This article is based on an article from the Japanese edition of Engadget and was created using the translation tool Deepl.
The Mac with Apple Silicon (which means Apple's own semiconductor) has been announced and is now shipping. Some of you may have gotten your hands on one as soon as this article is published on November 17.
Three series are now equipped with the Apple M1 processor, aka Apple Silicon. The MacBook Air, in particular, has replaced the Intel processor with the M1 processor entirely. The same goes for the lower-end models of the MacBook Pro and the Mac mini. In other words, the higher-end models of the Mac mini and MacBook Pro and above are still powered by Intel processors, but they have not been replaced by M1 processors.
What does this mean? I would like to start with the conclusions, leaving the results of the actual test for later.
Air's performance was better than expected
The most surprising thing about the announcement of Apple Silicon powered Macs was their use in a wide range of computer types, from the MacBook Air to the MacBook Pro to the Mac mini. Modern computers are dominated in design by the heat generated from the processor (SoC), and each product with the M1 has a previous model with the same basic mechanical design but with an Intel processor.
Among them, the MacBook Air completely replaces Intel with the M1, while the MacBook Pro and Mac mini still have models with Intel processors on top. The reasons for this will be discussed later, but in terms of processing processor heat, there is a big difference between the three products.
The new MacBook Air is probably designed for a thermal design (TDP, a measure of how much power a processor can consume, the higher the power, the better the performance) of around 10 watts. Incidentally, older models with Intel processors (Ice Lake) were designed for around 12-13 watts, I believe the switch to the M1 was made to eliminate the cooling fan, as the M1 can deliver performance at less than 10 watts.
You may remember that at the launch event, Apple touted the M1 as being twice as powerful at 10 watts compared to Intel's latest mobile chips (specifically the Core i7-1165G7) and capable of delivering the same performance as the full power of the Core i7-1165G7 at 25% less power consumption. In fact, the M1's performance is around 1750 for a single-core and 7600 for a multi-core, according to GeekBench 5. Even more practical benchmarks such as Cinebench R23 show higher performance than the Core i7-1165G7 at 28 watts, which is nearly three times the TDP.
That said, there were concerns that a fanless design would quickly build up heat and become "hot", which would degrade performance, but it sticks really well.
The unexpectedly small difference between Pro and mini
Meanwhile, in the MacBook Pro, the M1 replaced the processor in the lower model, which is equipped with only two ports of Thunderbolt 3. The enclosure design is also similar to the previous model, with no large slits on the bottom left and right sides and a 15-watt thermal design instead of 28 watts, in keeping with the Intel-era design. In the case of the Mac mini, the thermal design standard is 65 watts, as it also supports the six-core Coffee Lake-H. That's a lot of room to spare, isn't it?
So many people, including me, thought that the performance of the M1 would vary greatly depending on the heat treatment of the installed system, even though it has the same specifications. In reality, however, the performance of the MacBook Air was so high that the difference between the MacBook Pro and the M1 was so small that it had to be conditionally tested to get a clear figure. And the performance difference between the MacBook Pro and the Mac mini was also negligible.
And the difference in performance between the MacBook Pro and the Mac mini was also negligible: despite being capable of a 65-watt processor, the performance difference between the MacBook Pro and the Mac mini, which also has the same M1, is so small that it's safe to say they're exactly the same. Or, if all the CPUs and GPUs are working at close to 100% load at the same time, the Mac mini may perform better, but we didn't find any applications that made a difference during the test.
So even if it could dissipate more than 15 watts worth of heat, you can't expect a performance gain to match the cooling performance. The upper limit of the M1's clock frequency is 3.2 GHz (the lower limit is unknown), but 10 watts or so will get most of the performance out of it, and with a thermal design of 15-20 watts, the M1's performance can be almost fully exploited.
In conclusion, the difference in performance between MacBook Pro and Mac mini is "almost zero", and the difference is that the cooling fan is only "almost inaudible" when the load is increased. In fact, the temperature of the enclosure barely rises even if you continue to put a high load on the Mac mini.
Whether it's 10 or 15 watts, it's almost the same performance
When we received our test unit of the Apple Silicon powered Mac, the first thing we did was benchmark it. While it doesn't quantify the actual application experience, it does help us to understand their positioning.
As we have already reported the results we tested in Geekbench 5 with Apple Silicon support, the three products show about the same values, albeit in different enclosures and with different cooling systems. This means that under a load of GeekBench level, even a MacBook Air with a fanless design will not raise the temperature to the point where performance is degraded.
So it's not quite the same, but just slightly different. The difference can be seen in Cinebench R23 for Apple Silicon. This benchmark only uses the CPU, but it uses the CPU almost to its fullest and has a long processing time, so it's easy to see the true power of the CPU.
However, Cinebench also does not make a difference in just one lap. This benchmark measures the speed of rendering high-definition images in 3D, but with the release of R23, it has been changed to record the value of "10 minutes of continuous operation" as a score. In other words, it is possible to measure how much or how little the performance degrades when the load is continuously high. It is also possible to choose "Do not repeat", "10 minutes" or "30 minutes" in the advanced settings.
If you only want to produce a single image, all three products, including the MacBook Air, score about the same (about 1470 on a single core and 7800 on a multi-core). However, the MacBook Air scores slightly lower in the 10-minute repetition test of CineBench.
Measured over 30 minutes, the MacBook Air's score drops to about 6600 in multiprocessor mode (still better than the Tiger Lake Core i7-1165G7 in 28-watt mode, about 4900). The hottest part of the unit was 40 degrees (room temperature 25 degrees) for about 10 minutes, but by 30 minutes, the temperature rose to 42 degrees. Probably about 42-44 degrees is the upper limit. This is because other MacBook models are also at this level of temperature when they are at their hottest.
On the other hand, the score of the MacBook Pro only dropped to 7450 after 30 minutes of continuous loading. The result is such that there is almost no performance drop. The temperature is also lower than the MacBook Air at 35 degrees at the hottest part of the top surface and 37 degrees at the bottom. By the way, the performance of the Mac mini does not change, no matter how many minutes it is run continuously.
Whether you consider this difference to be large or small, if you start benchmarking side-by-side at the same time, you'll see that the difference starts to show up after about the fourth lap of Cinebench rendering.
In other words, it is safe to say that there is no difference between the MacBook Air and MacBook Pro, except for processing that requires very long spurts. There is also a 13-inch version of the MacBook Pro with four Thunderbolt 3 ports still in the lineup, but aside from specs like SSD and maximum memory capacity, there is no longer any reason to choose it in terms of performance.
Real power glimpsed from the speed of real apps
The new products are also significantly improved in terms of GPU performance, but this time we're focusing on CPU performance. We tested exporting a 12-minute full HD/30P video to H.264 with Final Cut Pro, and all three were close to the Core i9 model of the 16-inch MacBook Pro. The 2018 Late model of MacBook Pro (the first 13-inch model to be quad-core), which I normally use, was completely outclassed in performance. The 16-inch MacBook Pro surpasses the M1 due to its Core i9-based model, but the standard model is equal or maybe the M1 is faster.
2-minute Full HD/30P video H.264 export test (Final Cut Pro)
Mac mini 2min 55sec
MacBook Pro 2 minutes 57 seconds
MacBook Air 2 minutes 57 seconds
MacBook Pro (2020Mid, i9 model, 16 inches) 2 minutes 23 seconds
MacBook Pro (2018Late top model, 13 inches) 11 minutes 7 seconds
And after processing, even the fanless MacBook Air was only slightly warmer; the Cofee Lake-H 8-core processor in the 16-inch model has the cooling fan spinning hard. Of course, the manufacturing process is completely different and therefore incomparable, but I was surprised that this kind of processing didn't diminish the processing power of the MacBook Air.
Next, to check the speed of applications developed for the Intel CPU, I developed 20 RAW files taken with the Ricoh GR III in Adobe's Lightroom and exported them as JPEGs. Here, all M1-equipped models produced the same results.
Development test of 20 RAW files taken with a Ricoh GR III (Lightroom)
Mac with M1 22.5 seconds
MacBook Pro (2020Mid, i9 model, 16-inch) 20.3 seconds
MacBook Pro (2018Late top model, 13-inch) 36.4 seconds
Again, the 13-inch MacBook Pro model with Coffee Lake can't compete with M1 Macs. Rosetta2 emulation is said to cut about 30% of the performance, so if Lightroom supports Apple Silicon, we can expect an additional 30% performance improvement from here. Best of all, it's emulation, yet it's fast and consumes less power. The same chassis design makes the performance difference at the heart even more apparent.
Beyond "Performance Per Power"
Although we have focused our evaluation on CPU performance, the M1's GPU performance (for an internal GPU) is also extremely high, reaching a GeekBench 5 Compute score of 19,000. Considering that the external GPU, GeForce MX350, has a score of around 14,000, you can see that this is extremely high performance for an internal GPU. It also has a definite advantage over Intel's Xe architecture. Of course, the M1 has the advantage at the moment because it is a processor produced on TSMC's 5-nanometer process and integrates 1.8 billion transistors. It remains to be seen how far the M1 will be able to maintain its advantage when its rivals get the same tools.
However, the M1 is able to pull off more performance per power than ever before in both CPU and GPU. No, in fact, if you upscale an SoC for smartphones and re-arrange it for PCs, you might be able to get every processor to perform as well per power as the M1. In any case, it would certainly outperform Intel's (plus we'll include AMD here) performance per power.
There is no lineup of Macs that use the 11th-generation Intel Core, so it's not a fair comparison. That said, all of the M1-powered Macs generate very little heat, and even the MacBook Air is basically an extremely "cool" computer, although it was the only time I was running Cinebench that the temperature topped 40 degrees.
The M1 not only has a higher CPU performance than Intel's more power-hungry quad-core, but also outperforms its larger, six-core processor for PCs. It never felt "hot" throughout our testing, and there is no doubt that the claim of three times better performance per power than Intel's mobile processors.
Of the three series we evaluated, the MacBook Air is still the material that best brings out the features of the M1. Of course, the MacBook Pro is also excellent and, under some conditions, more capable than the Macbook Air. The on-board battery capacity is also large, and whether the 30,000 yen price difference is acceptable or not will depend on how you use it. However, the question of "what to do next" remains a theme.
"Scalability" to be the challenge of the future
This time, I did not expect the difference in performance between the three series to be so small, although I did not expect it to be entirely unexpected. The M1 has only one specification, and all of them have an upper limit of the clock frequency of 3.2 GHz. However, it is not really necessary to have more than one spec because they all work at the appropriate clock frequency at the time and the frequency will be dropped if there are thermal problems.
However, the heat dissipation capacity of the Mac mini is completely overspecified. There is a point where there is a good balance between performance and power consumption around 10 to 15 watts, and beyond that, the performance doesn't increase much even if you take the power consumption out of the box. Or maybe they dared to provide a Mac mini with the M1 as a test environment for developers (Of course, it depends on the application, and the Mac mini's combination of quietness and performance will be preferred in situations such as music production, where fan noise is a concern).
But for users who need performance on a higher-end MacBook Pro, Mac mini, or iMac, or even Mac Pro-level performance, the M1 doesn't provide the answer in terms of scalability as well as performance. Of course, it will be upscaled over time, as Apple has said that it will take two years to transition. It is possible to imagine a scenario where the company will expand next year, unify all mobile systems with Apple Silicon and offer a desktop solution the year after, or cover up to the iMac next year and only replace the Mac Pro the year after that, but we don't have any solid information at the moment.
I'm also happy to report that compatibility with Rosetta 2 was extremely high. In our tests at hand, Epic Games' Fortnite (which is also in dispute) has had some unusual termination, but we've had very few compatibility issues, including with other games. You can also search for and download apps for iPad and iPhone in the Mac App Store (some apps that are available for the Mac don't show up in the search). Some apps may not work, such as those that use sensors that are not available on the Mac, but I have the impression that the compatibility is quite high when running apps for the iOS/iPad OS on Big Sur.
Unofficially, Apple Silicon applications are starting to appear at the moment, and I'm told that the major software vendors' Mac apps that were promised in the keynote speeches will start to support Apple Silicon next month.
As a person who knows firsthand about the batteries that do not diminish during testing and the enclosure that does not get hot under heavy load testing, both the MacBook Air and MacBook Pro look appealing. Moreover, I could feel the difference to the extent that I thought they were "fast" physically. It seems to me that the time to buy is now, especially for users who are aiming at the MacBook Air.
This article is based on an article from the Japanese edition of Engadget and was created using the translation tool Deepl. The Japanese edition of Engadget does not guarantee the accuracy or reliability of this article.