Preprint
Aug 2026
Evaluating MFU as a Proxy for GPU Power for Energy-Aware Simulation of LLM Training
This work evaluates whether Model FLOPs Utilization (MFU) can serve as a portable, software-defined predictor of GPU power for LLMs, finding that a linear MFU-based power model fits every tested GPU as long as the workload is compute-bound, as in production LLM training.
Niklas Enskat, Philipp Wiesner
· 1 citation
· ⚡1