inference/research
verifiedmedium · 0.70updated 2026-10-15

kyuz0 Strix Halo 8060S gemma-4-31B-it-UD-Q4_K_XL Vulkan RADV pp512

301.53 tok/s

Fields

id
benchmarks/kyuz0-8060s-gemma-4-31b-it-ud-q4-k-xl-vulkanradv-pp512
Metric
prefill_tok_s
Value
301.53
Unit
tok/s
Model
gemma-4-31B-it-UD-Q4_K_XL
Methodology
kyuz0 amd-strix-halo-toolboxes llama-bench sweep (results.json 2026-05-16, Fedora 43 kernel 6.19.12, llama.cpp b9187/9193). Vulkan RADV, -fa 1 -ngl 99 --mmap 0, default ctx. pp512 batch-1 per-request.
Measured by
third_party
Reproducible
yes

Cross-references