· via Hacker News – Front Page (native)
Stela runs macOS 27 on x86-64 PCs with an arm64 translator and NVIDIA Metal driver
Stela v0.1 is a UEFI boot layer that models Apple silicon hardware, translates arm64e code for x86-64 CPUs, and ships a Metal driver for NVIDIA GPUs. macOS boots, but slowly, and the project warns reliability is low.
A project called Stela has surfaced on Hacker News's front page with an ambitious pitch: a general-purpose Apple Silicon layer for x86-64 PCs. According to the project's GitHub repository, Stela is a UEFI boot program that combines three components — an arm64-to-x86-64 code translator, a Metal graphics driver for NVIDIA GPUs, and a software model of Apple's custom silicon — so that current macOS can boot on non-Apple hardware.
The results, at version 0.1.0, are very early. The README states that macOS and most applications start, but that the system is slow and needs much more testing. Stela currently models a Mac mini with an M2 chip and runs macOS 27. A prominent warning notes that a user's current chances of it booting are low, with major hardware support and fixes still on the way.
Translating arm64e for Intel and AMD chips
Because Apple's modern Mac software targets the arm64e architecture, the heart of Stela is its translator. According to the README, it works as both an ahead-of-time and just-in-time compiler, covers more than 3,000 AArch64 instruction encodings, maps guest registers to host registers, and caches translated code. It runs on any x86-64-v2 CPU, from either Intel or AMD — hardware from roughly 2011 onward, per the project's requirements.
A Metal driver built on Mesa
Graphics are handled by a Metal driver supporting NVIDIA GTX 16 and RTX 20 series GPUs and newer. Rather than routing through an existing graphics stack, the driver compiles AIR, Metal's shader intermediate representation, directly to NIR, the intermediate representation used by the open-source Mesa graphics stack. The README says it supports the Metal3 and Apple7 GPU families.
Requirements and installation
The stated requirements are an x86-64-v2 CPU (about 2011 or newer), UEFI firmware with Secure Boot disabled, 8 GB of RAM, an NVIDIA GTX 16 or RTX 20 series GPU or newer, a USB stick of at least 8 GB, an empty GPT partition of at least 60 GB, and network access during installation, which downloads macOS.
Setup is handled by a helper tool: users download a release, run sudo ./stela setup on Linux or macOS or launch stela.exe on Windows, then choose the USB stick, the macOS version and the Mac model before confirming. This step erases the USB stick, and the PC subsequently boots from it. A separate update command can refresh sticks made with older versions of Stela.
Building from source requires Rust and LLVM with Clang 21 or newer. The first build fetches and compiles Linux, Mesa and LLVM sources, which the README cautions takes a long time.
Licensing and status
Stela contains no Apple software, the README notes, and users are responsible for following Apple's software license agreement and terms of use — placing the burden of licensing compliance on whoever runs it. The project's own code is licensed GPL-2.0-or-later, and development is coordinated through a Discord server.
Why it matters
Running macOS on commodity PC hardware has historically meant relying on the Intel builds of the operating system, a route that is closing as Apple completes its transition to its own silicon. Stela charts an entirely different path: model Apple's arm64 hardware in software, translate the CPU's instructions for the host machine, and translate the graphics API too.
That makes it a working, if rough, demonstration that a modern consumer operating system can be moved across CPU architectures purely through translation layers — conceptually the reverse of Apple's own Rosetta, which brings x86-64 code to its arm64 chips. At v0.1, with an explicit warning that boot success is unlikely, Stela is more engineering achievement than practical tool. But for anyone interested in binary translation, GPU API compatibility or the limits of what a UEFI boot layer can do, it is a project worth watching.
- #macos
- #apple-silicon
- #nvidia
- #metal
- #open-source
- #uefi