Supermicro expands its portfolio with fully integrated systems with NVIDIA A100 GPU server, offering 5 AI performance PetaFLOPS on a single 4U server

 Super Micro Computer, Inc., a global enterprise computing, storage, networking solutions and green computing technology, announced two new systems created for deep learning artificial intelligence (AI) applications that fully benefit from third generation NVIDIA HGX ™ technology with the new NVIDIA A100 ™ Tensor Core GPUs, as well as full support for the new NVIDIA A100 GPUs across the company's comprehensive portfolio 1U, 2U, 4U and 10U server company. The NVIDIA A100 is the first elastic multi-instance GPU that unifies training, inference, HPC and analytics.

"Expanding our market-leading portfolio of GPU systems and NVIDIA HGX-2 system technology, Supermicro is introducing a new 2U system implementing the new NVIDIA HGX ™ A100 4-GPU GPU card (formerly Redstone) and a new 4U system based on the new NVIDIA HGX A100 8 GPU network card (formerly called Delta), delivering 5 AI performance PetaFLOPS, "said Charles Liang, executive director and president of Supermicro. "As GPU-accelerated computing evolves and continues to transform data centers, Supermicro will provide customers with the latest advances in the system, to help them achieve maximum acceleration at all scales, optimizing server utilization. GPU. These new systems will significantly improve performance across all accelerated HPC workloads, data analysis, deep learning training and deep learning inference. "

As a balanced data center platform for HPC and AI applications, Supermicro's new 2U system leverages the NVIDIA HGX A100 4 GPU network card with four direct-connect NVIDIA A100 Core Tensor GPUs using PCI-E 4.0 for maximum performance and NVIDIA NVLink ™ for high speed GPU to GPU interconnects. This advanced GPU system accelerates computing, network and storage performance with compatibility for one PCI-E 4.0 x8 expansion slot and up to four PCI-E 4.0 x16 for high-speed GPUDirect RDMA network cards and storage, such as InfiniBand ™ HDR ™, compatible with bandwidth up to 200 Gb per second.

"AI models are exploding in complexity as they face challenges at the next level, such as accurate conversational AI, deep recommendation systems and personalized medicine," said Ian Buck , general manager and vice president of accelerated computing at NVIDIA. "By implementing the NVIDIA HGX A100 platform on its new servers, Supermicro offers customers powerful performance and considerable scalability, allowing researchers to train the most complex AI networks at speed never seen before."

Optimized for AI and machine learning, the new 4U system from Supermicro is compatible with eight Tensor Core A100 GPUs. The 4U form factor with eight GPUs is ideal for customers who want to scale up their implementation as their processing requirements expand. The new 4U system will feature an NVIDIA HGX A100 8 GPU card with eight A100 GPUs connected with NVIDIA NVSwitch ™ for up to 600 GB per second of GPU to GPU bandwidth and eight expansion slots for GPUDirect RDMA high-speed network cards . Ideal for deep learning training, data centers can use this expansion platform to create next-generation AI and maximize the productivity of data scientists, with support for ten x16 expansion slots.

Customers can expect a significant increase in performance from Supermicro's extensive portfolio of multiple 1U, 2U, 4U and 10U GPU servers when equipped with the new NVIDIA A100 GPU servers. For maximum acceleration, Supermicro's new A + GPU system is compatible with up to eight full-height and double-width (or single-width) GPUs via PCI-E 4.0 x16 CPU-to-GPU direct-connect strips without a PCI- switch. And for the lowest latency and highest bandwidth. The system is also compatible with up to three additional high-performance PCI-E 4.0 expansion slots for a number of uses, including high-performance network connectivity up to 100G. An additional AIOM slot is compatible with the AIOM Supermicro card or OCP 3.0 mezzanine card.

As a leader in AI systems technology, Supermicro offers thermal designs optimized for multiple GPU servers, which provide the highest performance and reliability for AI, Deep Learning and HPC applications. With 1U, 2U, 4U and 10U rack-mounted GPU systems, Utra, BigTwin ™ and GPU-compatible embedded systems, in addition to GPU blade modules for SuperBlade® 8U, Supermicro offers the industry's broadest and deepest selection GPU systems to power edge applications to the cloud.

To provide more security and unmatched edge performance, Supermicro plans to add the new NVIDIA EGX ™ A100 configuration to its edge server portfolio. The convergent accelerator EGX A100 combines an NVIDIA Mellanox SmartNIC with GPUs equipped with the new NVIDIA Ampere architecture, so that companies can run AI at the edge more safely.