Xcode benchmark (Flutter) MBP 16" 2019 2.3GHz vs MacBook Air 16GB 1TB

[chelsel]

Here are some real world benchmarks comparing compiling a moderate size Flutter application on a MacBook Pro 16" vs the new MacBook Air 16GB.

TLDR; Seeing about 20% better performance on the Air over the MBP even while running under Rosetta. I don't know how performance on a much larger project would be impacted or if throttling will kick in, but it seems for average size Xcode projects you'll see a performance boost using the MacBook Air.

Here is my test data:

MacBook Pro 2019 16" 2.3GHz 16GB 1TB

Pages and Terminal running.
Execute when CPU is idle

Executing these commands:

flutter clean
flutter build ios

Running pod install... 13.6s
Running Xcode build...
└─Compiling, linking and signing... 17.2s
Xcode build done. 138.5s

Running pod install... 11.6s
Running Xcode build...
└─Compiling, linking and signing... 17.4s
Xcode build done. 139.8s

Running pod install... 11.8s
Running Xcode build...
└─Compiling, linking and signing... 17.8s
Xcode build done. 138.6s

Average build time: 138.97s (3 Run average)

MacBook Air M1 16GB 1TB

Pages and Terminal running.
Execute when CPU is idle
Terminal is running using Rosetta mode

Executing these commands:

flutter clean
flutter build ios

Running pod install... 11.2s
Running Xcode build...
└─Compiling, linking and signing... 12.0s
Xcode build done. 111.5s

Running pod install... 11.0s
Running Xcode build...
└─Compiling, linking and signing... 11.9s
Xcode build done. 111.5s

Running pod install... 11.4s
Running Xcode build...
└─Compiling, linking and signing... 11.8s
Xcode build done. 112.3s

Average build time: 111.77s (3 Run average)

I hope this comparison benefits someone!

 

[dmccloud]

The one thing I noticed about the two graphs you posted was that the Intel processor seems to spread the workload equally and evenly across all 16 threads, which is to be expected with the x86 approach. On the other hand, the M1 has a lot more variance across its 8 cores (seemingly unrelated to which cores are the performance cores and which ones are the efficiency cores.) The other takeaway is that the M1 was faster in all phases (pod install, compiling/linking/signing and overall build time). By any chance, did you have Activity Monitor up and running while running these tests to see how either CPU or memory usage was fluctuating during these tests?

 

[chelsel]

did you have Activity Monitor up and running while running these tests to see how either CPU or memory usage was fluctuating during these tests

I did and didn't see anything alarming but I did not capture the data.