What is your DeltaT (coolant temp)?

[sblantipodi]

As title. What is your DeltaT?

Please tell us your DeltaT under max load for at least ten minutes and post your:
- max watt
- number of rads and fans, fan speer
- pump speed details
- ambient temp
- case in use
- CPU and GPU

My setup achieve this:

With 29°C ambient temp,
using 2x Corsair XR5 360mm and 1x Corsair XR7 360mm rad
with fans 10x QX120 fans 1300rpm
with a max load of 700w for 15 minutes
a single DC5 pump at 4000rpm
case Lian Li O11 Dynamic Evo RGB
7950X3D with an overclocked RTX4090

my DeltaT is 15°C

What is yours?

 

[Zarathustra[H]]

For purposes of "max load". How are you loading it up?


Prime95 + Unigine Heaven on a loop?

Come to think of it, heaven is kind of old at this point. What other good GPU benchmark can be put on a loop like that?

 

[LukeTbk]

Furmark stresstest is made to push the gpu to 100% I think, maybe not good at much else, put to put a maximum of heat from a gpu it could be the way to go.

 

[Zarathustra[H]]

Furmark stresstest is made to push the gpu to 100% I think, maybe not good at much else, put to put a maximum of heat from a gpu it could be the way to go.

I used to use Furmark way back in the day, but after it was declared a "power virus" by both sides of the GPU war, I decided maybe not to use it anymore.

I remembered that I had bought a copy of 3DMark on sale at some point, so I queued up Port Royal with settings turned up to high and ran in loop mode in a 2560x1440 window.

I then went into task manager and set process priority for Port Royal to "above normal" to ensure it got all the CPU it needed to max out the GPU.

Then I loaded up Prime 95, set it to use all cores, HT and let it run as well, going into task manager and setting process priority to "below normal" just to make sure prime 95 didn't prevent Port Royal from maxing out the GPU.

My specs are in sig for this test.

The 4090 had a mild overclock on it as follows:

Cooling loop is per this thread (the last configuration as it changed a few times)

The system has three pwm controlled D5 pumps. Two of them in an EK dual fan pump head push the coolant through the radiators. The radiators are two Alphacool UT60 Xflow 460mm radiators and one Alphacool XT45 420mm radiator up top. Each radiator slot is populated in push-pull with Noctua Industrial PPC-2000 PWM fans for a grand total of 16x 120mm fans and 6x 140mm fans. After the radiators the water from this sub-loop goes back tot he reservoir.

The third pump pushes the water through the two blocks, a Watercool Heatkiller IV Pro for Threadripper on the CPU and the GPU block is an EK Fullcover block for the 4090.

The return from each sub-loop is mixed in the reservoir so that the radiators still cool the water going into the blocks, despite being separate subloops.

I then waited 15 minutes and took a screen-shot.

Power use at the wall (using a Kill-A-Watt) was as follows:

The full screenshot with all of this going on looked like this:

With teo 1080p screens and one 4k screen its a little big for the forums, so zooms follow below:


Here is my custom Aquacomputer dashboard:

Notably a 1037W output over 15 minutes raised the temperature in the computer room by about 1.5C in 15 minutes it's a bloody space heater

Since my build uses long extension cables and sits in another room, I have no problem with th enoise from high fan speeds. I control both fan and pump speeds based on temperature using PWM and my two aquaero's. In the case of this test, since load was an extreme worst case, that kicked everyting into 100% speed.

Still couldn't hear anything from my office though.


The Rivatuner charts that show everything that was going on are here:

Analysis:

Looks like ambient was 25.6C.

Fluid temp entering the hot (block) side of the loop was 30.4C

Fluid temp exiting the hot (block) side of the loop was 33.0C

Normally I would suggest that your pump speed should be high enough that you don't have a measurable difference regardless of where in the loop you measure, but I guess even at full speed, when you are dumping over a kilowatt of heat into the loop, maybe that is unrealistic. In my case there is a measurable difference of about 2.6C between the water entering the block loop and the water leaving the block loop after having passed through the GPU and CPU blocks, even at max speed and a flow rate of 226 L/h (~1GPM). The flow rate through the radiators is insane (629 L/H, 2.77GPH) due to the rater low resistance and the dual pumps. If I had to do this again, I'd use one pump on the radiator side and two on the GPU side, but I'd probably need a different case for that, as that just wouldn't fit even in my Corsair 1000D.

I'm not 100% sure the flow rate numbers are reliable. The Aquacomputer flow rate sensors are among the best there are, but they still seem a biot wonky for me, and only 1 GPM seems very low for a single D5 pump at max speed only going through two blocks. And the radiator flow rate of 2.77GPM sees almost a little too high.

(I do prefer GPM as a unit for flow in wartercooling, but Aquacomputer - in their wisdom - allows two system-wide settings, imperial or metric. So I have to choose which bothers me least. Water temps in farenheit or flow rates in L/h. It winds up being the flow rate. I love just about everything esle about my Aquaero's except this system wide unit preference thing)

Anyway, depending on which temp you use (before blocks or after blocks) the delta between the loop temp and ambient is either 4.8C or 7.4C

To be comparable to a typical loop where flow rate is high compared to heat output, and temp is assumed to be at a steady state throughout the loop, it seems approriate to use an average of the two, so I'll go with 6.1C.

 

[sblantipodi]

For purposes of "max load". How are you loading it up?


Prime95 + Unigine Heaven on a loop?

Come to think of it, heaven is kind of old at this point. What other good GPU benchmark can be put on a loop like that?

a 3D Mark, steel nomad or port royal in stress test mode for 15 minutes should be enough even if CPU is not really maxed out in that test.

 

[NightReaver]

Man I need to get my loop back up and going. Apparently my case can fit 4x480.

 

[kamikazi]

I need to try this out tonight and post back later. My pump and fan settings are controlled by the liquid temp, so it's much slower to react to spikes in CPU and GPU usage, but I'll bet loading it up to 100% for 15 minutes should heat it up pretty good.

 

[kamikazi]

I need to try this out tonight and post back later. My pump and fan settings are controlled by the liquid temp, so it's much slower to react to spikes in CPU and GPU usage, but I'll bet loading it up to 100% for 15 minutes should heat it up pretty good.

I just ran the Aida64 Stability Test for 15 minutes. CPU and GPU stayed maxed at 100%. CPU maxed at 69C and GPU at 50C. I could hear the fans ramp up just a bit, but it was still super quiet. See my signature for specs. Looks like there is a 1.5C temp differential in the fluid between the hot side and the cold side of the radiator. There was about a 7C temp differential between ambient and the loop. If I used a different performance preset I have set up on my aquereo, I could tighten that up for sure. Flow rate was around .8 gallon per minute. Those T30 fans and P12 fans are pretty quiet. If I ramped those up along with the dual DDC pumps, I could keep things cooler, but I'm happy with the current settings for an all around setup.

 

[xDiVolatilX]

I used to use Furmark way back in the day, but after it was declared a "power virus" by both sides of the GPU war, I decided maybe not to use it anymore.

I remembered that I had bought a copy of 3DMark on sale at some point, so I queued up Port Royal with settings turned up to high and ran in loop mode in a 2560x1440 window.

I then went into task manager and set process priority for Port Royal to "above normal" to ensure it got all the CPU it needed to max out the GPU.

Then I loaded up Prime 95, set it to use all cores, HT and let it run as well, going into task manager and setting process priority to "below normal" just to make sure prime 95 didn't prevent Port Royal from maxing out the GPU.

My specs are in sig for this test.

The 4090 had a mild overclock on it as follows:

View attachment 675593

Cooling loop is per this thread (the last configuration as it changed a few times)

The system has three pwm controlled D5 pumps. Two of them in an EK dual fan pump head push the coolant through the radiators. The radiators are two Alphacool UT60 Xflow 460mm radiators and one Alphacool XT45 420mm radiator up top. Each radiator slot is populated in push-pull with Noctua Industrial PPC-2000 PWM fans for a grand total of 16x 120mm fans and 6x 140mm fans. After the radiators the water from this sub-loop goes back tot he reservoir.

The third pump pushes the water through the two blocks, a Watercool Heatkiller IV Pro for Threadripper on the CPU and the GPU block is an EK Fullcover block for the 4090.

The return from each sub-loop is mixed in the reservoir so that the radiators still cool the water going into the blocks, despite being separate subloops.

I then waited 15 minutes and took a screen-shot.

Power use at the wall (using a Kill-A-Watt) was as follows:

View attachment 675594


The full screenshot with all of this going on looked like this:

View attachment 675595

With teo 1080p screens and one 4k screen its a little big for the forums, so zooms follow below:


Here is my custom Aquacomputer dashboard:

View attachment 675596

Notably a 1037W output over 15 minutes raised the temperature in the computer room by about 1.5C in 15 minutes it's a bloody space heater

Since my build uses long extension cables and sits in another room, I have no problem with th enoise from high fan speeds. I control both fan and pump speeds based on temperature using PWM and my two aquaero's. In the case of this test, since load was an extreme worst case, that kicked everyting into 100% speed.

Still couldn't hear anything from my office though.


The Rivatuner charts that show everything that was going on are here:

View attachment 675597


Analysis:

Looks like ambient was 25.6C.

Fluid temp entering the hot (block) side of the loop was 30.4C

Fluid temp exiting the hot (block) side of the loop was 33.0C

Normally I would suggest that your pump speed should be high enough that you don't have a measurable difference regardless of where in the loop you measure, but I guess even at full speed, when you are dumping over a kilowatt of heat into the loop, maybe that is unrealistic. In my case there is a measurable difference of about 2.6C between the water entering the block loop and the water leaving the block loop after having passed through the GPU and CPU blocks, even at max speed and a flow rate of 226 L/h (~1GPM). The flow rate through the radiators is insane (629 L/H, 2.77GPH) due to the rater low resistance and the dual pumps. If I had to do this again, I'd use one pump on the radiator side and two on the GPU side, but I'd probably need a different case for that, as that just wouldn't fit even in my Corsair 1000D.

I'm not 100% sure the flow rate numbers are reliable. The Aquacomputer flow rate sensors are among the best there are, but they still seem a biot wonky for me, and only 1 GPM seems very low for a single D5 pump at max speed only going through two blocks. And the radiator flow rate of 2.77GPM sees almost a little too high.

(I do prefer GPM as a unit for flow in wartercooling, but Aquacomputer - in their wisdom - allows two system-wide settings, imperial or metric. So I have to choose which bothers me least. Water temps in farenheit or flow rates in L/h. It winds up being the flow rate. I love just about everything esle about my Aquaero's except this system wide unit preference thing)

Anyway, depending on which temp you use (before blocks or after blocks) the delta between the loop temp and ambient is either 4.8C or 7.4C

To be comparable to a typical loop where flow rate is high compared to heat output, and temp is assumed to be at a steady state throughout the loop, it seems approriate to use an average of the two, so I'll go with 6.1C.

Bro, the amount of effort you put into a post is astonishing. Good job.
Question, how did you get the keyboard and mouse to work from the other room?

 

[sblantipodi]

I just ran the Aida64 Stability Test for 15 minutes. CPU and GPU stayed maxed at 100%. CPU maxed at 69C and GPU at 50C. I could hear the fans ramp up just a bit, but it was still super quiet. See my signature for specs. Looks like there is a 1.5C temp differential in the fluid between the hot side and the cold side of the radiator. There was about a 7C temp differential between ambient and the loop. If I used a different performance preset I have set up on my aquereo, I could tighten that up for sure. Flow rate was around .8 gallon per minute. Those T30 fans and P12 fans are pretty quiet. If I ramped those up along with the dual DDC pumps, I could keep things cooler, but I'm happy with the current settings for an all around setup.

View attachment 676065
View attachment 676066

what CPU and GPU are you using? signature seems disabled.
AIDA is not really a good test for the GPU heat test...

 

[kamikazi]

what CPU and GPU are you using? signature seems disabled.
AIDA is not really a good test for the GPU heat test...

5950x and 3090. Agreed on testing, it was quick and dirty.

 

[Zarathustra[H]]

Bro, the amount of effort you put into a post is astonishing. Good job.
Question, how did you get the keyboard and mouse to work from the other room?

Thanks.

I just used one of these plugged into the computer, and one of these hubs in my office.


It's all in this thread.


You just have to do a little bit of trial and error, as if you daisy chain the wrong hubs together you get too much signal degradation and it doesn't work. These work for me. If you need more ports, that's where the trial and error starts, as some hubs will work just fine if you daisy chain them, and others just won't.

 

[kamikazi]

5950x and 3090. Agreed on testing, it was quick and dirty.

Inspired by Zarathustra[H] , I re-tested. I looped Port Royal and ran Prime95, 16 cores, small FFTs for 15 minutes. I pretty much got the same results as my last test.

Total System Power around 650 watts:

Max CPU temp around 72C, but it stayed around 67C while I was watching:

Max CPU Power around 172W total package:

Max GPU temp around 50C:

Max GPU Power almost 400W:

Coolant Temps, Ambient Temps, Fan Speeds, Pump Speeds, etc.:

1.4C delta between pre and post radiator cooling liquid
About 7C water to air delta
Pump at .80 gallons/minute
Phanteks T30 pull fans at 1074 rpm (55% PWM of medium setting on fan)
Arctic P12 push fans at 909 rpm (55% PWM)
Dual EK DDC pumps at 3906 rpm (percentages of PWM don't mean much here based on the way you set them up in the system)

This is a very quiet setup, even at cranked CPU and GPU output. The coil whine from the 3090 is more noticeable.

My system doesn't pull 1000 watts like Z's, but I have the cooling for it if needed.