LeaseWeb: Llama-2-7B Q4_K_M on dual EPYC 9334, TTFT ~4.8s (512-tok prompt)
4800.0 ms (time-to-first-token, 512-token prompt)
Fields
id
benchmarks/leaseweb-epyc9334x2-llama2-7b-q4-ttftMetric
ttft_ms
Value
4,800
Unit
ms (time-to-first-token, 512-token prompt)
Model
Llama-2-7B Q4_K_M
Methodology
THIRD-PARTY MEASURED (LeaseWeb blog, Fran Baigorri Hauen, 2026-04-05). HPE ProLiant DL385 Gen11, 2x EPYC 9334 (Zen4, 32c/64t each, 2.7GHz, 210W, 128MB L3/socket), 64GB DDR5 total (12-channel/socket). llama-bench (llama.cpp) local run + OpenLLM/llmperf API-level check. Test config: 512-token prompt, 128 tokens generated, 24 CPU threads for LLM rows. Values read off published charts/prose; CPU util 20-30%. Worst printed TTFT in the set (4.8s).
Measured by
third_party
Reproducible
no
notes
Printed upper bound.
Notes
Printed upper bound.