llama.cpp CUDA on NVIDIA A10G (A10 AWS variant): Llama-2-7B Q4_0 noFA tg128 t/s (#15013 comment)
108.29 tok/s (tg128, per-request, batch 1)
Fields
id
benchmarks/llamacpp-cuda-a10g-llama27b-q40-nofa-tg128Metric
decode_tok_s
Value
108.29
Unit
tok/s (tg128, per-request, batch 1)
Model
llama-2-7b.Q4_0.gguf (TheBloke, 3.56 GiB, 6.74B params)
Methodology
COMMUNITY-SCOREBOARD COMMENT (llama.cpp discussion #15013, reported row, NOT on the maintainer-curated board). llama-bench -m llama-2-7b.Q4_0.gguf -ngl 99 -fa 0. Build commit 571d0d5 (b10068), reporter @Hastwell. Per-request batch-1. Device: NVIDIA A10G (AWS custom A10 variant, sm_86, 24GB GDDR6, 300W, ~498 GB/s; VRAM reported 22502 MiB).
Measured by
third_party
Reproducible
yes
notes
A10G is the AWS-specific A10 variant (2-slot, 300W, same ~498GB/s). Comment-level row, not on the curated scoreboard. Complements the two existing FA-on A10G records.
Notes
A10G is the AWS-specific A10 variant (2-slot, 300W, same ~498GB/s). Comment-level row, not on the curated scoreboard. Complements the two existing FA-on A10G records.