Home > Workload Solutions > Container Platforms > Red Hat OpenShift Container Platform > Archive > GPU-Accelerated Machine Learning with OpenShift Container Platform > Lab setups
The following table shows similarities and differences between the Dell lab setup and the NVIDIA setup against which we compare our results:
Table 1. Hardware and software stack details for MLPerf Training worker node
Component |
Specification |
|
Dell EMC system |
NVIDIA system (published) |
|
GPUs—memory |
4 x V100 GPUs—32 GB |
DGX-1 (8 x V100 GPUs)—32 GB |
Container orchestration |
OpenShift 4.3 |
Not applicable |
Operating system |
RHCOS 4.3 |
Ubuntu 18.04 |
Framework |
PyTorch NVIDIA Release 19.05 |
PyTorch NVIDIA Release 19.05 |
NVIDIA CUDA and CUDA driver |
|
|
CUDNN |
CUDNN 7.6.0 |
CUDNN 7.6.0 |
NCCL |
NCCL 2.4.6 |
NCCL 2.4.6 |
MLPerf Training |
v0.6 |
v0.6 |
Prometheus |
Prometheus 2.14.0 |
Not applicable |
Grafana |
Grafana 6.4.3 |
Not applicable |