Cooling AMD EPYC With Noctua Coolers: NH-U9 TR4-SP3, NH-U12S TR4-SP3, NH-U14S TR4-SP3

[phoronix]

Phoronix: Cooling AMD EPYC With Noctua Coolers: NH-U9 TR4-SP3, NH-U12S TR4-SP3, NH-U14S TR4-SP3

While Austrian PC cooling manufacturer Noctua is mostly known for their high-end enthusiast/gamer oriented cooling products, they do offer a line-up of heatsinks for both AMD Ryzen Threadripper and EPYC server processors. In this article we are trying out the NH-U9 TR4-SP3, NH-U12S TR4-SP3, and NH-U14S TR4-SP3 heatsinks on a 32-core / 64-thread AMD EPYC platform for seeing how well these air coolers will do with AMD's promising server platform.

http://www.phoronix.com/vr.php?view=26107

[baka0815]

What about the loudness of those?

[GraysonPeddie]

I'm interested in noise levels as well. Even power consumption for that matter at idle.

[Michael]

What about the loudness of those?

All of the Noctua 92/120/140mm fans tested are extremely quiet. Whoops, forgot to point that out in greater detail but unfortunately don't have the means of providing any useful sound measurements.

[wujj123456]

I just got my EPYC 7401P and NH-U14S. The temperature is awesome, very similar to what published in article. I can really only hear the fan if my ear is less than half meter from the heatsink, and I use an open test bed.

[thelongdivider]

Why are the 2 fan versions performing worse?

[AdamOne]

Is Blender Game Engine doing multi-threading at the same level as the test in the article does?

[linuxgeex]

With the Blender 3D modeling software it's highly threaded and does make wonderful use of all 64 CPU threads. Still here the NH-U9 TR4-SP3 and its larger siblings had no troubles keeping up with cooling the EPYC processor.

It's a shame that you didn't benchmark it long enough to reach steady-state so that you could legitimately come to the conclusion you presented.



What this graph shows is a steady rise in temperature, which isn't at all a clear indication that the cooling solution is coping well.

[coder]

Why are the 2 fan versions performing worse?

The 2x 120 mm fan version is only worse than the single 140 mm fan. But always better than 1x 120 mm fan.

I wouldn't have guessed the single 140 mm would be quite so strong, but otherwise not surprising.

[coder]

Is Blender Game Engine doing multi-threading at the same level as the test in the article does?

I highly doubt it. At decent resolutions, with decent quality rendering options enabled, game engines tend to be GPU-limited, even on far less powerful CPUs.

In contrast, the non-realtime rendering benchmark he's using is doing rendering on the CPU, itself.