AI/ML●●Solid
35B MoE LLM and other models locally on an old AMD crypto APU (BC250)
Kernel ttm.pages_limit workaround unlocks 16GB UMA for Vulkan inference on repurposed crypto hardware.
Dark HorseNiche Gem
akandr
202mo ago
Back to browse
GitHub Repository
AMD BC-250 (PS5 APU) setup guide — Ollama + Vulkan inference, poor man's AI assistant via Signal, stable-diffusion.cpp image generation
132 starsPython
Tok/s on a 35B MoE model using a $100 AMD crypto APU and Vulkan
by akandr·Mar 23, 2026·2 points·1 comment
AI Analysis
●MidNiche Gem
Clever hardware hack but this is a config guide, not a shipped tool.
Strengths
- •Gets serious 35B models running on $100 consumer hardware
- •Vulkan inference path works around CUDA dependency
Weaknesses
- •Documentation for existing tools, not novel software
- •35 GitHub stars suggests limited adoption or polish
Category
Target Audience
Hobbyists running local LLMs on budget hardware
Similar To
Ollama · llama.cpp · stable-diffusion.cpp
Similar Projects
AI/ML●●Solid
I ran Qwen3.5 35B on my iPhone at 5.6 tok/SEC
Runs 19.5GB Qwen3.5 on 12GB RAM iPhone via memory swapping.
WizardryBold Bet
alexintosh
422mo ago
AI/ML●●●Banger
NVFP4 on Desktop Blackwell – 122B MoE on a Single RTX PRO 6000 31 tok/s
Bypasses NVIDIA's artificial FP4 lock—122B MoE on single desktop GPU at 31 tok/s.
WizardryDark Horse
jcartu
202mo ago
Productivity●●●Banger
Vocalinux // 100% offline voice typing for Linux
Linux finally gets offline voice typing; Ctrl-tap + Vulkan GPU support vs cloud-dependent alternatives.
Solve My ProblemDark Horse
jatinkrmalik
403mo ago
AI/ML●●●Banger
SwiftLM – Qwen Chat on iPhone, 100B+ Moe on M5 Pro 64GB (Native Swift)
Native Swift inference with SSD streaming runs 100B MoE models without kernel panics.
WizardryNiche Gem
aegis_camera
122mo ago
AI/ML●Mid
Open Access Qwen3.6-35B-A3B-UD-Q5_K_M with TurboQuant
Temporary public endpoint for Qwen3.6-35B quant on a spot instance.
Ship It
freakynit
421mo ago