R9700 1-GPU Vulkan RADV Qwen3.8-27B UD-Q4_K_XL depth-0 decode

Donato Capitella, local-llm-benchmarks.dev speed dataset (results.json, run run-ed5d7b8176368aa2, completed 2026-10-02). Protocol deterministic-prime-extend-completions-v1, runner local-llm-speed v1, measured at the API endpoint of an OpenAI-completions server, concurrency 1, 2 repeats after 1 warmup. Depth 0: no primed prefix (actual_cached_depth_tokens 0, recomputed_history_tokens 0), so the whole 2048-token prompt is computed. Model file Qwen3.8-27B-UD-Q4_K_XL.gguf from unsloth/Qwen3.8-27B-GGUF, KV cache type default, speculative decoding off. Single GPU (execution_label '1 GPU'). llama.cpp version 0.5.0-dev build 11347 commit 5fc4f3c8c. Vulkan RADV, container image r9700-llama-vulkan-radv, served with --cache-ram 0 (workaround for llama.cpp issue 19839, Vulkan prompt-cache serialization). Decode figure is client-observed generation throughput over the 256 generated tokens after the first, so TTFT is excluded. Per-request, not aggregate.

Same host and model file as the paired prefill/decode record; the two are one run split by phase. Not comparable with the repo's llama-bench pp512/tg128 rows for this card: different prompt length, different measurement point (API endpoint, TTFT excluded from decode), and a server config rather than llama-bench.