verifiedmedium · 0.75updated 2026-10-14

Strix Halo 8060S (HIP (ROCm)) Qwen 3 32B Q8_0: pp512 prefill tok/s

226.1 tok/s (pp512, per-request batch-1)

Fields

benchmarks/strix-halo-8060s-qwen3-32b-q80-hip-pp512

Metric

prefill_tok_s

Value

226.1

Unit

tok/s (pp512, per-request batch-1)

Model

Qwen 3 32B

Methodology

THIRD-PARTY MEASURED (lhl/strix-halo-testing llm-bench summary table). Hardware: Framework Desktop, AMD Ryzen AI Max+ 395 (Strix Halo) + Radeon 8060S iGPU, 128GB LPDDR5x-8000 unified memory. llama.cpp ~b5863, HIP (ROCm) backend, flags hipBLASLt; llama-bench -fa 1 --mmap 0 -ngl 99, pp512 + tg128, per-request batch-1. Peak resident 33683 MiB. Numbers verbatim from the best-backend summary table.

Measured by

third_party

Reproducible

yes

Cross-references

Strix Halo 8060S (HIP (ROCm)) Qwen 3 32B Q8_0: pp512 prefill tok/s

Fields

Cross-references

Accelerators (2)

Engine (1)

Interconnect (1)