11:02
2026-10-10
dev.to
large-language-models
On-Device Inference Debugging (Part 2): Threads on Big Cores, CPU at Full Clock — Still Slow
A developer benchmarking Qwen2.5-1.5B on a Pixel 4 (Snapdragon 855) ruled out thread scheduling and CPU frequency throttling as the dominant cause of the phone's 0.5 tok/s inference speed, finding tha…