MoE-CAP: Benchmarking Cost, Accuracy and Performance of Sparse Mixture-of-Experts Systems

Abstract

The sparse Mixture-of-Experts (MoE) architecture is increasingly favored for scaling Large Language Models (LLMs) efficiently, but it depends on heterogeneous compute and memory resources. These factors jointly affect system Cost, Accuracy, and Performance (CAP), making trade-offs inevitable. Existing benchmarks often fail to capture these trade-offs accurately, complicating practical deployment decisions. To address this, we introduce MoE-CAP, a benchmark specifically designed for MoE systems. Our analysis reveals that achieving an optimal balance across CAP is difficult with current hardware; MoE systems typically optimize two of the three dimensions at the expense of the third-a dynamic we term the MoE-CAP trade-off. To visualize this, we propose the CAP Radar Diagram. We further introduce sparsity-aware performance metrics-Sparse Memory Bandwidth Utilization (S-MBU) and Sparse Model FLOPS Utilization (S-MFU)-to enable accurate performance benchmarking of MoE systems across diverse hardware platforms and deployment scenarios.

Publication
In *Annual Conference on Neural Information Processing Systems (NeurIPS'25) *
Yao Fu
PhD Student
Yeqi Huang
PhD Student
Zhan Lu
PhD Student
Leyang Xue
PhD Student (Primary supervisor Mahesh Marina)
Congjie He
PhD Student
Man-Kit Sit
PhD Student
Dayou Du
PhD Student (Primary supervisor Jianyi Cheng)
Tairan Xu
PhD Student
Luo Mai
Luo Mai
Associate Professor

My research interests include computer systems, machine learning systems and data management.

Related