Apple has detailed a new system called Minions that uses the idle GPUs in Apple Silicon Macs to create a virtual GPU cloud. The system combines untapped processing power to handle complex AI workloads, particularly large-scale machine learning inference. Apple describes Minions as a dynamic virtual GPU cloud that aggregates idle Apple Silicon GPUs across Macs, and its researchers have already been using it in-house.
The underlying idea is simple. Macs have powerful GPUs that spend most of their time idle, especially during everyday computing tasks. Minions is designed to put that unused GPU capacity to work, turning underused Macs into a cluster of AI-ready machines. Both the Mac mini and Mac Studio support clustering through Remote Direct Memory Access (RDMA) over Thunderbolt 5 connections, which allows compute and memory from multiple machines to be combined on a single workload. Minions extends that concept.
Apple's documentation describes four key components. A Minions Server acts as the central coordinator, handling requests through a web interface and APIs and queuing jobs in a PostgreSQL metadata store. Minion Executors run on macOS devices, carry out subtasks on local Apple Silicon GPUs, and write results to storage. If a failure occurs, such as a connectivity problem or power loss, the Minion Watchdog re-queues the job on another machine. A dashboard provides real-time monitoring of the whole process.
Apple says its testing showed some inference jobs that previously took more than 10 hours now finish in minutes, without additional infrastructure and with an easy path to scale. For businesses, cutting processing time without buying new hardware is an obvious benefit, especially as prices continue to rise.
Source: AppleInsider · Summarized by HeadlinesBriefing