Puzzle 75B-A9B / ROCm native PP512/TG128

on AMD Strix Halo 128GB · llama.cpp / ROCm / isolated Puzzle port b10023

Oct 13, 2026

Throughput

22.7 t/s gen · 323.5 tokens/s prompt processing (prefill)Prompt processing measures input prefill speed; it does not tell you the time to first token.

Quant

Q4_K_M (GGUF)

Flash Attention

System RAM

128 GB

Summary

User reports Puzzle 75B-A9B at 22.7 t/s generation and 323.5 t/s prompt processing on an AMD Strix Halo 128GB system with a single Radeon 8060S iGPU. Setup is llama.cpp with the YanissAmz Q4_K_M quant, full offload, batch 2048, flash attention and 10 threads under ROCm. The figures are separate 512-token prompt and 128-token generation microbenchmarks, not one end-to-end request; a separate ROCm thread sweep reports 10 threads at 22.14 t/s.