SPEChpc 2021 Benchmarks on Ice Lake and Sapphire Rapids Infiniband Clusters: A Performance and Energy Case Study

TL;DR

This study evaluates the performance, power, and energy characteristics of the SPEChpc 2021 benchmarks on Intel Ice Lake and Sapphire Rapids clusters, revealing bottlenecks and energy strategies relevant to HPC workloads.

Contribution

It provides a comprehensive performance and energy analysis of SPEChpc 2021 benchmarks on two modern Intel HPC clusters, highlighting key bottlenecks and energy efficiency insights.

Findings

Memory bandwidth is a common bottleneck.

Race-to-idle is effective for energy minimization.

Memory-bound codes benefit from Sapphire Rapids energy efficiency.

Abstract

In this work, fundamental performance, power, and energy characteristics of the full SPEChpc 2021 benchmark suite are assessed on two different clusters based on Intel Ice Lake and Sapphire Rapids CPUs using the MPI-only codes' variants. We use memory bandwidth, data volume, and scalability metrics in order to categorize the benchmarks and pinpoint relevant performance and scalability bottlenecks on the node and cluster levels. Common patterns such as memory bandwidth limitation, dominating communication and synchronization overhead, MPI serialization, superlinear scaling, and alignment issues could be identified, in isolation or in combination, showing that SPEChpc 2021 is representative of many HPC workloads. Power dissipation and energy measurements indicate that the modern Intel server CPUs have such a high idle power level that race-to-idle is the paramount strategy for energy to solution and energy-delay product minimization. On the chip level, only memory-bound code shows a clear advantage of Sapphire Rapids compared to Ice Lake in terms of energy to solution.