inference/research
verifiedmedium · 0.70updated 2026-10-14

Embedded LLM - Benchmarking LLM Inference on Intel Arc Pro B60 (4x B60 TP=4, vLLM vs LLM-Scaler)

benchmark

Fields

id
sources/embeddedllm-arc-pro-b60-vllm-llmscaler-benchmark
URL
https://embeddedllm.com/blog/benchmarking-llm-inference-intel-arc-pro-b60
Publisher
Embedded LLM (Sparkle-provided 4xB60 workstation)
Kind
benchmark
Accessed
2026-10-14
notes
Feb 14 2026. 4x Arc Pro B60 TP=4, Qwen3-VL-30B-A3B-Instruct, vLLM 0.12 vs Intel LLM-Scaler 1.2. Concurrency 16/32/64, 1K->1K and 1K->2K. TTFT + TPOT tables; ~1000 tok/s aggregate peak at c=64.

Notes

Feb 14 2026. 4x Arc Pro B60 TP=4, Qwen3-VL-30B-A3B-Instruct, vLLM 0.12 vs Intel LLM-Scaler 1.2. Concurrency 16/32/64, 1K->1K and 1K->2K. TTFT + TPOT tables; ~1000 tok/s aggregate peak at c=64.

Cross-references

This record references no other records.