Same budget: MacBook Air + 48GB VRAM desktop rig vs 64GB MacBook Pro for local LLMs
gappyvalley · reddit · 2026-09-09
A local-LLM user with a desktop running RTX 5080 + 5060 Ti (31GB usable VRAM, currently serving Qwen 27B Q6 and Qwen A35B) weighs two same-cost options: a MacBook Air M5 24GB plus an R9700 32GB to bring desktop VRAM to 48GB (with Tailscale remote access), or a 16-inch MacBook Pro M5 Pro with 64GB unified memory. The tradeoff is portable big-model capability versus a light laptop backed by a desktop AI rig.
More from Infra
- Nextcloud: The Open-Source Stack That Replaces Google Workspace for $0 Per User — JafarNajafov · 2026-09-09
- Whole-house AI: local Gemma 4 12B with XIAO S3 Sense sensor nodes — Paimaamu · 2026-09-09
- Google Pours $15B Into Finnish AI Infrastructure With 22-Year Nuclear Power Deal — mark_k · 2026-09-09
- Cyber Threats to Boost Zero Trust Network Adoption After Hugging Face-Style Incidents — bookwormengr · 2026-09-09
- CLI reads Apple Notes SQLite directly, 2500x faster than AppleScript-based integrations — autocar24 · 2026-09-09
- NVIDIA launches CUDA Rust: two tracks to write GPU kernels in Rust — ducha_aiki · 2026-09-09