https://www.localxpu.com/Notes from running local LLMs on Intel Arc cards: two Arc Pro B60s in a desktop, with the setup, the numbers and what broke.en-us https://www.localxpu.com/articles/serving-27b-two-b60s/https://www.localxpu.com/articles/serving-27b-two-b60s/Thu, 01 Oct 2026 12:00:00 GMTTwo B60s in a normal desktop now serve Qwen3.8-27B at about 53 tokens/s for one user and 250 across eight, after a graph-capture change, a P2P kernel module and a oneCCL memory leak.