At Computex, NVIDIA has announced the certification of dozens of additional servers to run NVIDIA AI Enterprise software, indicating a rapid expansion of the NVIDIA-Certified Systems program, which now includes more than 50 server systems from prominent manufacturers around the world.
NVIDIA Certified server systems, with NVIDIA’s accelerators and GPUs included, will be built by companies like HPE, Lenovo, Dell, Gigabyte, Asus, among many others.
Enterprises can manage a variety of demanding workloads in traditional data centers and hybrid clouds with NVIDIA-Certified server systems. These include deploying an AI-ready enterprise platform that scales AI workloads using the NVIDIA AI Enterprise suite of AI and data analytics tools on VMware vSphere, NVIDIA Omniverse Enterprise for design collaboration and sophisticated simulation, and Red Hat OpenShift for AI development.
The solutions would also integrate seamlessly with Cloudera data engineering and machine learning, allowing models to be delivered in minutes rather than hours.
“Enterprises across every industry need to support their innovative work in AI on traditional data center infrastructure,” said Manuvir Das, head of Enterprise Computing at NVIDIA. “The open, growing ecosystem of NVIDIA-Certified Systems provides unprecedented customer choice in servers validated by NVIDIA to power world-class AI.”
NVIDIA A100, A40, A30, and A10
NVIDIA-Certified server systems go through extensive testing and follow NVIDIA’s design best practices in terms of performance, security, and scalability. The server systems include NVIDIA A100, A40, A30, or A10 Tensor Core GPUs, as well as NVIDIA BlueField-2 DPUs or NVIDIA ConnectX-6 adapters, and are available at a variety of pricing and performance levels.
Dell Technologies, HPE, Nettrix, and Supermicro are the latest companies to offer newly certified servers based on the NVIDIA HGX accelerated computing platform for advanced AI training and cloud computing services.
With four or eight NVIDIA A100 GPUs, NVIDIA NVLink GPU interconnects, NVIDIA InfiniBand networking, and NVIDIA’s AI and HPC software stack, these would deliver industry-leading AI performance.
Boosting Security: DPUs BlueField-2
New servers utilizing BlueField-2 DPUs, or data processing units, from several of the world’s major server system manufacturers will debut later this year in a new category of NVIDIA-Certified Systems to allow “breakthrough” networking, storage, and security performance.
A single BlueField-2 DPU may deliver data center services that could need up to 125 CPU cores by offloading activities off the CPU, freeing up server CPU cycles to run a wide range of business-critical applications.
Servers with Arm CPUs Arriving in 2022
Gigabyte and Wiwynn aim to deliver new servers with Arm Neoverse-based CPUs, as well as NVIDIA Ampere architecture GPUs or BlueField-2 DPUs, or both, to expand the Arm ecosystem into the enterprise for high-performance AI computing. The servers will be available next year and will be submitted for NVIDIA certification as soon as they are available.
Gigabyte and NVIDIA have partnered to deliver an Arm HPC Developer Kit, which is a complete hardware and software platform for developing HPC, AI, and scientific computing applications. The platform includes an Arm Neoverse-based Ampere Altra CPU from Ampere Computing, two A100 GPUs, two BlueField-2 DPUs, and the NVIDIA HPC SDK, which has been validated by NVIDIA to fulfill severe HPC application requirements.
NVIDIA Ampere architecture GPUs are now available in NVIDIA-Certified server systems. Later this year, server systems with NVIDIA BlueField-2 DPUs will be available. In 2022, server systems with Arm CPUs will become available.