iPad + AMD R9700 eGPU Runs Qwen3.8-27B at 159 tok/s via Open-Source LSE Engine
TheOriginalG2 · reddit · 2026-10-09
LemonSeed Studio demos on-device inference on an iPad Pro with an AMD Radeon AI PRO R9700 in a Thunderbolt enclosure, running Qwen3.8-27B Q4 at 131k context. Decode hits 158.5 tok/s with DFlash2 speculative decoding (31.2 baseline), nearly identical across iPadOS/macOS/Linux, while Strix Halo tops out at 64.4. The LSE engine embeds unmodified Linux amdgpu/amdkfd drivers as a PCIDriverKit extension, records forward passes as graphs, fuses ops, and picks the fastest kernels by on-device measurement. Open source, with an OpenAI-compatible server, CLI, and C library.
More from Infra
- Self-hosted SearXNG behind 67 proxy chains, accessible only via tailnet — haydendevs · 2026-10-09
- GlobalFoundries Signs 5-Year TSMC Deal to Make AI Chip Interposers in New York — teortaxesTex · 2026-10-09
- NVIDIA pitches RTX Spark PCs for local fine-tuning, AI coding agents and deployment — danielhanchen · 2026-10-09
- Air Street GP: failed frontier labs are pivoting to data centers serving Chinese models — johncoogan · 2026-10-09
- Domingos: China zooms past America in data center capacity — pmddomingos · 2026-10-09
- Ocean compute may beat land data centers: model and dashboard test wave-powered sea compute — nwilliams030 · 2026-10-09