Evaluating On-Node GPU Interconnects for Deep Learning Workloads

Evaluating On-Node GPU Interconnects for Deep Learning Workloads