Wednesday, Dec 16, 2020, 16:11 Hardware

Breaking in the new MacBook Air M1 — how well does it handle a continuous load?

After having already tested the new Mac mini’s chip and performance, it’s now the MacBook Air’s turn, we’ll be testing the version with 8 GPU/CPU cores. Connoisseurs already know that the MacBook Air is the only M1 Mac with two GPU/CPU versions. The two available M1 MacBook Airs consist of the base version, with 7 GPU/CPU cores, and the optional “performance version,” which brings 8 GPU/CPU cores to the table—costing 250$ more, harkening upon the so-called “binning.” In addition to the 2 current versions and in order to minimize waste during production, chips with defective cores will still be installed and sold in devices at a downmarket price, rather than landing in the trash.



MacBook Air M1 vs. Mac Pro and other Macs
In order to reproduce a realistic usage-scenario, we used our own application Logoist as a benchmark tool to measure various aspects of the performance. Performance was more than sufficient during simultaneous CPU and GPU core stress from triangulation, as well as geometric rendering, drawing of shapes, and the execution of shader code. The apparent result: As expected, even the compact MacBook Air manages to out-perform all other Intel Macs, with the passively cooled notebook leaving even an iMac Pro with ten xeon cores and monstrous cooling in the dust.

Above: the entire test, below: the CoreML Benchmark (Sekunden = seconds)

But how will it perform under heavy demand?
With the MacBook Air, however, we’re more interested in a different kind of benchmark. Apple themselves have stated that the true performance difference lies in the manner of cooling. The MacBook Pro and Mac mini can continually perform well under load, with their fans eventually sorting for additional heat dissipation. Due to the MacBook Air’s compact design, this isn’t possible—the result? In contrast to actively cooled Macs, the MacBook Air throttles its chip in order to ensure that the device’s temperature doesn’t rise to an unsafe level. The next few benchmarks will show whether or not the MacBook Air takes a performance hit as a result. What we’re looking out for is how long the MacBook Air needs to run for this “chip-throttling” to occur (our benchmarks above consisted of ten “passes”).

Chip performance sinks continually when under maximum stress (x-axis = seconds per pass, y-axis = total time passed)

Heat Build-up and Throttling
Even after 20 minutes of intensive testing our MacBook Air was at the most warm to the touch—with the upper housing barely reaching a temperature of 100 degrees. In comparison to MacBook Pro 2014, which began shooting its fans off quite heavily after 2 minutes, where the upper housing managed to reach a temperature of more than 140 degrees.

Heat could be felt throughout the entire housing of the MacBook Air after 30 minutes, but the ultrabook continued to function at a high level of performance, suffering only a 15 percent performance hit.

The last benchmark score, before we ended the test after 47 minutes: The housing remained uniformly warm, but not hot. The MacBook Air meets the promised computational performance and quite well for a passively cooled silent notebook. The chip was only rarely, if ever, throttled at a value of more than 25% and only in the most preposterous situations for which barely anyone would purchase an ultrabook for. Thus, the MacBook Air more than meets the performance standard of almost every other Intel Mac.

How the performance curve of a MacBook Pro with an Intel Chip looks:
Lastly, the comparison to the MacBook Pro 2014 with i7 QuadCore (2,2 GHz)

The same test with a MacBook Pro (x-axis = seconds per pass, y-axis = total time passed)

Something that was immediately noticeable during the Intel chipped MacBook Pro’s test run: The processor became hot extremely quickly, resulting in a serious dip in performance. The initial performance value proved unreachable even with maximum cooling—thus, the performance remained consistently below the starting value after 5 minutes.

M1 — Active vs. Passive Cooling
Our test also documented the performance difference between actively and passively cooled M1 Macs. The difference, although not noticeable during the first test run, makes itself apparent after consistent longterm load. This is exactly what Senior Vice President of Software Engineering, Craig Federighi, referred to as “thermal capacity,” which differs between the MacBook Air (which places a throttle on the chip to passively “cool”) and the MacBook Pro (which provides continuous consistent performance thanks to “active” cooling).

More articles you might enjoy to read: