White Papers

Deep Learning Performance: Scale-up vs Scale-out
Architectures & Technologies Dell EMC | Infrastructure Solutions Group
45
Figure 40. Multi-node training PowerEdge C4140-V100-SXM2- Configuration-K with IntelXeon4116 cpu,
Multi-node training PowerEdge C4140-V100-SXM2 Configuration-M with IntelXeon6148 cpu, versus
single-node training non Dell 8xV100-16GB-SXM2
In the Figure 40 we can see how the system C4140-V100-SXM2 Configuration-M outperforms in terms of
training time in different batch sizes compared the other systems.
7.4 Other Explored Aspects
This section shows the results of aspects explored during this project such as the hyper
parameter tuning, learning rate effect on single-node and multi-node mode, and critical kernels
executed in the TensorFlow benchmarks. These aspects could be subject of deeper study for
future projects.