Home > AI Solutions > Artificial Intelligence > White Papers > Performance of Dell Servers Running with NVIDIA Accelerators on MLPerf™ Training v2.0 > Multi-node
Because training on multiple nodes is important, most of our results are submitted on multiple PowerEdge XE8545 nodes with ResNet/BERT/Mask R-CNN models. These results scale from 1/2/4/8/16 to 32 nodes, which helps you understand cluster scalability better. Because each PowerEdge XE8545 server houses four NVIDIA A100 accelerators, we see the multicard, multinode scaling runs from 4/8/16/32 to 128 NVIDIA A100 accelerators.