llama.cpp CUDA Titan V llama-2-7b Q4_0 tg128 FA
112.17 tok/s (tg128)
Fields
id
benchmarks/llamacpp-scoreboard-titan-v-fa-tg128Metric
decode_tok_s
Value
112.17
Unit
tok/s (tg128)
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 on (-fa 1). Build e56abd2, reporter @Hedede. Printed mean 112.17. Re-read of the discussion page on 2026-10-17.
Measured by
third_party
Reproducible
yes
notes
Titan V, 12 GB / HBM2 / 3072 bit, build e56abd2, reporter @Hedede. token generation cell.
Notes
Titan V, 12 GB / HBM2 / 3072 bit, build e56abd2, reporter @Hedede. token generation cell.