Nvidia announced that Dell EMC, Gigabyte, HPE, Inspur and Supermicro are now shipping servers using Nvidia A100 Tensor Core GPUs under a new certification approach.
The dozen servers shipping from the five companies all carry an Nvidia-Certified System badge indicating that they meet Nvidia’s top design practices to “tackle the toughest tasks in machine learning, data analytics and more,” said Adel El-Hallak, director of product management for NGC at Nvidia in a blog.
The servers all deliver AI capabilities with the GPUs connected to fast Mellanox networks, he said.
The certification program means that enterprises can confidently acquire the servers to run their AI workloads at scale, Nvidia said in response to questions from reporters. “We have support from the OEMs as they recognize the value of provide Nvidia-Certified Systems to help enterprises with their accelerated computing needs and have agreed to jointly work with us on this,” Nvidia said.
OEMs don’t pay Nvidia to participate in the certification program. Software support under the certification system is priced per system, which varies based on the system’s configuration, Nvidia said. In one example, for volume servers, 2 GPUs A100 systems are priced at $4,299 per system with a three-year support term that customers can renew.
Both Nvidia and the various OEMs have designated responsibilities for customer support under a joint agreement. Nvidia said if a server customer needs support and calls the OEM first, but the OEM determines the problem is an Nvidia software issue, then Nvidia wants the customer to open a support case with Nvidia and provide the server vendor’s case number for collaboration.
And, Nvidia said if a customer calls Nvidia first they can use the enterprise support portal.
If Nvidia finds that the customer’s problem is an issue that the OEM is reponsible for, Nvidia will request the customer open a case with the OEM vendor.
Nvidia took the announcement of the shipping servers as an opportunity to tout how AI is being used by major corporations. American Express, for example, if using AI models for real-time fraud detection, while Ford is using generative adversarial networks to build up data it needs to test self-driving cars. Dominos is also using AI to improve predictions of when orders will be ready for pizzas it delivers.
Walmart is finding strategy insights in its data gathered at 11,000 stores and additional distribution and logistics systems, which totals 2.5 every hour. AI models used to sift through dta have grown nearly 30,000 times in just five years.
Nvidia cited Gartner research showing that 37% of all organizations have AI in production, a number that will double to 75% by 2024.
RELATED: What’s cookin’ at Nvidia: A100 graphics card pops out of Huang’s oven