GPUs for research computing and AI

GPUs are specialised computer components designed for processing digital images and other parallel processing tasks. In a research context, GPUs are commonly used for machine learning and generative AI.

• GPUs can complete large and complex analyses faster than using CPUs alone.
• Certain software packages work best with (or require) GPUs.

• Not all software benefits from running on a GPU, e.g. software that uses a single core.

• Request a research virtual machine consultation with experts from the Centre for eResearch for advice and support relating to accessing and using GPUs.
• Nectar self-service virtual machines provide access to a specific GPU for a block of time. Current GPUs include NVIDIA A100s, P40s, T4s, and V100s.
• The REANNZ (previously NeSI) High Performance Computing (HPC) service provides access to GPUs.
• Research programming and workflow development includes the development of specialised software which makes use of GPUs.

• Introduction to deep learning workshop builds on skills learned in Programming with Python workshop and Introduction to machine learning workshop, to get started with applying neural networks to research applications.
• Introduction to the command line workshop offers important skills to navigate filesystems, run software, and automate repetitive tasks, to more efficiently use virtual machines and High Performance Computing services.

Research Virtual Machines Support Services
Email: research-vm@auckland.ac.nz

Chris Seal
Senior eResearch Solutions Specialist, Centre for eResearch
Email: c.seal@auckland.ac.nz