macuda
CUDA on an Apple Silicon Mac over Thunderbolt with no NVIDIA driver: a C userspace NVIDIA driver (libtinynv) plus cudart/cublas shims running unmodified llama.cpp and stable-diffusion.cpp on an RTX 5090 eGPU
Details
- External ID
- 1372373127
- Source
- GITHUB
- Company
- —
- Product
- macuda
- Website domain
- github.com
- Launched
- Sept. 16, 2026
- Cohort
- —
- Upvotes
- 30
- Upvotes percentile
- 0.7248911094030234
- Tags
- apple-silicon, cuda, egpu, gpu-driver, llama-cpp, macos, nvidia, thunderbolt, tinygrad
- Fetched at
- Sept. 20, 2026, 5:44 p.m.
- Updated at
- Sept. 20, 2026, 5:44 p.m.
Enrichment
- Theme
- gpu compute and acceleration tools
- Vertical
- Horizontal
- Function
- Model & infra
- Audience
- Developer
- AI stance
- AI feature
- Project type
- Hobby / open-source project
- Normalized one-liner
- cuda support for apple silicon egpus
- Manually corrected
- False
Could you build this?
No Writing a custom C userspace NVIDIA driver and CUDA runtime/cublas shims over Thunderbolt for Apple Silicon requires elite low-level reverse engineering, hardware architecture expertise, and GPU driver development.
What it would actually take: Requires deep reverse-engineering of NVIDIA's proprietary GPU command submission protocols, memory mapping, PCIe/Thunderbolt BAR management, and undocumented microarchitectural interfaces (e.g. Blackwell/Ada). The stack involves writing custom kernel/userspace drivers in C/C++ (libtinynv) interacting via macOS IOKit/PCIDriverKit, alongside reimplementing subsets of CUDA runtime (cudart) and BLAS interfaces. This demands specialized systems programming expertise in OS kernels, GPU hardware registers, and binary reverse engineering.
Competitors
Other products that read as similar to this one — 310 launches clear the similarity bar, closest 8 shown.
Attention rank: #93 of 311 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 322 days after the earliest competitor.
- mvcc · github · 2026-09-23 · 51 upvotes · similarity 0.65
- maclator · github · 2026-09-30 · 18 upvotes · similarity 0.56
- ppcosxkvm · github · 2026-09-27 · 47 upvotes · similarity 0.56
- RUDA · github · 2026-09-09 · 14 upvotes · similarity 0.55
- CUDA/graphics in QEMU-KVM VMs without passing the Nvidia card to them · hn · 2026-09-08 · 5 upvotes · similarity 0.54
- d4r · github · 2026-09-27 · 92 upvotes · similarity 0.54
- CUDA-for-AMD-Windows · github · 2026-09-13 · 230 upvotes · similarity 0.54
- laya-apple · github · 2026-09-22 · 12 upvotes · similarity 0.50
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a model & infra tool for Fintech yet.