inference/research
verifiedmedium · 0.70updated 2026-10-15

Level1Techs forum — Intel Arc Pro B70 x4 vs RTX 3090 x4, vLLM Qwen3.5-27B 50-concurrent-request throughput + TTFT

forum

Fields

id
sources/level1techs-intel-b70-vs-rtx3090-vllm-50req
Publisher
Level1Techs forum
Kind
forum
Accessed
2026-10-15
notes
As relayed by hardware-corner.net B70 launch article: 4x B70 vLLM Qwen3.5-27B, 50 concurrent requests, ~1024-token context: ~369 tok/s sustained output throughput (peak ~550), avg TTFT ~11.4 s. Same-test 4x RTX 3090: ~348 tok/s, TTFT ~18.7 s. Forum measurement post, low-formal-methodology.

Notes

As relayed by hardware-corner.net B70 launch article: 4x B70 vLLM Qwen3.5-27B, 50 concurrent requests, ~1024-token context: ~369 tok/s sustained output throughput (peak ~550), avg TTFT ~11.4 s. Same-test 4x RTX 3090: ~348 tok/s, TTFT ~18.7 s. Forum measurement post, low-formal-methodology.

Cross-references

This record references no other records.