A workload whose speed is limited by how fast data can move through a link, not by how fast the accelerator can compute.
Continue to AI University →