llama.cpp Vulkan Intel Arc A770 16GB llama-2-7b Q4_0 pp512 FA
1073.85 tok/s (pp512)
Fields
id
benchmarks/llamacpp-vulkan-a770-llama27b-q40-fa-pp512Metric
prefill_tok_s
Value
1,073.85
Unit
tok/s (pp512)
Model
llama-2-7b.Q4_0.gguf (TheBloke, 3.56 GiB, 6.74B params)
Methodology
COMMUNITY-SCOREBOARD (llama.cpp discussion #10879, maintainer-curated Vulkan backend). llama-bench llama-2-7b.Q4_0.gguf (TheBloke, 3.56 GiB, 6.74B params), pp512 and tg128, batch 1, ngl=99 full offload, Vulkan, flash attention on. Build a69d54f. Chip Intel Arc A770. Per-request batch-1, not aggregate. Printed mean 1073.85 +/- 29.68.
Measured by
third_party
Reproducible
yes
notes
Flash-attention-on scoreboard row.
Notes
Flash-attention-on scoreboard row.