A new project called Stela is offering PC owners a way to load macOS 27 on their own x86-64 hardware by presenting a virtual M2 Mac Mini to the operating system. The effort comes from developer Tyler Smith and has drawn swift attention from enthusiasts. Three releases have appeared since the first 0.1 version arrived in barely half a day.

Smith has cautioned that Stela is slow and not very stable, warnings that have not deterred the technically inclined from trying it. In the project's Discord server, he said the macOS image used with Stela is not modified at all, and that the software supplies the translation and hardware behavior underneath so macOS can go through the exact same restore process it would on a real Mac. The server has seen a substantial influx of users, many posting photos of their PCs near-booting macOS. Most questions have concerned support for particular hardware, though the excitement is palpable, according to the report.

Getting started requires downloading Stela from its release pages, providing a USB stick to boot the PC with, and supplying a copy of macOS. Stela translates AArch instructions into x86-64 using a combination of ahead-of-time and just-in-time compilation for the CPU, and it translates Apple's Metal graphics layer calls to Mesa's NIR instructions.

The hardware requirements are modest, calling only for an x86-64-v2 processor such as a 2008 Intel Core chip or a 2011-vintage AMD FX. Smith notes that AVX2 and AVX-512 are used when present. Users also need 8 GB of RAM, UEFI support, a GPT partition or an entire drive, and an Nvidia card from the 16xx series through present-day 5000 models.

Wi-Fi is not supported yet and Bluetooth is limited, but the list of working Ethernet controllers is extensive, covering models from Intel, Realtek, Broadcom, Aquantia and Atheros, as well as USB Ethernet sticks. Smith intends for Stela to eventually support all the hardware the Linux kernel does.

The report notes that pre-translating much of macOS to x86-64 is not completely insane to imagine with AI automation, though such an endeavor carries caveats. Much code, such as a browser's JIT JavaScript compilation, does not exist on the disk image to begin with. No matter how efficient a translation layer is, the operating system still needs to cryptographically accept it and run it, perhaps requiring further modification.

More hardware news from TechManNews.