Apple's Quiet Nod to Nvidia on Macs Opens a Door for AI Developers

A longstanding barrier between Apple hardware and Nvidia's industry-standard computing tools has just developed a crack. This week, Apple approved a third-party driver that allows recent Nvidia desktop GPUs to work with Apple Silicon Macs for the first time since the transition began.

The driver, called Miracle and developed by a company named Sapling, operates as a sanctioned System Extension. It lets Mac Studio or MacBook Pro models offload compute tasks to an external Nvidia graphics card via Thunderbolt. Crucially, it supports CUDA, Nvidia's parallel computing platform that underpins most machine learning development, scientific simulation, and advanced rendering workflows.

Apple's approval is notable. The company has spent years promoting its integrated M-series chips as a complete solution, making the decision to allow this bridge a significant, if silent, concession. The driver is compute-only, meaning it won't drive displays, and is aimed squarely at professional tasks like training neural networks or processing large datasets.

The timing aligns with immense pressure from the AI boom. Developers who create models often rely on CUDA, a reality that has forced many to maintain a separate Windows or Linux machine alongside their Mac. This driver offers a potential path to consolidate that setup, keeping developers within Apple's ecosystem.

Performance questions remain, as Thunderbolt bandwidth is a fraction of a direct PCIe connection. Yet for individual developers prototyping or fine-tuning models, even reduced performance may be a worthwhile trade-off for the convenience. This isn't a full corporate détente between Apple and Nvidia, but it is a pragmatic opening that addresses a real pain point for a key professional audience.

Source: Webpronews