Apple unleashes a swarm of Mac Minions to share GPU resources for large AI tasks
The company unites resources from disparate macOS devices
Apple has unveiled a system called Minions to help its employees access GPU resources for AI tasks more efficiently. This distributed virtual GPU cloud system utilizes idle GPUs across Macs to execute large-scale machine learning inference workloads. Previously, some inference jobs faced queue times of over ten hours due to limited centralized GPU availability. However, with Minions, tasks can often start within minutes by utilizing idle macOS devices.
The recently released Mac mini and Mac Studio support clustering using RDMA over Thunderbolt 5, which combines compute and memory to accommodate larger AI workloads. Minions creates a virtual GPU cloud from connected Mac hardware, forming a fault-tolerant pool of inference workers in a shared distributed inference platform. This redistribution of idle resources on the network allows teams to make better use of GPU resources.
Minions consists of four main components: the Minion Server, which provides the coordination layer; the PostgreSQL Metadata Store, which maintains task metadata; the Minion Executor, which runs inference tasks on a Mac's GPU and uploads results to S3; and the Minion Watchdog, which monitors the process for failures and reassigns tasks if necessary. Database statistics are displayed on Grafana-based dashboards for real-time visibility.
Apple's focus on hardware availability and utilization differs from Nvidia clusters, which prioritize throughput, latency, and efficient GPU utilization for large models.
Written by urgent.news from The Register's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
