Review: Noctua NH-D15 chromax.black

Test Methodology and Performance

Notes

The arrival of our 2020 Test Platform allows us to start afresh with CPU cooler testing. The new PC features a 16-core, 32-thread AMD Ryzen 9 3950X processor and 32GB of dual-channel G.Skill Trident Z Neo DDR4 memory set to run at 3,200 using the built-in profile.

A Fractal Design Define 7 chassis is configured with its three stock Dynamic X2 140mm fans connected to the integrated hub and attached to a single motherboard header. Coolers are tested using default out-the-box settings, and to increase the challenge while minimising noise, all fan headers are set to the 'Silent' profile within the Asus BIOS.

When testing liquid coolers, the pump is connected to the motherboard's dedicated AIO header, and the radiator is installed in the roof of the chassis with fans configured to push air up through the radiator and out of the enclosure. Fractal Design's vented top panel is used with liquid coolers, and the sound-dampened solid panel is in place when testing traditional air coolers.

Actual CPU temperature is recorded and we also graph the delta temperature (CPU temperature minus ambient temperature). Last but not least, to give you an idea of cooler acoustics, we use a PCE-318 noise meter to measure overall system noise in both idle and load states.

Performance

Our benchmarking begins with the 105W AMD Ryzen 3950X CPU running at stock speeds. We use the popular and open-source Blender creation suite to render the taxing Victor scene and record the average CPU temperature from the last five minutes of 100 per cent load.

We've been intrigued to see how the air vs. liquid battle would play out on our new test platform, and though we are not surprised to find the Noctua in the lead, we didn't expect the gap to be quite so large. 16 cores working flat-out results in plenty of heat output, and though your own performance may vary - cooling results can swing significantly depending on chassis and fan configuration - the giant air cooler delivers a 10 per cent improvement over the 360mm radiator in our setup.

Might the AIO need re-seating? We tried, with Noctua thermal paste no less (the H150i doesn't include any extra in the box), and still recorded an average temperature of 74.4ºC. We're of the opinion that the relatively large AMD chip isn't ideally suited to the AIO contact plate, and the Noctua's good-sized copper base ultimately fares better with the Ryzen die arrangement.

At stock settings, it is good to see that both coolers keep noise levels down to a minimum. The Noctua is fractionally quieter when idle (thanks in part to being used with the chassis' solid top panel) and the Corsair edges it under load. There's very little between the two, and neither can be considered intrusive.

Here's where things became interesting. Upping the ante in a simple manner, we chose to raise the multiplier to 43x on all cores, while increasing voltage to 1.3V, generating around 100W of extra power for the coolers to deal with. We know the chip can do it, and the bump in juice is enough to push the coolers to their limit.

As it turns out, that limit was quickly breached by the all-in-one liquid cooler, which failed to keep temperature the right side of 100ºC, resulting in the Asus motherboard forcing a system shutdown. No such problems for the Noctua, which remained stable throughout the Blender benchmark with temperature climbing to just under 86ºC.

Looking ahead, it will be interesting to see which coolers manage to pass the overclocking test, and we'll have some latest-generation liquid coolers installed inside the test bench in the very near future.

Is overclocking ill-advised on the Ryzen 9 3950X? That's a whole other debate. While the Noctua can manage 1.3V coursing through the chip, it has to ramp-up fan speed significantly, resulting in a noticeable shift in noise levels. Over 40dB can be deemed clearly audible and quite distracting in an otherwise quiet room.