Mac Studio M5 Max (64GB) vs 2x RTX 4060 Ti 16GB for local LLMs
retained text
I’m looking at the Mac Studio M5 Max (40-core GPU, 64GB). I currently have dual RTX 4060 Ti 16GB (32GB total) setup. If you run a Mac what prefill and decode speeds do you get? I currently get on the 2x 4060 Ti: |Model|Quant|Ctx|Prefill (t/s)|Decode (t/s)| |:-|:-|:-|:-|:-| |Qwen3.6-35B-A3B|UD-Q4\_K\_XL|262k|3060|122.5| |Qwen3.8-27B (dense)|UD-Q4\_K\_XL|200k|960|34.3|
document fetched 2026-10-30 00:44 · canonical https://reddit.com/r/LocalLLaMA/comments/1xio8j7/mac_studio_m5_max_64gb_vs_2x_rtx_4060_ti_16gb_for