ENERGETIC

Energetic = ENergy-aware hEteRoGenEous compuTIng at sCale

What’s the scope of the work?

There is a natural heterogeneity between UKRI DRI resources used in high performance computing due to variations in design choices as well as a heterogeneity designed into individual DRI resources due to the use of ‘accelerator’ special-purpose technologies (e.g graphics processing units [GPU] and field programmable gate array [FPGA] architectures). High performance computing is a significant contributor to energy usage. However, the energy-to-solution varies between these architectures. The ENERGETIC sandpit project set out to investigate whether the use of accelerators would give significant energy savings compared to CPU-only computations, and to determine which classes of code are most suited to accelerator technologies.

In this project, approaches to measuring energy usage on computer central processing units (CPUs) and associated accelerator technologies were investigated by a literature review and practical “hands-on” experimentation. Additionally a stakeholder workshop of around 25 attendees was held to discuss the matter of fair energy usage measurements across architectures.

The full report can be found here: DOI: 10.5281/zenodo.7692272

More information on the Energetic webpage

Representatives:
Michael Bane - Manchester Metropolitan University
Deepayan Bhowmik - Newcastle University
Oliver Brown - EPCC/University of Edinburgh
Jamie Quinn - University College London

Initial Abstract: DOI: 10.5281/zenodo.6787467

Final Report: DOI: 10.5281/zenodo.7692272