What is a Computer Cluster?
The concept of a cluster involves using multiple computers simultaneously to speed up the processing of complex tasks. In a YouTube video, Alex Ziskind took this on by linking five Mac mini M4s using Thunderbolt connections. Unlike traditional Ethernet networks (1 to 10 Gb/s), Thunderbolt can achieve speeds of 40 to 80 Gb/s, allowing for faster and smoother data transfer between devices.
Apple Silicon: Notable Advantages
The Apple Silicon architecture used in the Mac mini offers benefits over traditional approaches. Thanks to unified memory, the CPU and GPU share optimized access to the system’s memory, avoiding the limitations of standard graphics cards. Furthermore, a cluster of Mac minis consumes less energy than a PC with a high-end graphics card, making it a more economical solution.
To manage this cluster, Ziskind relies on MLX, an open-source project by Apple dedicated to machine learning research. Unlike Xgrid, a former Apple technology that was discontinued, MLX focuses on smaller clusters, more suitable for individuals or small businesses.
However, the performance achieved is not outstanding. For instance, a single Mac M4 Pro delivers better results than two combined Mac mini M4s in certain tasks, such as training machine learning models.
A Niche Solution
The Mac mini cluster finds its use in specific cases, such as processing large datasets or research requiring parallel resources. But this technology is complex to implement and not very accessible to the average user. It also presents specific challenges, such as thermal management and limitations of Thunderbolt ports, which can complicate the use of such a solution.
This approach is both intriguing and appealing, but it should be reserved for projects requiring very specific computing resources. For most users, a high-end Mac M4 Max or Pro will always be a better option.