Super Micro Computer, Inc., a global leader in enterprise computing, storage, networking solutions, and green computing technologies, will deliver two Intel-powered models for high-demand cloud gaming, media delivery, AI, and ML workloads Support for new accelerators from Intel and Intel Habana, enabling customers to deploy with the latest acceleration technologies from Intel and Intel Habana.
Supermicro uses the Building Block Solutions approach to design new systems and bring new technologies to market quickly. This approach allows the latest GPU and acceleration technologies to be easily incorporated into existing designs, or to be quickly adapted to match higher performance components when needed.
Sandra Rivera, Executive Vice President and General Manager, Data Center and AI Business Unit, Intel Corporation, said, “Supermicro introduces systems featuring our latest Gaudi2 and Arctic Sound-M accelerators to help deliver state-of-the-art AI and media processing capabilities. Supermicro’s Gaudi AI training The servers will accelerate deep learning training for some of the fastest growing workloads in the data center.”
Supermicro systems with Arctic Sound-M GPUs will meet the demanding requirements of cloud gaming, media transcoding and streaming, virtual desktop infrastructure (VDI), simulation and visualization, machine learning and content creation. Using the industry’s first hardware AV-1 encoder and open source media software stack, Intel Arctic Sound-M delivers dramatic performance improvements over pure software video transcoding and delivery solutions, while Includes acceleration for VDI environments.
Supermicro will initially feature Intel Arctic Sound-M GPUs in 2U 2-node single-processor Intel systems (3 GPUs per node), 4U 10xGPU systems, and CloudDC servers, with other systems to be released later this year. The new Supermicro AI training server contains two Intel 3rd Generation Intel Xeon Scalable processors and eight Habana Gaudi2 accelerators for high-performance AI training environments.
Habana Labs Gaudi2
Habana Labs Gaudi2 excels in many workloads including vision applications such as image classification, object detection, natural language processing (NLP) models, and recommender systems. This new 8U chassis AI training server will be the first commercial implementation of the latest Habana Gaudi2 (HL-225). Combining two 3rd Gen Intel Xeon Scalable processors and up to 8TB of DRAM, this server accelerates AI training to a new level of performance. There are 24 hot-swappable drive bays in the server, which support high-capacity local high-performance storage, and also provide a lot of IO. Additionally, the Gaudi2 accelerator offers an air-cooled (HL-225) option. Habana Gaudi2 helps you resize quickly and easily. Each Habana Gaudi2 accelerator can communicate with other Gaudi2 accelerators at 700 GB per second via ROCE, and with other Gaudi2 accelerators in other servers at 2.4 TB per second. In addition, each server has 6 QSFP-DD ports, which can be easily scaled out to handle larger models and data sets.
For in-depth information on Supermicro, please refer to the official website www.supermicro.com