two points are all that is needed to completely define the pump and radiator speed curve.

I see I have neglected the cooling load of that RTX 3080 Waterforce, which means I will have to find a way to stress test the cpu and gpu concurrently at step 3 above.

What else have I missed or got wrong and is there a easier or more reliable way to define the speed curve?

Discussion, criticism, and advice are open!!

TIA

Larry
To All,

I am rapidly reaching the point of needing to order my PSU! Given I9-9900ks and RTX 3080...both likely overclocked to some irrational degree, is an sfx 750 enough? (There highly rated Corsair SFX units avalable at 750, 850, and 1000 watts available).

Whatcha all recommend?

TIA

Larry