llama.cpp ROCm on RX 9070 XT: Llama-2-7B Q4_0 FA-off pp512
5055.19 tok/s (pp512, per-request, batch 1)
Fields
id
benchmarks/llamacpp-rocm-9070xt-llama27b-q40-nofa-pp512Metric
prefill_tok_s
Value
5,055.19
Unit
tok/s (pp512, per-request, batch 1)
Model
llama-2-7b.Q4_0.gguf (TheBloke, 3.56 GiB, 6.74B params)
Methodology
COMMUNITY-SCOREBOARD (llama.cpp discussion #15021, maintainer-curated). llama-bench -m llama-2-7b.Q4_0.gguf -ngl 99 -fa 0. AMD Radeon RX 9070 XT (16 GB / GDDR6 / 256 bit). Build commit 583cb83, reporter @Hadrianneue. Mean +/- std (±109.58). Single-GPU run per thread rules. Per-request batch-1, NOT aggregate.
Measured by
third_party
Reproducible
yes
notes
Maintainer-verified scoreboard row from #15021. Identical model/test/flags across all GPUs. tg128 = decode, pp512 = prompt processing. RDNA4 prefill is strong - ahead of MI100 on compute-bound pp.
Notes
Maintainer-verified scoreboard row from #15021. Identical model/test/flags across all GPUs. tg128 = decode, pp512 = prompt processing. RDNA4 prefill is strong - ahead of MI100 on compute-bound pp.