Decode throughput
Gemma 4 12B · Itinerary · token/s
Mean and min–max range. Select a point to inspect its run and evidence.
Intel Arc Pro B50n = 3 independent runs
19.46SD 0.02
RX 9060 XT custom blowern = 3 independent runs
35.46SD 0.38
Updated 09/09, 18:22 (UTC+8)
Blower GPU tests
AI inference
Grouped by model and workload. Read decode, prefill and latency separately, retaining each result's evidence level.
Intel Arc Pro B50
RX 9060 XT custom blower
Search this category only. Menus and charts update as you type.
Descriptive comparison under recorded conditions