Hey everyone,

I wanted to share a cooling-control setup I built using a Quadro. I currently have four physical temperature sensors and three main water-cooling control groups: intake, exhaust, and pump.

This took a bit of trial and error because none of my connected devices were reporting RPM quite the way I expected. Since my pump is not powered directly by the controller, I also had to estimate some of its behavior. For the intake and exhaust fans, I estimated RPM using the fan model’s published performance data combined with the commanded power level. I used a similar approach for the pump, building a rough but useful estimate of power usage based on RPM and commanded output.

By collecting intake air temperature, internal case temperature, coolant temperature, and exhaust temperature, I was able to calculate temperature deltas and use them to better understand loop efficiency and overall case airflow effectiveness. Then using creative overrides, I set loss of signal power presents. Notice, my override is the primary commanded control.

One of the biggest benefits was finding a good balance between intake and exhaust airflow. In my specific setup, a slight negative-pressure configuration appears to cool more efficiently. It seems to move heat out of the case quickly enough that the intake radiator can continue pulling in cooler air effectively.

I also built some “emergency” control logic. Since I’m running a 5800X3D, the CPU is very heat-dense and can spike in temperature much faster than the coolant temperature can react. Because of that, I created mathematical overrides for the normal fan and pump curves. If the CPU approaches my upper temperature limits, the system forces higher fan and pump speeds even if the coolant temperature itself has not risen much yet.

After watching the data for a couple of weeks and making small adjustments, I was able to tune the setup into what feels like a really efficient cooling profile for my system.

The software itself is very capable, but I do wish there were more freedom on the design side. XAML feels a bit restrictive, and I think something like a Python wrapper could open up a lot of creative dashboard and visualization options. For example, displaying 09 instead of 9? A lot of the data and mathematics were done with virtual sensors.

Also, slightly unrelated: does anyone know how to force leading zeroes on single-digit date/time fields?

Water-cooling & Aesthetics Components​:

• Alphacool AIO Quick Disconnect (QD) Hoses – 90° (x2)
• Alphacool AIO Quick Disconnect (QD) Hoses – Straight (x2)
• Alphacool Core 100 Aurora Pump
• G1/4 Temperature Sensor
• Open Air Temperature Sensors (x3)
• Molex PSU Cable
• Aquacomputer Quadro Controller
• RGB Splitters (Generic)
• Cooler Master Vertical GPU Mount
• Corsair Hydro XR5 420mm x 30mm Radiator
• Bykski Full-Coverage GPU Water Block (Sapphire Pulse 24GB IC flashed to Nitro+)
• Heatkiller IV AM4 CPU block
• New 24 pin motherboard cable with LEDs,
• 8-Pin PCIe GPU Power 180-degree adapters.
• Added a 4-way fan splitter for the exhaust fans to use a single DC header.
• Combine all intake fans on a single DC header.
• Combine all exhaust fans on a single DC header.
• Mayhems XT-1 Coolant (1000ml)
  

PC Components:

• CPU: 5800X3D
• RAM: 64GB GSkill CAS 16
• GPU: Saphire Pulse 7900XTX 24GB running Nitro+ Unlocked Firmware
• Storage: Samsung 990 Pro 2TB NVMe Gen 4
• Motherboard: MSI B550i Gaming Wi-Fi
• Network: MSI Hearld Wifi 7 Broadcom chipset (FastConnect 7800)
• Intake: 3x 140mm fans
• Exhaust: 4x 120mm fans

Thanks for the kind words and advice.


In my specific setup, I’m running a fairly large radiator, so during CPU-heavy workloads the coolant temperature does not rise very much compared to gaming loads, where the GPU can add roughly 500W into the loop. I’ve actually seen better CPU-only temperatures when the pump runs slower. My assumption is that the reduced flow gives the coolant more time in the radiator, although I realize the full behavior depends on the entire loop and airflow setup.


I also keep my exhaust fans about 5% higher than my intake fans because my temperatures appear to be better with slight negative case pressure.


For example, right now my cooling-efficiency delta is 3.3°C, and my case-airflow delta is 3.2°C, with the pump at 54%. If I manually raise the pump to 90%, that delta actually increases by almost 1°C. It is not a huge difference, but it does show that liquid cooling, radiator performance, case layout, fan type, fan size, airflow balance, and pump speed all work together as a system. Even my idle CPU temperatures reflect this: around 41°C at 50% pump speed versus about 54°C at 95% pump speed.


Because of that, I have the pump tied primarily to CPU temperature, with an emergency trigger if the CPU reaches 84°C. When that happens, the fans ramp up as well to help bring coolant temperature down faster and increase case airflow. Even if the coolant temperature is not rising dramatically during CPU-only workloads, the added airflow also helps evacuate heat from inside the case.


I also want to clarify that this is not just guesswork. I logged data continuously for about two days and then analyzed the results. My current configuration is based on that testing and the patterns shown in the data.


Regarding Jacob’s suggestion to switch the controller to PWM mode, that is not an option with my current controller. I’m using the entry-level model, which is basic but still very capable. It only supports DC control. My assumption is that because the fans are daisy-chained, the RPM reporting is being thrown off. As for the pump, I’m not completely sure. The 4-pin cable is plugged into the controller’s 4-pin header, so I would expect RPM reporting to work, but it does not appear to read correctly. My workaround is accurate enough for my purposes, so I’m comfortable using it for now.

Zitat von »Remayz«

A few notes from what i saw, in no particular order
for the fans, 5% PWM difference is basically the same. even on your screenshots you can see they spin at the same speed. Most fans have a wide margin for max speed tolerance from the factory like +-10%.
Since you have 3 intakes obstructed by a radiator, and 4 free exhausts, there's nothing to do really, it's already naturally negative pressure.

With a 5800 X3D, i wouldn't bother either about controlling anything with CPU temperature it's toasty regardless just because of the hot IO die, the Vcache on top and the massive IHS. The water loop infuences it less that the Vcore does. All it will do is render your pump noisy if you don't add a massive hysteresis on the constantly fluctuating CPU temp.

For your fan and pump control with input offsets, it's a bit odd You are forcing yourself into a flat fan or pump curve basically. you can achieve the same using a regular fan curve and leaving it set to flat.
Instead of setting the output % manually and controlling the offset, you can set the % with a fan curve and manually control the offset, leaving you the option to have a non linear output with a curved... curve. Since you have a single radiator for a X3D CPU + a 7900XTX, it may help as the CPU will always have a wiuldly fluctuating temperature that you may want to filter out with a flatter low end of the curve, and the GPU will run the water fairly warm under load with a single radiator.

Finally for emergency cooling, you can simply use the bbuilt in profile switching of profiles on Aquasuite. Just set the switching to your sensor of choice, and if say the temperature exceeds a certain threshold, it'll go to a whole other profile with more agressive cooling, until the temperature is under control.

Edit : i forgot to add, that's just my 2 cents it's custom loop for a reason, do it as you like!

So, what I'm hearing is. "You took that hardest route possible to reach the goal." Classic me, I guess! A second profile does seem easier, but I'll have to look into that I don't know how to do that. Honestly everything at full bore is still almost 20 decibels lower than what I had for air cooling. 7900XTX too is a temp hog with the 500w firmware I put on it. I put two fireballs in my system. My idle temps for the GPU are lower than that of the CPU when not gaming. 7900XTX loves the water im giving it. I have a 15% overclock on the card and it doesn't miss a beat.