Apple MacBook Pro 14 2021 Laptop Review: The performance of the M1 Max is limited

Apple offers five different M1 versions for the MacBook Pro 14. The base model offers 8 CPU cores, all the other four models 10 CPU-cores. The biggest difference is the number of GPU cores, starting with 14-16 for the M1 Pro and 24 or 32 for the M1 Max. If you select the M1 max, you automatically have to pay about 500 Euros for the upgrade to 32 GB RAM as well and the model is about 1000 Euros more expensive than the entry-level version. The following table shows the five different configurations.

Our review unit with the M1 Max and 24-core GPU is the second fastest model you can get, but the CPU performance is identical to the M1 Pro models with 10 CPU cores. The only advantage of the M1 Max is the increased GPU performance and it can be configured with up to 64 GB RAM. Contrary to the review unit of the MacBook Pro 16 with the M1 Max, the MBP 14 does not offer an optional High-Performance mode.

We already mentioned that almost all M1 Pro and M1 Max SoCs have the same CPU part with 10 cores, only the entry-level model is equipped with a slower 8-core model. The 10 cores are distributed across three clusters: One efficiency cluster with 2 frugal cores (up to 2064 MHz) and two performance clusters with 4 fast cores each (up to 3228 MHz). We checked the performance values like the power consumption and clocks with the power metrics function and the performance is completely stable, both on battery as well as mains.

The results in the Cinebench R15 Multi loop are on par with the other M1 chips with 10 CPU cores and 28 % ahead of the base model with the 8-core M1 Pro. Cinebench R15, however, does not run natively on the M1 SoC and has to be emulated via Rosetta 2, which results in a higher power consumption of 33W for the CPU cores (40W package power) as well as a slightly reduced performance.

Cinebench R23 and Geekbench, which run natively on the M1 MacBook, are better for the comparison with modern AMD and Intel CPUs. We see a power consumption of 29W (34W package) in multi-core benchmarks and the results are very competitive. The performance is just slightly behind the new Ryzen 9 6900HS in the Asus ROG Zephyrus, which uses much more power. The same applies for the Core i7-12700H, which consumes up to 100W (55W sustained) in the case of the Alienware x14. This means neither the latest chips from AMD, nor Intel, can keep up with Apple in terms of efficiency.

Single-core benchmarks pretty much paint the same picture, and only the new Alder Lake Core i7 with the new performance cores can establish a lead. However, the consumption of the i7 is significantly higher at 25W (IA cores) compared to the M1 Max at ~4W (~9W package). More CPU benchmarks are available here.

It is hardly surprising that the subjective performance impression of the MacBook Pro 14 with the M1 Max processor is extremely good. All inputs are executed instantly and even demanding application like Adobe Photoshop or Premiere Pro are launched without noteworthy loading times. Emulated applications via Rosetta 2 work good as well, but the subjective performance is often noticeably worse.

We also have to differ in the synthetic benchmarks, since only CrossMark and obviously the browser benchmarks run natively on the M1 SoC. The browser test results in particular are very good and the MacBook Pro 14 is also ahead of the Surface Laptop Studio in Photoshop and Premiere Pro, even though we have to use the Intel versions of the apps for the benchmarks. You immediately notice that the Apple silicon versions run better and the real-world advantage should be even bigger than the benchmarks suggest.

While the M1 Pro is either available with 14 or 16 GPU cores, the M1 Max features 24 or 32 GPU cores. We observed a maximum power consumption of 44W for the GPU during our review of the MacBook Pro 16 with the 32-core M1 Max and now around 35W for the 24-core M1 Max. The maximum GPU clock was 1.267 MHz during our benchmarks and the MBP 14 is roughly 20 % slower than 32-core model and ~50 % faster than the entry-level version with the 14-core GPU, respectively.

This power consumption is similar to the Nvidia’s GeForce RTX 3050 Ti, which runs at 35-50W in the case of the Schenker Vision 14, for example. If we look at the 3DMark Wildlife Extreme score, however, the performance is more comparable to a fast 130W TGP version of the GeForce RTX 3060.

The performance is also completely stable, both under sustained workloads as well as on battery power. More GPU benchmarks are listed here.

Gaming is still a tricky subject. In addition to Apple’s Arcade collection, there are some native titles for the M1 (like World of Warcraft or farming Simulator 22). It is also possible to find some compatible (emulated) games via Steam or Epic Games, which were used for our benchmarks below. They once again show a performance deficit of around 20 % compared to the 32-core M1 Max. The software CrossOver is another possibility to use the full resources of the M1 SoC and it allows the installation of Windows apps like Steam including the games library. We had no problems to install Witcher 3 via Steam and it ran well with maximum settings (without Nvidia HariWorks), but we cannot offer our usual benchmark results due to the lack of fps monitoring. Games usually do not offer the typical 16:9 or even 16:10 resolutions and you will have to launch a game with an external 16:9 screen attached. You can then select the standard resolutions, even when you detach the external screen.

The two fans can reach the same speeds (left side: 5807, right side: 6241 rpm) and the same maximum fan noise of 49.5 dB(A) like the entry-level model, but the fan behavior has changed a bit since we reviewed the base version. The fans usually stay below 4000 rpm and up to 36 dB(A), which is a very quiet fan murmur. This is also the case in our stress test with native apps and we can only provoke the maximum fan speed with sudden changes of the load situation (when we stop the GPU load in the stress test and the CPU suddenly gets more power, for example). The fans are a bit louder with emulated apps, where the fans are often in the range between 40 up to 45 dB(A).

However, these are worst-case scenarios, and you will usually not hear more than a very quiet murmur. You also benefit from the very defensive fan curve, because the fans only start spinning as soon as the chip reaches 95 °C, which takes more than a minute at maximum CPU load, for example. You can expect around 30 dB(A) when you only stress the CPU, which was the case for Cinebench R23 Multi after three minutes. We did not hear any other electronic sounds like coil whine.

Surface temperatures and the distribution are very similar to the base model of the MacBook Pro 14. The laptop is usually very cool in everyday scenarios and you can even put it on your lap under load without any limitations, because the bottom side only warms up to 44 °C. The upper center area of the keyboard on the other hand warms up to almost 50 °C, which can be uncomfortable during typing. The power adapter tops out at ~46 °C, which is not critical.

We managed to get observe a maximum package power of 86-90W when we reviewed the larger MacBook Pro 16 with the 32-core M1 Max. The CPU performance of the smaller M1 Max is identical and the GPU consumes about 10W less according to our measurements, so we expected a package power of around 75-80W in combined workloads. This is not the case though and the performance is much more limited compared to the larger MBP 16.

Both the CPU and the GPU consume about 22W (54W package) at the beginning of our stress test, but this value drops to just 12W for the CPU (~2 GHz on all cores) after a couple of seconds, while the GPU maintains around 20W (~900 MHz) for a total package power of ~42W. This is comparable to the entry-level model of the MacBook Pro 14 with the M1 Pro, but the power distribution differs a bit (~20W CPU & 15W GPU). This means the M1 Max offers slightly more GPU performance during the stress test, but also less CPU performance and it is pretty far away from its potential performance. The optional energy-saver mode limits the package power to just 22W (5W CPU & 9W GPU).

All in all, this is a rather disappointing results since the fans certainly have plenty of headroom. If you pay a lot of money for the M1 Max, you will probably want to use the performance in combined workload scenarios, which is not possible here. The power consumption already hits 97W at the beginning of the stress test, which is the limit of the PSU. Apple does not offer a more powerful power adapter for the MacBook Pro 14, so we expect the MBP 14 with the 32-core M1 Max will show a similar behavior. Unfortunately, we could not test the 140W power adapter due to supple issues, we would have liked to check the performance with the more powerful PSU.

Our idle measurements show that the M1 Max consumes slightly more power than the base model of the MBP 14 with the M1 Pro. We record a maximum consumption of 97.7W for a short moment, which already seems to be limited by the 96W power adapter. The consumption levels off at around 76W after a couple of seconds. We would have liked to test the behavior with the 140W PSU, but we unfortunately could not get one for testing.