Beyond Bandwidth: SMT Scaling for LLM Inference on IBM POWER

Zenodo (CERN European Organization for Nuclear Research) 2026 Preprint Open access

Read full text
Authors: H Blanco
DOI: 10.5281/zenodo.18607857
Cited by: 0

Abstract

This paper presents a cross-generational characterization of Simultaneous Multi-Threading (SMT) scaling for Large Language Model (LLM) inference on IBM POWER processors. Testing POWER9, POWER10, and POWER11 with models from 1.1B to 7B parameters under strict single-core isolation.

Fields of Study & Subjects

Parallel Computing and Optimization Techniques Low-power high-performance VLSI design Real-Time Systems Scheduling

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
11 February 2026
Language
EN
Document type
Preprint