Nvidia launches program for certified AI servers

Nvidia is offering a certification program to steer customers and partners to buy GPU servers specifically tailored for artificial intelligence (AI) workloads from top-tier OEMs and ODMs.

The program, called Nvidia-Certified Systems, is similar to those offered by Intel and AMD. It tells customers that these servers with these configurations are best optimized for running AI workloads at peak performance and in large-scale deployments.

Nvidia kicked off the program with 14 certified GPU servers from Hewlett Packard Enterprise, Dell Technologies, Gigabyte, Supermicro, and Inspur. All systems feature the A100 Ampere GPU and Mellanox smartNICs for moving massive amounts of data.

That’s just a start. The company said there are 70 systems from 11 system makers engaged in the program. There is no cost for OEMs and other partners to participate in the certification program.

In order to achieve certification, the servers must pass tests on deep-learning training and inference, machine learning algorithms, intelligent video analytics, and network and storage offload, all of which are based on software from Nvidia’s NGC catalog.

Nvidia-Certified Systems come with as many as eight A100 GPUs and high-speed InfiniBand or Ethernet network adapters. Others are mainstream AI systems tailored to run AI at the edge of the corporate network.

OEMs certify the systems using Nvidia Mellanox cables, switches and network cards such as ConnectX-6 InfiniBand or Ethernet adapters and BlueField-2 DPUs. In addition to high throughput at low latency, these adapters support multiple layers of security, from a hardware root of trust at boot time to connection tracking for applications.

All Nvidia-Certified Systems are available with enterprise support across the full software stack, including support for open-source code.