Apple Silicon Macs Gain Official Support for External GPUs

Apple has officially authorized the use of external Graphics Processing Units (eGPUs) on Macs equipped with M-series silicon by approving a specialized driver extension developed by Tiny Corp [1]. This development allows users to connect external NVIDIA and AMD graphics cards to Apple Silicon Macs via Thunderbolt or USB4 ports without the need to disable System Integrity Protection (SIP) [3].

Key takeaways

• Apple has signed and approved a Tiny Corp driver that enables external GPU support for Arm-based Macs [4].
• The solution supports AMD RDNA3+ and NVIDIA Ampere+ series GPUs connected through Thunderbolt or USB4 interfaces [3].
• The driver is specifically optimized for AI workloads, such as training and running Large Language Models (LLMs), rather than gaming [1].
• Users no longer need to bypass system security protocols like SIP to utilize the hardware [3].
• The installation process remains technical, requiring users to compile the driver via Docker [4].

Enabling AI Workloads on Apple Silicon

The integration of third-party GPUs marks a significant shift for the Apple hardware ecosystem, which had effectively dropped eGPU support following the transition to proprietary silicon in 2020 [3]. Tiny Corp’s solution, known as TinyGPU, is built upon the tinygrad neural network framework and allows for communication between macOS and external discrete GPUs [3]. By providing an Apple-approved DriverKit extension, the software allows researchers to leverage high-performance external hardware for complex data processing and machine learning tasks [1].

While the approval simplifies the security requirements, the setup is not a traditional plug-and-play experience for the average consumer [4]. Users must follow a specific workflow that includes installing the driver extension, configuring the GPU compiler, and running inference through the tinygrad environment [3]. Furthermore, the current implementation lacks acceleration for direct video output through the eGPU, confirming that the technology is intended for computational research rather than general graphics or gaming applications [1].

Why it matters

This development addresses a long-standing limitation for developers who rely on the macOS environment but require the compute power of NVIDIA or AMD architectures for AI development [4]. By allowing these drivers through its DriverKit framework, Apple has signaled a willingness to support specialized third-party compute hardware, potentially reducing the need for researchers to rely solely on cloud-based compute resources [3]. While it remains unclear if this move will eventually lead to broader GPU support for non-AI tasks, it provides a functional path for the Mac to serve as a more capable node in heterogeneous AI development setups [3].