PT tested two configurations of this solution: one with four GPUs and another with two GPUs. The four-GPU solution supported 40 simulated data science knowledge workers, while the two-GPU configuration supported 20. In both cases, the solution supported the hardware-maximum of 10 virtual users per GPU, and all users were able to complete a compute-intense machine learning inference workload (MLPerf Inference v1.0).
Data science knowledge workers often consume a large share of a company’s compute resources, so being able to customize a server’s GPU computing power can help organizations to meet their precise needs.
To learn more about how the Dell EMC PowerEdge R750xa with NVIDIA A100 PCIe GPUs can…