An Empirical-cum-Statistical Approach to Power-Performance Characterization of Concurrent GPU Kernels

TL;DR

This paper presents an empirical and statistical framework for analyzing power and performance in concurrent GPU kernels, introducing a benchmark suite and demonstrating significant energy savings through concurrency.

Contribution

It develops a multi-kernel throughput workload framework and benchmark suite, enabling systematic power-performance analysis of concurrent GPU workloads for exascale data centers.

Findings

Concurrency reduces energy consumption by up to 33%.

Power dissipation is not significantly affected by concurrency.

The benchmark suite captures diverse kernel workload behaviors.

Abstract

Growing deployment of power and energy efficient throughput accelerators (GPU) in data centers demands enhancement of power-performance co-optimization capabilities of GPUs. Realization of exascale computing using accelerators requires further improvements in power efficiency. With hardwired kernel concurrency enablement in accelerators, inter- and intra-workload simultaneous kernels computation predicts increased throughput at lower energy budget. To improve Performance-per-Watt metric of the architectures, a systematic empirical study of real-world throughput workloads (with concurrent kernel execution) is required. To this end, we propose a multi-kernel throughput workload generation framework that will facilitate aggressive energy and performance management of exascale data centers and will stimulate synergistic power-performance co-optimization of throughput architectures. Also, we demonstrate a multi-kernel throughput benchmark suite based on the framework that encapsulates symmetric, asymmetric and co-existing (often appears together) kernel based workloads. On average, our analysis reveals that spatial and temporal concurrency within kernel execution in throughput architectures saves energy consumption by 32%, 26% and 33% in GTX470, Tesla M2050 and Tesla K20 across 12 benchmarks. Concurrency and enhanced utilization are often correlated but do not imply significant deviation in power dissipation. Diversity analysis of proposed multi-kernels confirms characteristic variation and power-profile diversity within the suite. Besides, we explain several findings regarding power-performance co-optimization of concurrent throughput workloads.