Cooling 2023 test protocol


Since 2018, the PC processor market has seen a good number of generations come and go. And with good reason: since the last update of our cooling test protocol, we’ve had three generations of Ryzen CPUs and four on the Intel side. In fact, at the time of our last update, Intel was just about to release its Core i 9000, while AMD was only at Ryzen 1000.

So why the change of protocol? Quite simply because, on the one hand, the number of cores has increased yet again. On the other, CPU design has become radically different. While Intel retains a monolithic design on its thirteenth generation, AMD has switched to MCM since the Ryzen 3000. All this changes the way processors heat up… Not to mention their power consumption, which has risen sharply. These are the reasons that have prompted us to rethink our methodology.

So, in order to measure the behavior of AIOs and fans on Intel and AMD platforms, we’ve opted for the following two configurations:

Configurations for our new cooling protocol :

  • AMD :
    • AMD Ryzen 7 7900X
    • MSI X670E Carbon WiFi
    • Kingston Fury Beast DDR5 2×16 GB (KF560C40BBAK2-16)
    • GTX 690 for display
    • NZXT C1000 Gold V2
    • Thermaltake Commander F6 rheobus
  • Intel :
    • Intel Core i5 13600K + Thermalright Contact Frame
    • MSI Z790 Carbon WiFi
    • Kingston Fury Beast DDR5 2×16 GB (KF560C40BBAK2-16)
    • GTX 690 for display
    • NZXT C1000 Gold V2
    • Thermaltake Commander F6 rheobus

Protocole test cooling 2023

Between the two platforms, we note the addition of a Contact Frame positioned on the Intel processor. The aim is to improve the contact surface with the heatsinks we test. All fans and AIOs are compared with this device.

Room temperature:

Please note that all our CPU cooling tests are carried out outside the case, in a room maintained at around 20°C (±2°C). So, in winter, we turn up the heat, and in summer, as the tests are carried out in Brittany, it’s not a problem because it rains all the time and there’s never any sun. Joking aside, in summer we take our readings when temperatures are most favorable, in the morning for example.

Ventilation management:

For our new protocol, fan management is still entrusted to our Thermaltake F6 rehobus. For all-in-one watercooling, the fans are connected to the rheobus, as is the pump.

We set the cooling between three settings:

  • High speed: 12V fans and 12V pump
  • Medium speed: 8V fans and 12V pump.
  • Low speed: 5V fans and 12V pump.

Compared with our previous protocol, we leave the pump running at full speed, while the fan runs at low speed. The power consumption of the 13600K and 7900X is nowhere near that of our old 1700X.

Nevertheless, the aim remains to have three clearly defined noise levels.

Thermaltake Commander FC6

As our rheobus allows us to adjust the ventilation to the nearest tenth of a volt, we’ll be running the fans according to three profiles: 5V, 8V and 12V. So why 8V? To be, more or less, in the middle of the fans’ operating range.

However, some fans may not be suitable for 5V. In which case they either don’t start up, or they do start up, but run at a ridiculously low speed, outside their specifications. The fan is then operating in a clearly unsuitable electrical environment. If such a case arises, we’ll increase the voltage slightly, so that we can once again comply with the fan’s specs. We set a minimum of 500 RPM, unless the data sheet indicates a lower minimum.

The aim behind this reflection and adaptation is to obtain results corresponding to real uses and possible cooling settings.

CPU profiles :

Next, we want to take readings with higher and lower power consumption, to better show the behavior of CPU cooling. We have therefore defined four stable profiles in our X670E Carbon WiFi and Z790 Carbon WiFi, each corresponding to three different power consumption profiles:

  • AMD :
    • 65W:  Multiplier: x24.5 and VCore at 1,000V
    • 100W: Multiplier factor: x33 and VCore at 1.132V
    • 150W: Multiplier factor: x39 and VCore at 1.246V
    • 200W: Multiplier factor: x40 and VCore at 1.360V
  • Intel :
    • 65W: E-Core/P-Core multiplier: x30/x30 and VCore at 1.105V
    • 95W: E-Core/P-Core multiplier: x35/x35 and VCore at 1.210V
    • 130W: E-Core/P-Core multiplier: x35/x45 and VCore at 1.270V
    • 170W: E-Core/P-Core multiplier: x40/x50 and VCore at 1.350V
Review: Elegoo Saturn 4 Ultra

Of course, there are ways of optimizing profiles in terms of frequency/consumption/heating, but that’s not the objective here. Our aim is that, at the time of our readings, power consumption was always around 100W/150/200W on our AMD platform. The same goes for Intel, with 95W, 130W and 170W, all of them stable.

This gives us an idea of the performance of cooling systems at different power consumption levels. For example, 65W will only be used for aircooling heatsinks and 120 mm AIOs. This corresponds to the consumption of low-power CPUs such as the Core i3 13400F.

Between 95W and 100W, we find the same power consumption as the Ryzen 5 7600X and Ryzen 7 7800X3D. Between 130W and 150W, we’re now around the power consumption of a Ryzen 9 7950X3D or a Core i9 10900K. Finally, at 170W-200W, we’re on a par with the @stock consumption of a Ryzen 9 7900X or a Core i7 12700K.

Unfortunately, we would have liked to have gone further in terms of power consumption on our Core i5 13600K. Unfortunately, we’re forced to note that beyond 170W, it’s difficult for us to obtain consistent results with our AIOs, as the temperatures reached are so high.

Our approach:

Before fitting a new heatsink, we clean our CPUs and apply Arctic MX-4 as thermal paste. We connect the fans to the rheobus, just like the pump on an AIO.

At start-up, we select the 100W/95W profile for the first series of readings, 150W/130W for the second and 200W/170W for the last series. Heating times will last around 10 mins per series, and will be broken down as follows:

6 mins of heating, during which the temperature is allowed to stabilize.
At the end of these 6 mins, we’ll have two series of averaged readings of 2 mins each.
We’ll repeat this period of 2-minute readings until the temperature has stabilized (stopped rising).

To stress our Ryzen 9 7900X and Core i5 13600K, we’ll run them under CPUID’s PowerMax. We’ll run an SSE-type test, which we’ll leave running for an unlimited time. The aim is to generate a linear load using 100% of our processors to see how the cooling systems behave.

For the temperature readings, we use the HWiNFO64 software and record the Tdie temperature. There are 9 temperature readings: 3 fan modes with 3 different CPU settings.

Noise measurements are taken during the heating period. The sound level meter is positioned 40 cm from the sound source in the direction of the fan suction. The advantage is twofold: we can measure as many differences as possible from one cooling cycle to the next, and we are not disturbed by air movements.

In order to measure only cooling noise, we block the graphics card fan and ensure that the semi-passive power supply is not triggered at this time.

These measurements will enable us to offer you temperature graphs, noise graphs and graphs showing the evolution of the temperature/noise ratio according to fan mode. In this way, we’ll go into more detail about the behavior of one cpu cooler or AiO versus another.

And off we go!

NZXT Kraken Elite 360 écran

You’ve already had a chance to see what it’s like, albeit partially, in our tests of the Asus AIO Ryuo III 240 and 360, and the NZXT Kraken Elite 360 and Kraken 240 RGB.

We’ll soon be publishing an article featuring the readings of these kits on the Intel platform, to see how they perform on our Core i5 13600K.