kyuz0 Strix Halo 8060S Qwen3.6-35B-A3B-UD-Q8_K_XL ROCm 6.4.4 32k tg32
39.88 tok/s
Fields
id
benchmarks/kyuz0-8060s-qwen3-6-35b-a3b-ud-q8-k-xl-rocm644-32k-tg32Metric
decode_tok_s
Value
39.88
Unit
tok/s
Model
Qwen3.6-35B-A3B-UD-Q8_K_XL
Methodology
kyuz0 amd-strix-halo-toolboxes llama-bench sweep (results.json generated 2026-05-18, Fedora 43 kernel 6.19.12-200, llama.cpp build 1a68ec937 / b9193). ROCm 6.4.4, -fa 1 -ngl 99 --mmap 0. Long-context cell: context depth 32768 tokens, generation of 32 tokens. ONE request, batch size 1, so the number is per-request, not aggregate across concurrent requests. Mean of the repeated runs; std 0.0 tok/s. Model file 35.8 GiB, 34.66 B params, quant Q8_K_XL.
Measured by
third_party
Reproducible
yes
notes
Not in the earlier kyuz0 filing, which covered only the default-context pp512/tg128 cells. Same results.json, the longctx32768 and longctx65536 rows.
Notes
Not in the earlier kyuz0 filing, which covered only the default-context pp512/tg128 cells. Same results.json, the longctx32768 and longctx65536 rows.