Generative AI SuperCluster integrates with NVIDIA AI Enterprise and NIM microservices, and provides immediate ROI benefits and lower AI operating costs through massively scalable computing units , thereby simplifying AI for rapid deployment.
Supermicro, Inc. provides comprehensive IT solutions for AI, cloud, storage and 5G/edge areas Manufacturer launches ready-to-deploy liquid-cooled AI data center. This data center is specially designed for cloud-native solutions. It uses SuperCluster to accelerate the application of generative AI by enterprises from all walks of life. It is optimized for the NVIDIA AI Enterprise software platform and is suitable for the development and deployment of generative AI. Through Supermicro’s 4U liquid cooling technology, NVIDIA’s recently launched Blackwell GPU can fully exploit the AI performance of 20 PetaFLOPS on a single GPU, and can provide 4 times the AI training performance and 30 times the inference compared to earlier GPUs. efficiency and additional cost savings. In line with this first-to-market strategy, Supermicro recently launched a complete product series based on the NVIDIA Blackwell architecture, supporting the new NVIDIA HGX TM B100, B200 and GB200 Grace Blackwell Superchip.
Liang Jianhou, President and CEO of Supermicro, said: “Supermicro continues to lead the industry in building and deploying AI solutions with rack-level liquid cooling technology. Liquid-cooling configurations in data centers can be designed for almost free and can continue to reduce usage. The advantage of power provides customers with additional value. Our solutions are optimized for NVIDIA AI Enterprise software, meet customer needs in various industries, and provide world-scale efficient global manufacturing capabilities, so we can shorten delivery time and improve efficiency. Quickly provide ready-to-use liquid or air-cooled computing clusters with NVIDIA HGX H100 and H200, and upcoming B100, B200 and GB200 solutions. From liquid cooling panels to CDUs and cooling towers, our full range of rack-level liquid cooling. The cooling solution can reduce the continuous power consumption of the data center by up to 40%.”
At Computex 2024 in Taipei, Supermicro demonstrated upcoming systems optimized for NVIDIA Blackwell GPUs, including a 10U air cooling system and a 4U liquid cooling system based on NVIDIA HGX B200. In addition, Supermicro will also offer 8U air-cooled NVIDIA HGX B100 systems, NVIDIA GB200 NVL72 racks (featuring 72 GPUs interconnected through NVIDIA NVLink switches) and new NVIDIA MGX™ systems (supporting NVIDIA H200 NVL PCIe GPUs and the latest releases NVIDIA GB200 NVL2 architecture).
NVIDIA founder and CEO Jensen Huang said: “Generative AI is driving the reset of the entire computing stack. The new data center will use GPUs to accelerate computing and optimize for AI. Supermicro has designed top-notch NVIDIA accelerated computing and network solutions to enable multi-trillion dollar global data centers to be optimized for the AI era.”
With the rapid development of Large Language Model (LLM) and the continuous launch of open source models such as Meta’s Llama-3 and Mistral’s Mixtral 8x22B, it is easier for enterprises to obtain and use today’s most advanced AI models. Simplifying AI infrastructure and providing access in the most cost-effective manner is critical to supporting today’s rapid AI transformation. Supermicro’s cloud-native AI SuperCluster bridges the gap between the convenience and portability of instant access to the cloud, and enables AI projects of any size to move smoothly from the evaluation test stage to actual operations through NVIDIA AI Enterprise. This provides the flexibility to run and securely manage data anywhere, including on-premises systems or large local data centers.
As enterprises from all walks of life rapidly implement experimental use of generative AI, Supermicro works closely with NVIDIA to ensure a complete and flexible push from experimentation and evaluation pilots of AI applications to operational deployment and large-scale data center AI. This smooth process is achieved through rack- and cluster-level optimization of the NVIDIA AI Enterprise software platform, making the process from initial exploration to scalable AI implementation smoother and more seamless.
Managed services involve infrastructure selection, data sharing and generative AI policy control. NVIDIA NIM microservices, part of NVIDIA AI Enterprise, deliver managed generative AI and the benefits of open source deployment without bugs. Its multifunctional inference runtime environment uses microservices to accelerate the deployment of various types of generative AI, from open source models to NVIDIA basic models. In addition, NVIDIA NeMo TM enables customized model development through data utilization, advanced customization, and Retrieval-Augmented Generation (RAG) to achieve enterprise-level solutions. Integrated with Supermicro SuperCluster with NVIDIA AI Enterprise, NVIDIA NIM provides the fastest way to deploy scalable, accelerated generative AI jobs.
Supermicro’s current generative AI SuperCluster products include:
- Liquid-cooled Supermicro NVIDIA HGX H100/H200 SuperCluster with 256 H100/H200 GPUs as a 5-rack-scale scalable compute unit (including 1 dedicated network cabinet)
- Air-cooled Supermicro NVIDIA HGX H100/H200 SuperCluster with 256 HGX H100/H200 GPUs as a 9-rack-scale scalable compute unit (including 1 dedicated network cabinet)
- Supermicro NVIDIA MGX GH200 SuperCluster with 256 GH200 Grace TM Hopper Superchips as 9 rack-scale scalable computing units (including 1 dedicated network rack)
Supermicro’s SuperCluster supports NVIDIA AI Enterprise, which includes NVIDIA NIM microservices and the NVIDIA NeMo platform, enabling end-to-end generative AI customization and targeting NVIDIA Quantum-2 InfiniBand and the new NVIDIA Spectrum with 400Gb/s per GPU network speeds -X Ethernet platform optimization, scalable to large computing clusters with tens of thousands of GPUs.
Supermicro’s upcoming SuperClusters include:
- Supermicro NVIDIA HGX B200 SuperCluster, liquid-cooled
- Supermicro NVIDIA HGX B100/B200 SuperCluster, air-cooled
- Supermicro NVIDIA GB200 NVL72 or NVL36 SuperCluster, liquid-cooled
Supermicro’s SuperCluster solutions are optimized for LLM training, deep learning, and large-scale and high-batch inference. Supermicro‘s L11 and L12 verification testing and field deployment services provide customers with a seamless experience. Customers receive plug-and-play scalable units that can be easily deployed in the data center to realize benefits faster.
SOURCE: Businesswire