English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Growing deployment of power and energy efficient throughput accelerators(GPU) in data centers demands enhancement of power-performance co-optimizationcapabilities of GPUs. Realization of exascale computing using acceleratorsrequires further improvements in power efficiency. With hardwired kernelconcurrency enablement in accelerators, inter- and intra-workload simultaneouskernels computation predicts increased throughput at lower energy budget. Toimprove Performance-per-Watt metric of the architectures, a systematicempirical study of real-world throughput workloads (with concurrent kernelexecution) is required. To this end, we propose a multi-kernel throughputworkload generation framework that will facilitate aggressive energy andperformance management of exascale data centers and will stimulate synergisticpower-performance co-optimization of throughput architectures. Also, wedemonstrate a multi-kernel throughput benchmark suite based on the frameworkthat encapsulates symmetric, asymmetric and co-existing (often appearstogether) kernel based workloads. On average, our analysis reveals that spatialand temporal concurrency within kernel execution in throughput architecturessaves energy consumption by 32%, 26% and 33% in GTX470, Tesla M2050 and TeslaK20 across 12 benchmarks. Concurrency and enhanced utilization are oftencorrelated but do not imply significant deviation in power dissipation.Diversity analysis of proposed multi-kernels confirms characteristic variationand power-profile diversity within the suite. Besides, we explain severalfindings regarding power-performance co-optimization of concurrent throughputworkloads.

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