RTX PRO 5000 Blackwell vLLM: Qwen/Qwen3-8B aggregate output throughput, 1000 requests
2743.63 tok/s (aggregate output tokens, 1000 concurrent)
Fields
id
benchmarks/dm-pro5000-vllm-qwen3-8b-c1000-outMetric
tps_aggregate
Value
2,743.63
Unit
tok/s (aggregate output tokens, 1000 concurrent)
Model
qwen3-8b
Methodology
Database Mart vLLM bench on one RTX PRO 5000 Blackwell 48 GB GDDR7 ECC (Server Edition, not the 72 GB workstation SKU). Model Qwen/Qwen3-8B, quantization column '16' (bf16/fp16 not distinguished; format_id null, no fp16 record exists). vLLM commit not printed. Single GPU, tensor parallel 1. 1000 requests, input length 100 tokens, output length 600 tokens, benchmark duration 178.45 s, total input tokens 100000, total generated tokens 489611. AGGREGATE across the concurrent requests, not a per-request decode rate.
Measured by
third_party
Reproducible
yes
notes
Output (tokens/s) column. Aggregate across 1000 requests, not the per-request eval rate.
Notes
Output (tokens/s) column. Aggregate across 1000 requests, not the per-request eval rate.