llama.cpp CUDA RTX 4090 llama-2-7b Q4_0 pp512 noFA
11992.7 tok/s (pp512)
Fields
id
benchmarks/llamacpp-scoreboard-rtx-4090-nofa-pp512Metric
prefill_tok_s
Value
11,992.7
Unit
tok/s (pp512)
Model
llama-2-7b.Q4_0.gguf (ggml-org, 3.56 GiB, 6.74B params)
Methodology
COMMUNITY-SCOREBOARD (llama.cpp discussion #15013, maintainer-curated CUDA backend). llama-bench llama-2-7b.Q4_0.gguf (ggml-org, 3.56 GiB, 6.74B params), pp512 and tg128, n_gpu_layers 99, batch 1, full GPU offload, CUDA. Per-request batch-1, not an aggregate across concurrent requests. Flash attention off (-fa 0). Build 2241453, reporter @lhl. Printed mean 11992.7. Re-read of the discussion page on 2026-10-17.
Measured by
third_party
Reproducible
yes
notes
RTX 4090, 24 GB / GDDR6X / 384 bit, build 2241453, reporter @lhl. prompt processing cell.
Notes
RTX 4090, 24 GB / GDDR6X / 384 bit, build 2241453, reporter @lhl. prompt processing cell.