Qwen3.6 27B
on AMD Radeon 8060S integrated GPU (Strix Halo; system RAM unreported) · llama.cpp b9843 (86b94708f) / stock ROCm
Summary
User reports Qwen3.6 27B at 359.4 t/s prompt processing on an AMD Ryzen AI Max+ 395 with Radeon 8060S graphics.
Setup is llama.cpp b9843 with UD-Q8_K_XL, f16 K/V cache and flash attention, one participating Radeon with ngl=999.
The figure is the stock batch2048/ubatch512 arm at depth 512, a PP-only synthetic run with no generation measured; the same arm falls to 124.740378 t/s at depth 100000.