< Back to all clusters
[TECHNOLOGY] · 3 sources

started · updated

iPhone 17 Pro Max used to boost MacBook Pro AI performance

A developer has demonstrated a method to enhance the AI processing capabilities of a MacBook Pro M4 Pro by utilizing an iPhone 17 Pro Max as a secondary computational resource. Using an open-source software project called ‘backburner’ available on GitHub, the setup distributes the workload of running the Qwen 3.6 27B large language model across both devices.

The process involves the MacBook Pro executing the first 40 layers of each 256-token batch before transferring the data via a 10 Gb/s USB-C connection to the iPhone. The iPhone then uses its A19 Pro chip's GPU to process the remaining 24 layers (layers 41 to 64). This parallelization allows the MacBook to begin the next batch immediately, increasing prefill performance by up to 44% depending on the context size.

Beyond layer processing, the iPhone also assists with memory management. When the context window exceeds 64,000 tokens, the iPhone takes over the management of the KV cache and attention calculations for older data. Additionally, the iPhone's Neural Engine is utilized to optimize historical conversational context, helping to reduce latency during long sessions.

Entities

Apple · GitHub · MacBook Pro · Qwen · iPhone 17 Pro Max

Claims

What the coverage asserts, and how many sources carry each claim.