Wednesday, June 5, 2024

Dive into the Future: Supermicro's Liquid-Cooled AI SuperClusters




Picture this: Supermicro is taking the AI world by storm with its latest innovation, introducing rack-scale, plug-and-play liquid-cooled AI SuperClusters. These cutting-edge systems are optimized for NVIDIA Blackwell and NVIDIA HGX H100/H200. Supermicro’s radical innovations in the AI era are turning heads, making liquid-cooling not just efficient but practically free. And yes, there's a bonus!

Supermicro, based in San Jose, Calif., and Taipei, Taiwan, is a total IT solution provider for AI, cloud, storage, and 5G/edge. They’ve rolled out a ready-to-deploy liquid-cooled AI data center. This game-changing solution is designed for cloud-native applications, accelerating generative AI adoption across various industries. Their SuperClusters are optimized for the NVIDIA AI Enterprise software platform, making the development and deployment of generative AI smoother than ever.


Hold onto your seats because Supermicro’s 4U liquid-cooled systems are here to blow your mind. With NVIDIA's recently introduced Blackwell GPUs, these systems can unleash 20 PetaFLOPS of AI performance on a single GPU. That’s right, 20 PetaFLOPS! They showcase a jaw-dropping 4X better AI training and 30X better inference performance than previous GPUs, all while cutting costs. Supermicro's first-to-market strategy is evident with their complete line of NVIDIA Blackwell architecture-based products, including the new NVIDIA HGX™ B100, B200, and GB200 Grace Blackwell Superchip.

Charles Liang, the president and CEO of Supermicro, couldn't have said it better: "Supermicro continues to lead the industry in creating and deploying AI solutions with rack-scale liquid-cooling." Their data centers promise to be virtually free, offering customers bonus value with reduced electricity usage. Supermicro’s solutions are optimized with NVIDIA AI Enterprise software, ensuring efficiency and global manufacturing capacity. This means faster delivery times for their liquid-cooled or air-cooled turnkey clusters with NVIDIA HGX H100 and H200, as well as the upcoming B100, B200, and GB200 solutions. From cold plates to CDUs to cooling towers, Supermicro’s rack-scale total liquid cooling solutions can cut ongoing data center power usage by up to 40%.



At COMPUTEX 2024, Supermicro is set to unveil its next big thing: systems optimized for the NVIDIA Blackwell GPU, including a 10U air-cooled and a 4U liquid-cooled NVIDIA HGX B200-based system. But that’s not all. They’re also rolling out an 8U air-cooled NVIDIA HGX B100 system and Supermicro's NVIDIA GB200 NVL72 rack, which contains 72 interconnected GPUs with NVIDIA NVLink Switches. They’re even introducing the new NVIDIA MGX™ systems supporting NVIDIA H200 NVL PCIe GPUs and the recently announced NVIDIA GB200 NVL2 architecture.

Jensen Huang, the founder and CEO of NVIDIA, sums it up perfectly: "Generative AI is driving a reset of the entire computing stack — new data centers will be GPU-accelerated and optimized for AI." Supermicro’s cutting-edge NVIDIA accelerated computing and networking solutions are enabling global data centers to be optimized for this AI era.



The rapid development of large language models and the constant introduction of new open-source models like Meta's Llama-3 and Mistral's Mixtral 8x22B have made state-of-the-art AI models more accessible for enterprises. The need to simplify AI infrastructure and provide cost-efficient accessibility is more crucial than ever. Supermicro’s cloud-native AI SuperCluster bridges the gap between the convenience of instant cloud access and portability. Leveraging the NVIDIA AI Enterprise, it allows AI projects to move seamlessly from pilot to production at any scale. This ensures flexibility to run anywhere with securely managed data, whether in self-hosted systems or on-premises large data centers.

As enterprises across industries rapidly experiment with generative AI use cases, Supermicro is collaborating closely with NVIDIA to ensure a seamless and flexible transition from experimenting and piloting AI applications to production deployment and large-scale data center AI. This is achieved through rack and cluster-level optimization with the NVIDIA AI Enterprise software platform, enabling a smooth journey from initial exploration to scalable AI implementation.



Managed services often compromise infrastructure choices, data sharing, and generative AI strategy control. However, NVIDIA NIM microservices, part of the NVIDIA AI Enterprise, offer managed generative AI and open-source deployment benefits without these drawbacks. Its versatile inference runtime with microservices accelerates generative AI deployment across various models, from open-source to NVIDIA's foundation models. Additionally, NVIDIA NeMo™ enables custom model development with data curation, advanced customization, and retrieval-augmented generation (RAG) for enterprise-ready solutions. Combined with Supermicro's NVIDIA AI Enterprise-ready SuperClusters, NVIDIA NIM provides the fastest path to scalable, accelerated generative AI production deployments.

Supermicro’s current generative AI SuperCluster offerings include liquid-cooled and air-cooled Superclusters with 256 H100/H200 GPUs as scalable compute units. They also offer Supermicro NVIDIA MGX GH200 SuperClusters with 256 GH200 Grace™ Hopper Superchips. These solutions are end-to-end generative AI customization-ready and optimized for NVIDIA Quantum-2 InfiniBand and the new NVIDIA Spectrum-X Ethernet platform with 400Gb/s networking speed per GPU, scaling out to a large cluster with tens of thousands of GPUs.



Looking ahead, Supermicro's upcoming SuperCluster offerings include liquid-cooled and air-cooled options for NVIDIA HGX B200 and B100/B200 systems, as well as the liquid-cooled NVIDIA GB200 NVL72 or NVL36 systems. These solutions are optimized for LLM training, deep learning, and high-volume batch-size inference, providing seamless plug-and-play scalable units for easy deployment in data centers and faster time to results.

Supermicro, a global leader in Application-Optimized Total IT Solutions, has been committed to delivering first-to-market innovation for enterprise, cloud, AI, and 5G Telco/Edge IT infrastructure. Founded and operating in San Jose, California, Supermicro’s expertise in motherboard, power, and chassis design enables them to develop and produce next-generation innovations from cloud to edge for their global customers. Their award-winning portfolio of Server Building Block Solutions® allows customers to optimize for their exact workload and application needs. Whether it’s air-conditioned, free air cooling, or liquid cooling, Supermicro’s flexible and reusable building blocks support a comprehensive set of form factors, processors, memory, GPUs, storage, networking, power, and cooling solutions.

Visit www.supermicro.com/ai for more information and dive into the future with Supermicro’s groundbreaking liquid-cooled AI SuperClusters!

No comments:

Post a Comment