MMODELYST
Papers/Instant GPU Efficiency Visibility at Fleet Scale
PAP

Instant GPU Efficiency Visibility at Fleet Scale

May 20, 2026

arXiv
Abstract

We present Overall FLOP Utilization (OFU), a hardware-level, precision-agnostic GPU efficiency metric for AI workloads on HPC systems, derived from two on-chip performance counters: Tensor Pipe Activity and SM clock frequency. OFU requires no application instrumentation and works across GPU generations and numeric precisions. We characterize five properties of the OFU approximation -- tile quantization, floating-point precision scaling, clock sampling noise, Tensor Core clock domains, and non-tensor undercounting -- through controlled GEMM experiments on H100 and GB200 across FP16, TF32, FP8, and NVFP4. After tile-quantization correction, OFU predicts application-level MFU to within <=2 percentage points. Against 608 production training jobs, OFU achieves r = 0.78 correlation with application-level MFU and surfaces two framework-level FLOPs miscalculations. Deployed across large-scale GPU fleets, OFU has detected a 2.5x efficiency regression and tracked precision-dependent utilization changes in mixed-precision pretraining. Our evaluation and operational experience suggest OFU is a practical, deployment-ready complement to application-level MFU for continuous fleet-wide efficiency monitoring.

Select text to highlight · click a highlight to remove · saved in this browser only
Authors
Connor Pedersen, Dong H. Ahn, Michel Migdal, Collin Neale, Nik Konyuchenko
Your notes (browser-local)
saved
arXiv:2605.20799