UBC ARC's Sockeye cluster provides shared high-performance computing resources, including CPUs, GPUs, and software tools, to support machine learning research. These resources enable researchers to develop, train, evaluate, and run machine learning models for applications ranging from image analysis and predictive modelling to genomics and molecular research. Machine learning is Sockeye's largest workload category, accounting for 42% of total GPU-hours.
Source: UBC ARC internal GPU usage analysis, April–August 2026.
Explore machine learning research capabilities available on Sockeye:
Note: Supported workloads depend on model size, computational requirements, GPU memory, and resource availability. Researchers are encouraged to consult ARC Support for guidance on computing resources and workflows best suited to their research needs.
Getting Started
To access machine learning computing resources on Sockeye, apply for a UBC ARC Sockeye allocation. Allocation requests must be submitted by a Principal Investigator (PI).
Apply for a Sockeye Allocation
Technical Resources
Explore technical documentation and guidance for accessing and using Sockeye.
Additional Computing Resources
Need additional computing capacity?
UBC ARC is modernizing its digital research infrastructure, with new computing resources expected to become available in 2027.
UBC researchers can access additional computing resources through the Digital Research Alliance of Canada, including larger allocations through its annual resource allocation competition. ARC can assist with applications. Researchers should review their data classification, as specific requirements may apply to certain research data classifications.