llamaperf
Submit

Qwen3.6 35B (3B active)

on AMD Radeon 8060S integrated GPU (Strix Halo; system RAM unreported) · llama.cpp b9843 (86b94708f) / stock ROCm

Oct 13, 2026
Share
Throughput
1044.1 tokens/s prompt processing (prefill)Prompt processing measures input prefill speed; it does not tell you the time to first token.
Quant
UD-Q8_K_XL (GGUF)
KV cache
F16
Flash Attention
on

Summary

User reports Qwen3.6 35B-A3B at 1044.1 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, flash attention on, and one participating Radeon. The figure is the stock batch2048/ubatch512 run at depth 512, measured as 1044.087853 +/- 3.814818 t/s; generation throughput and TTFT were not measured.