The MLPerf inference benchmark measures how fast a system can perform deep learning inference using a trained model in various deployment scenarios. The following figure represents the Offline and Server scenarios of the MLPerf Inference benchmark with an exponentially scaled y axis:

Figure 5.              Resnet50, SSD-Resnet34, RNN-T, BERT, DLRM Offline, and 3D UNET performance of the PowerEdge R750xa server

Key takeaways include:

• The PowerEdge R750xa server can run all models and meet the target accuracy requirements that MLPerf inference v1.0 defined.
• The PowerEdge R750xa server renders promising performance relative to other results.
• All the results in official table are verified by MLCommons™. As of the time of submission, PowerEdge R750xa server results are in line 44. 

Note: Due to time constraints, 3D-UNet and DLRM results were not submitted. Figure 5 includes unverified results of these benchmarks.