• Home
  • Article
  • The results of Apple M1 Pro and M1 Max exceeding Google Colab by 54% are recommended to your article.

The results of Apple M1 Pro and M1 Max exceeding Google Colab by 54% are recommended to your article.

The author Daniel Bourke is a machine learning engineer living in Australia. His published articles on various machine learning are also popular in Medium, some of which have been introduced on AINOW (see the official website for more information about him). In an article published on Medium, "Apple's M1 Pro and M1 Max outperformed Google Colab by up to 54%," he reported on MacBook Pro's machine learning development performance benchmark, including M1 Pro and M1 Max.

In October 2021, Apple released M1 Pro and M1 Max, whose M1 chip is equipped with machine learning model development agency "Apple Neural Engine". Bourke has conducted benchmark tests on M1 chips and similar tests on M1 Pro and M1 Max. The contents of the tests and their results can be summarized in the following table.

Benchmark tests for M1 Pro and M1 Max implemented by Bourke

Test content:

Test results:

さまざまなサイズとエンコーディングの動画の書き出しProResエンコーディングの場合、M1 ProとM1 Maxは短時間で書き出す
機械学習モデル生成機能CreateML使用時の訓練時間M1 ProとM1 Maxは短時間で訓練完了。M1 ProとM1 Maxの性能差は少ない。
TensorFlowを用いた小規模モデルの訓練時間M1チップシリーズで大きな性能差はない。
TensorFlowを用いた大規模モデルの訓練時間M1 ProとM1 Maxは短時間で訓練完了し、Google Colabより早かった。M1 Maxがもっとも早い。

According to the above benchmark results, from the point of view of machine learning model development, Bourke evaluates the MacBook Pro series of M1 chips as follows.

Evaluation of MacBook Pro Series on M1 Chip from the Development of Machine Learning Model

By the way, Mr. Bourke is considering replacing the 13-inch M1 MacBook Pro currently in use with the 14-inch M1 Pro MacBook Pro.

In addition, the following articles are translated by direct contact with Mr. Daniel Bourke after obtaining the translation permission. In addition, the content of the translated article is his point of view and does not represent specific countries, regions and organizations, nor does it represent the principled propositions of the translator and the AINOW editorial department.