Ciru Lab / NPU-only benchmark board

XDNA2 token telemetry

Local AMD Ryzen AI NPU measurements, presented beside public FastFlowLM and Procyon reference context. TOPS, prefill, and decode are kept as separate signals.

AMD XDNA 2 badge
NPU / FastFlowLM-NPU / tg128 smoke

Local NPU Scatter

Fastest Full TG128 Rows

Published Qwen3.5 Decode Ladder

Published Qwen3.5 Prefill Ladder

Readout

NPU Catalog Rows

ModelBackendStatusPP tok/sTG tok/sWall tok/sPromptOutput

Reference Benchmarks And Architecture Notes