Supermicro Extends 8-GPU, 4-GPU, and MGX Product Strains with Help for the NVIDIA HGX H200 and Grace Hopper Superchip for LLM Functions with Sooner and Bigger HBM3e Reminiscence – New Progressive Supermicro Liquid Cooled 4U Server with NVIDIA HGX 8-GPUs Doubles the Computing Density Per Rack, and As much as 80kW/Rack, Lowering TCO
SAN JOSE, Calif.and DENVER, Nov. 13, 2023 /PRNewswire/ — Supercomputing Convention (SC23) — Supermicro, Inc. (NASDAQ: SMCI)a Whole IT Resolution Supplier for AI, Cloud, Storage, and 5G/Edge, is increasing its AI attain with the upcoming assist for the brand new NVIDIA HGX H200 constructed with H200 Tensor Core GPUs. Supermicro’s trade main AI platforms, together with 8U and 4U Common GPU Techniques, are drop-in prepared for the HGX H200 8-GPU, 4-GPU, and with almost 2x capability and 1.4x increased bandwidth HBM3e reminiscence in comparison with the NVIDIA H100 Tensor Core GPU. As well as, the broadest portfolio of Supermicro NVIDIA MGXTM methods helps the upcoming NVIDIA Grace Hopper Superchip with HBM3e reminiscence. With unprecedented efficiency, scalability, and reliability, Supermicro’s rack scale AI options speed up the efficiency of computationally intensive generative AI, giant language Mannequin (LLM) coaching, and HPC functions whereas assembly the evolving calls for of rising mannequin sizes. Utilizing the constructing block structure, Supermicro can shortly carry new expertise to market, enabling clients to develop into extra productive sooner.
Supermicro 4U 8GPU H200
Supermicro can be introducing the trade’s highest density server with NVIDIA HGX H100 8-GPUs methods in a liquid cooled 4U system, using the newest Supermicro liquid cooling answer. The trade’s most compact excessive efficiency GPU server permits information heart operators to cut back footprints and power prices whereas providing the best efficiency AI coaching capability obtainable in a single rack. With the best density GPU methods, organizations can cut back their TCO by leveraging cutting-edge liquid cooling options.
“Supermicro companions with NVIDIA to design essentially the most superior methods for AI coaching and HPC functions,” stated Charles Liangpresident and CEO of Supermicro. “Our constructing block structure permits us to be first to market with the newest expertise, permitting clients to deploy generative AI quicker than ever earlier than. We will ship these new methods to clients quicker with our worldwide manufacturing amenities. The brand new methods, utilizing the NVIDIA H200 GPU with NVIDIA® NVLink™ and NVSwitch™ high-speed GPU-GPU interconnects at 900GB/s, now present as much as 1.1TB of high-bandwidth HBM3e reminiscence per node in our rack scale AI options to ship the best efficiency of mannequin parallelism for immediately’s LLMs and generative AI. We’re additionally excited to supply the world’s most compact NVIDIA HGX 8-GPU liquid cooled server, which doubles the density of our rack scale AI options and reduces power prices to attain inexperienced computing for immediately’s accelerated information heart.”
Study extra in regards to the Supermicro servers with NVIDIA GPUs
Supermicro designs and manufactures a broad portfolio of AI servers with completely different kind components. The favored 8U and 4U Common GPU methods that includes four-way and eight-way NVIDIA HGX H100 GPUs are actually drop-in prepared for the brand new H200 GPUs to coach even bigger language fashions in much less time. Every NVIDIA H200 GPU comprises 141GB of reminiscence with a bandwidth of 4.8TB/s.
“Supermicro’s upcoming server designs utilizing NVIDIA HGX H200 will assist speed up generative AI and HPC workloads, in order that enterprises and organizations can get essentially the most out of their AI infrastructure,” stated Dion Harrisdirector of knowledge heart product options for HPC, AI, and quantum computing at NVIDIA. “The NVIDIA H200 GPU with high-speed HBM3e reminiscence will be capable of deal with large quantities of knowledge for quite a lot of workloads.”
Moreover, the not too long ago launched Supermicro MGX servers with the NVIDIA GH200 Grace Hopper Superchips are engineered to include the NVIDIA H200 GPU with HBM3e reminiscence.
The brand new NVIDIA GPUs enable acceleration of immediately’s and future giant language fashions (LLMs) with 100s of billions of parameters to slot in extra compact and environment friendly clusters to coach Generative AI with much less time and likewise enable a number of bigger fashions to slot in one system for real-time LLM inference to serve Generative AI for tens of millions of customers.
At SC23, Supermicro is showcasing the newest providing, a 4U Common GPU System that includes the eight-way NVIDIA HGX H100 with its newest liquid-cooling improvements that additional enhance density and effectivity to drive the evolution of AI. With Supermicro’s trade main GPU and CPU chilly plates, CDU(cooling distribution unit), and CDM (cooling distribution manifold) designed for inexperienced computing, the brand new liquid-cooled 4U Common GPU System can be prepared for the eight-way NVIDIA HGX H200, which can dramatically cut back information heart footprints, energy price, and deployment hurdles by means of Supermicro’s absolutely built-in liquid-cooling rack options and our L10, L11 and L12 validation testing.
For extra info, go to the Supermicro sales space at SC23
About Tremendous Micro Laptop, Inc.
Supermicro (NASDAQ: SMCI) is a world chief in Utility-Optimized Whole IT Options. Based and working in San Jose, CaliforniaSupermicro is dedicated to delivering first to market innovation for Enterprise, Cloud, AI, and 5G Telco/Edge IT Infrastructure. We’re a Whole IT Options producer with server, AI, storage, IoT, swap methods, software program, and assist companies. Supermicro’s motherboard, energy, and chassis design experience additional permits our growth and manufacturing, enabling subsequent technology innovation from cloud to edge for our world clients. Our merchandise are designed and manufactured in-house (within the US, Taiwanand the Netherlands), leveraging world operations for scale and effectivity and optimized to enhance TCO and cut back environmental impression (Inexperienced Computing). The award-winning portfolio of Server Constructing Block Options® permits clients to optimize for his or her precise workload and software by deciding on from a broad household of methods constructed from our versatile and reusable constructing blocks that assist a complete set of kind components, processors, reminiscence, GPUs, storage, networking, energy, and cooling options (air-conditioned, free air cooling or liquid cooling).
Supermicro, Server Constructing Block Options, and We Maintain IT Inexperienced are emblems and/or registered emblems of Tremendous Micro Laptop, Inc.
All different manufacturers, names, and emblems are the property of their respective homeowners.
Picture – https://mma.prnewswire.com/media/2274161/03_2023__4U_8GPU_H200_1080.jpg
Brand – https://mma.prnewswire.com/media/1443241/Supermicro_Logo.jpg
