Close ad

When Apple introduced the iMac Pro last year, apart from the price, many people wondered how Apple would solve the cooling problem. All in one form factor is not a very ideal solution for cooling demanding components that will be under heavy load for a long time. The cooling limits of classic iMacs are a sufficient example. However, Apple has denied that the cooling in the new iMac Pros has been completely redesigned. It now includes two independent cooling circuits (CPU and GPU blocks). The fans and radiator are also new. They tested the updated cooling circuit on the Appleinsider server and found that it is definitely not without problems.

They summarized their detailed article in a video, which you can view below this paragraph. For testing, they used the "basic" configuration of the new iMac Pro, which has an 8-core Xeon (3,2GHz, 4,2GHz Boost), an AMD Vega 56 GPU, 32GB DDR4 RAM and a 1TB NVMe SSD. When idle, the new iMac Pro is completely silent. You will not know about it during normal work, which is not at all demanding on the components inside - i.e. browsing the web, some e-mails, etc.

Surprisingly, this state does not change even when 4K video was rendered in Final Cut Pro X on the tested model. Even under heavy load, the iMac Pro was very quiet, and even when the fans were running, there was no hum from the insides of the machine. Compared to a normal 5K iMac, this is said to be a huge difference. However, this "quiet operation" also has its drawbacks. As it seems, when designing cooling settings and fan cooling curves, Apple prefers low noise at the expense of cooling performance.

In the case of the classic Cinebench R15 CPU benchmark (achieved score of 1682 points), the processor reached a frequency of 3,9GHz. In each subsequent test, however, there was a temporary underclocking to 3,6GHz, due to a decrease in the temperature of the chip. The processor reached the limit of 94 degrees relatively quickly under load, after reaching which classic throttling occurs. These drops in frequency lasted for about two seconds, after which the processor rose again to 3,9. The more Cinebench was repeated, the more often the processor underclocked. So Apple has set the maximum speed of the fans due to the noise of the cooling, and the train does not go beyond that. Currently, it is not possible to set the performance curves of the cooling fans to your liking.

CPU throttling appeared again while editing the video. In this case, it took about three minutes for the CPU to reach 93-94 degrees. At that moment, the repeated frequency reduction from 3,9 to 3,6 GHz began. This behavior was repeated throughout the testing (in this case during 4K video rendering), which lasted about 7 minutes and the processor temperature was between 90 and 94 degrees.

The cooling system becomes loud when the GPU needs to be cooled in addition to the CPU. In case of load on both the processor and the graphics card, the cooling noise is at the same level as in the case of a classic 5K iMac. If the cooling system has to cool the graphics card as well, the processor will reach its limit temperature (94 degrees) much faster. Earlier this will lead to throttling and reduced performance. In the case of combined load, the processor starts to underclock to 3,3GHz and returns back to 3,6GHz. The frequency of 3,9GHz is unattainable with the combined load, at least with the default cooling. The graphics card reached 74 degrees in the tests, and the tests showed that there is underclocking and loss of performance even here when the system is under maximum load. This is roughly 10%.

Testing by Appleinsider pointed out a few things. First of all, it is clear that Apple prefers the silent operation of its devices, even if this means that the components work at extreme temperatures and are underclocked. A huge disadvantage is the impossibility of customizing cooling and creating custom curves and cooling profiles. As soon as this becomes possible, it will probably be reflected in the performance in practice. It should also be taken into account that some of the benchmarks in this stress test do not represent the real load that the iMac Pro will face. For example, Cinebench or the combination of CPU+GPU testing is only used for testing. On the other hand, I would expect the authors to also focus on a classic stress test in such a test. What would the processor frequency look like after two hours of load? Anyway, you can now get a fairly clear idea of ​​how the new iMac Pro performs in terms of its cooling performance.

Source: Appleinsider

.