Comparison of the Usability of Apple M1 Processors for Various Machine Learning Tasks

Area of Science:

• Computer Science
• Machine Learning Engineering

Background:

• Machine learning research requires robust computational resources.
• Apple's transition to M1 series chips presents new hardware possibilities for on-device ML tasks.

Purpose of the Study:

• To evaluate the usability of different Apple MacBook Pro models for basic machine learning research.
• To compare the performance of Intel-based (i5) MacBooks against M1, M1 Pro, and M1 Max chip variants.

Main Methods:

• Developed a Swift script utilizing Apple's Create ML framework for training and evaluating four distinct machine learning models.
• Conducted performance tests across four MacBook Pro configurations: Intel (i5), M1, M1 Pro, and M1 Max.
• Measured training and evaluation times for each model and hardware configuration.

Main Results:

• M1 series MacBook Pros demonstrated superior performance compared to the Intel-based model.
• Specific hardware architectures of M1 chips showed significant benefits in machine learning task execution.
• Performance differences were quantified and presented through comparative tables.

Conclusions:

• M1, M1 Pro, and M1 Max MacBook Pros are highly suitable for basic machine learning research, offering substantial performance gains over Intel-based predecessors.
• The Create ML framework on macOS provides an efficient environment for on-device machine learning model development.
• Hardware architecture plays a critical role in optimizing machine learning research performance on laptops.