The first 13 systems from five OEMs to be certified have been announced, but NVIDIA expects that before long, about 70 systems from 15 OEMs will meet their engineering requirements for certification.
NVIDIA has launched a program that they are terming a first of its type, that will test and certify OEM systems to validate their suitability for modern accelerated AI workloads. The NVIDIA-Certified Systems program launches with 13 systems from five OEMs — Dell Technologies, GIGABYTE, Hewlett Packard Enterprise, Inspur and Supermicro. These are just the first to be certified however, and NVIDIA expects a significant amount of other systems and vendors to participate.
“The NVIDIA-Certified program validates systems that are certified to optimally run workloads from our NVIDIA NGC catalogue,” said Adel El-Halak, Director of Product Management, NVIDIA NGC. “This is our treasure trove of software, our hub for GPU-accelerated software.”
The NVIDIA-Certified program validates that OEM systems will run workloads at scale using this software, and gives customers a fast and easy way to procure these systems, and obtain support for their software through NVIDIA.
“We believe that this is first of a kind program,” El-Halak said. “It’s AI-centric and meant to ensure AI-based workloads will run optimally at scale. Everything is tested against software that comes from the NGC catalogue.”
El-Halak said that market momentum for these solutions running accelerated AI solutions has reached the point that such a program became necessary.
“Nearly 40% of organization have AI deployed and in production, a number Gartner says will double in the next three years,” he stated. The program will ensure the systems meet standards NVIDIA has set to be functional enough to serve different and divergent workflows, while delivering performance, scalability and performance, and security.
“Doing AI at scale is difficult,” El-Halak stated. “In the past, it has largely been a ‘do it yourself’ system. We’ve replaced that by figuring out what it takes to scale efficiently and developing criteria that identifies systems that can deploy and scale easily. It takes a complex process and makes it as predictable as possible.”
The thirteen systems that have been validated as of today are: Dell EMC PowerEdge R7525 and R740 rack servers; GIGABYTE R281-G30, R282-Z96, G242-Z11, G482-Z54, G492-Z51 systems; HPE Apollo 6500 Gen10 System and HPE ProLiant DL380 Gen10 Server; Inspur NF5488A5, and Supermicro A+ Server AS -4124GS-TNR and AS -2124GQ-NART.
“Launching a program isn’t exciting unless you have participants, and we have 13 systems from five OEMs that are certified now,” El Halak said. “We expect up to 70 systems from about 15 OEMs to be qualified in the near term. 11 have reached out so far. What we are highlighting today is the ones who have already passed and have systems available today.”
Different types of vendors are also expected to participate going forward.
“We will also be certifying more of the ecosystem stack, so companies like VMware and Red Hat are likely to be folded in as part of this,” El-Halak indicated.
System builders like Microway or Nortek who build systems that aren’t simply using validated designs of the ODMs would also be eligible. There is no fee for any of the companies whose designs are validated through the program.
Organizations who purchase validated designs are eligible for NVIDIA software support, which will be made available through the OEMs. The cost will depend on the number of GPUs in the system.
At the same time, El-Halak emphasized that the program hasn’t been designed to sell software support.
“This is primarily an engineering program, not a marketing program,” he stressed. While the OEMs will certainly use their validation by NVIDIA in their own marketing, NVIDIA’s concern is that the systems meet NVIDIA’s requirements.
“The program is a differentiator because we tap into real workloads, and we are also testing for scale,” he stated.