As an Amazon Associate I earn from qualifying purchases @AMAZON
2018 MacBook Pro: Sustained CPU Usage vs Performance
Related: 2017 MacBook Pro, 2018 MacBook Pro, GPU, laptop, MacBook, MacBook Pro, Macs, Photoshop, post processing, thermal throttling, Thunderbolt, video
MPG tested a fully-loaded Apple 2018 MacBook Pro 15-inch 2.9 GHz 32GB 2TB Radeon Pro 560X. The only better model is the same configuration with a 4TB SSD.
While testing the 2018 and 2017 (and 2016) MacBook Pro, a pattern was noticed in that performance declines with each iteration of some tests. For example, with repeated iterations of of the Photoshop sharpening test, the 2017 MacBook Pro was seen to decline in performance by up to 28% after as little as 30 seconds. No such pattern was seen on the iMac 5K or 2013 Mac Pro.
Investigation with the 2017 MacBook Pro determined that turns out that this is a GPU throttling behavior. Disabling the GPU resulted in identical performance when tasks use the CPU instead. This little secret is not mention in the “brag sheet” aka specifications—not so nice.
The question addressed here is whether and by how much the 2018 MacBook Pro declines in performance as compared to the 2017 MacBook Pro.
Degraded performance with sustained usage
Checked on macOS 10.13.6
Below are times for iterations of the Photoshop sharpening test, repeated 500 times.
At first the time is in the range of ~6 .3 seconds, but within 11 iterations it declines to ~7.1 seconds and (on average) stays that way, never again going as fast as initially.
This is a far better result than the 28% decline seen with the 2017 MacBook Pro. Since the decline is less than 10%, this performance is a major improvement over previous models in terms of sustained consistent performance.
Creating an air gap under the MacBook Pro and then using a fan to blow air under it did not result in any difference in behavior at an ambient temperature of about 75°F. What happens when after a year or so dust starts to build up inside the case? Performance is likely to suffer more and more over time, since the ability of the MacBook Pro to cool itself will decrease with time as a fine layer of dust builds up in the sealed clamshell design.
The Intel Power Gadget graph shows low CPU utilization of around 15%, showing that the sharpening is mostly done using the GPU.
Below, the 2017 MacBook Pro shows a much greater decline in performance, going from about 2.6 seconds to about 3.6 seconds (per-iteration runtime is faster than the 2018 MacBook Pro because a much larger test file was used).