Supermicro first announced support for NVIDIA HGX™ A100 configurations in May with its new high-density 2U and 4U servers based on the four- and eight-way NVIDIA HGX A100 boards. With A100 GPUs now also available in PCIe form factor, customers can expect a major performance boost across Supermicro's extensive portfolio of multi-GPU servers when they are equipped with the new NVIDIA A100 GPUs.
"Expanding our industry-leading portfolio of GPU systems with full support for the new NVIDIA A100 PCIe GPUs means customers can choose highest performing and most optimised server based on their specific applications," said Charles Liang, CEO and president of Supermicro. "Designed to accelerate a vast range of compute-intensive applications, our new systems with PCIe Gen 4 deliver fully optimised support for the new NVIDIA A100 to boost performance up to 20x on some accelerated workloads."
Supermicro's new 4U A+ GPU system supports up to eight NVIDIA A100 PCIe GPUs via direct-attach PCI-E 4.0 x16 CPU-to-GPU lanes without any PCI-E switch for the lowest latency and highest bandwidth to deliver maximum acceleration. The system also supports up to two additional high-performance PCI-E 4.0 expansion slots for a variety of uses, including high-performance networking connectivity up to 200Gb/s. An additional AIOM slot supports a Supermicro AIOM card or an OCP NIC 3.0 card. These systems will also be NGC-Ready, providing customers a seamless way to develop and deploy their AI workloads at scale. NGC-Ready systems are validated for functionality and performance for the AI stack from NVIDIA's NGC registry.
"As the first elastic, multi-instance GPU that unifies training, inference, HPC, and analytics, the NVIDIA A100 is built to handle workloads of all sizes," said Paresh Kharya, Director of Product Marketing of Accelerated Computing at NVIDIA. "NVIDIA A100 PCIe GPU servers allow customers to tackle some of the most complex data science, high performance computing and AI challenges."