Question

A mid-sized company wants to purchase GPU servers from a major OEM vendor rather than buying NVIDIA-branded DGX systems, but the IT team is concerned about whether the hardware configuration will reliably run accelerated AI workloads and NVIDIA software. Which approach best addresses this concern while allowing the company to use their preferred OEM vendor?

Accepted Answer

Purchase servers that carry the NVIDIA-Certified Systems designation from the OEM vendor

Answer

Buy any OEM server with a supported GPU and install the CUDA toolkit manually to guarantee compatibility

Answer

Require the OEM to ship servers only from the DGX SuperPOD reference architecture

Answer

Select consumer-grade GPUs since they use the same CUDA cores as data center GPUs