Cover Image

The Illusion of Power Capping in LLM Decode: A Phase-Aware Energy Characterisation Across Attention Architectures

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ma, Bole, Afzal, Ayesha, Eitzinger, Jan, Wellein, Gerhard
Format:	Preprint
Published:	2026
Subjects:	Distributed, Parallel, and Cluster Computing Artificial Intelligence Machine Learning Performance
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

GROMACS Unplugged: How Power Capping and Frequency Shapes Performance on GPUs
by: Afzal, Ayesha, et al.
Published: (2025)

Analytic Roofline Modeling and Energy Analysis of LULESH Proxy Application on Multi-Core Clusters
by: Afzal, Ayesha, et al.
Published: (2024)

Move the Query, Not the Cache: Characterizing Cross-Instance Latent Attention Redistribution Across GPU Fabrics
by: Ma, Bole, et al.
Published: (2026)

Wattlytics: A Web Platform for Co-Optimizing Performance, Energy, and TCO in HPC Clusters
by: Afzal, Ayesha, et al.
Published: (2026)

Diagnosing Overhead in Dispatch Operations: Cross-architecture Observatory
by: Ma, Bole, et al.
Published: (2026)

Microarchitectural comparison and in-core modeling of state-of-the-art CPUs: Grace, Sapphire Rapids, and Genoa
by: Laukemann, Jan, et al.
Published: (2024)

CloverLeaf on Intel Multi-Core CPUs: A Case Study in Write-Allocate Evasion
by: Laukemann, Jan, et al.
Published: (2023)

Irminsul: MLA-Native Position-Independent Caching for Agentic LLM Serving
by: Ma, Bole, et al.
Published: (2026)

Cache Blocking of Distributed-Memory Parallel Matrix Power Kernels
by: Lacey, Dane C., et al.
Published: (2024)

Leyline: KV Cache Directives for Agentic Inference
by: Ma, Bole, et al.
Published: (2026)

Alya towards Exascale: Optimal OpenACC Performance of the Navier-Stokes Finite Element Assembly on GPUs
by: Owen, Herbert, et al.
Published: (2024)

On the Challenges of Energy-Efficiency Analysis in HPC Systems: Evaluating Synthetic Benchmarks and Gromacs
by: Machado, Rafael Ravedutti Lucio, et al.
Published: (2025)

Energy-Aware Computing in the Year 2026
by: Tchakoute, Roblex Nana, et al.
Published: (2026)

RAPID-LLM: Resilience-Aware Performance analysis of Infrastructure for Distributed LLM Training and Inference
by: Karfakis, George, et al.
Published: (2025)

Fine-Grained Energy Prediction For Parallellized LLM Inference With PIE-P
by: Dutt, Anurag, et al.
Published: (2025)

THEAS: Efficient Power Management in Multi-Core CPUs via Cache-Aware Resource Scheduling
by: Muhammad, Said, et al.
Published: (2025)

LMDeploy Accelerates Mixed-Precision LLM Inference with TurboMind
by: Zhang, Li, et al.
Published: (2025)

GhostServe: A Lightweight Checkpointing System in the Shadow for Fault-Tolerant LLM Serving
by: Jayakody, Shakya, et al.
Published: (2026)

DUAL-BLADE: Dual-Path NVMe-Direct KV-Cache Offloading for Edge LLM Inference
by: Jeong, Bodon, et al.
Published: (2026)

Characterizing and Optimizing LLM Inference Workloads on CPU-GPU Coupled Architectures
by: Vellaisamy, Prabhu, et al.
Published: (2025)

Prompt-Aware Scheduling for Low-Latency LLM Serving
by: Tao, Yiheng, et al.
Published: (2025)

Scalable Systems and Software Architectures for High-Performance Computing on cloud platforms
by: Ramesh, Risshab Srinivas
Published: (2024)

On the Partitioning of GPU Power among Multi-Instances
by: Vamja, Tirth, et al.
Published: (2025)

Understanding Power Consumption Metric on Heterogeneous Memory Systems
by: Proaño, Andrès Rubio, et al.
Published: (2024)

Architecture Specific Generation of Large Scale Lattice Boltzmann Methods for Sparse Complex Geometries
by: Suffa, Philipp, et al.
Published: (2024)

Temporal Load Imbalance on Ondes3D Seismic Simulator for Different Multicore Architectures
by: Solórzano, Ana Luisa Veroneze, et al.
Published: (2024)

The Energy Cost of Execution-Idle in GPU Clusters
by: Lei, Yiran, et al.
Published: (2026)

AcceleratedKernels.jl: Cross-Architecture Parallel Algorithms from a Unified, Transpiled Codebase
by: Nicusan, Andrei-Leonard, et al.
Published: (2025)

"Two-Stagification": Job Dispatching in Large-Scale Clusters via a Two-Stage Architecture
by: Yildiz, Mert, et al.
Published: (2025)

SmartWatts: Self-Calibrating Software-Defined Power Meter for Containers
by: Fieni, Guillaume, et al.
Published: (2020)

EfiMon: A Process Analyser for Granular Power Consumption Prediction
by: León-Vega, Luis G., et al.
Published: (2024)

ProTrain: Efficient LLM Training via Memory-Aware Techniques
by: Yang, Hanmei, et al.
Published: (2024)

Communication-Aware Diffusion Load Balancing for Persistently Interacting Objects
by: Taylor, Maya, et al.
Published: (2026)

Minos: Systematically Classifying Performance and Power Characteristics of GPU Workloads on HPC Clusters
by: Jain, Rutwik, et al.
Published: (2026)

Unleashing the Power of Preemptive Priority-based Scheduling for Real-Time GPU Tasks
by: Wang, Yidi, et al.
Published: (2024)

Towards Portability at Scale: A Cross-Architecture Performance Evaluation of a GPU-enabled Shallow Water Solver
by: Villalobos, Johansell, et al.
Published: (2025)

KEET: Explaining Performance of GPU Kernels Using LLM Agents
by: Davis, Joshua H., et al.
Published: (2026)

Extracting Practical, Actionable Energy Insights from Supercomputer Telemetry and Logs
by: Cornelius, Melanie, et al.
Published: (2025)

FAILS: A Framework for Automated Collection and Analysis of LLM Service Incidents
by: Battaglini-Fischer, Sándor, et al.
Published: (2025)

Efficient allocation of image recognition and LLM tasks on multi-GPU system
by: Lawenda, Marcin, et al.
Published: (2025)