lhl 8060S Mistral-Small-3.1-24B-Instruct-2503-UD-Q4_K_XL ROCm tg1 decode
13.16 tok/s
Fields
id
benchmarks/lhl-8060s-mistral-small-31-24b-instruct-2503-ud-q4-rocm-fa-tg1Metric
decode_tok_s
Value
13.16
Unit
tok/s
Model
Mistral-Small-3.1-24B-Instruct-2503-UD-Q4_K_XL
Methodology
lhl strix-halo-testing llm-bench (github.com/lhl/strix-halo-testing, Mistral-Small-3.1-24B-Instruct-2503-UD-Q4_K_XL/results.jsonl). llama.cpp llama.cpp-hip, backend ROCm, -fa 1, ngl=99, tg1 batch-1 decode at depth 0 on Ryzen AI Max+ 395 (Radeon 8060S gfx1151). System RAM peak 26589MiB, avg power 47W, max temp 53.0C. Per-request batch-1.
Measured by
third_party
Reproducible
yes