- Solution
GIGAPOD - AI Supercomputing Solution
Scalable, Turnkey AI Supercomputing Solution One GIGAPOD is composed of multiple racks populated with GIGABYTE GPU servers acting as one powerful cluster that accelerates everything AI. Unleash a Turnkey AI Data Center with High Throughput and an Incredible Level of Compute GIGABYTE has been pivotal in providing its technology leaders with a supercomputing infrastructure built around powerful GIGABYTE GPU servers that house either NVIDIA H200 Tensor Core GPUs or AMD Instinct™ MI300 Series accelerators. GIGAPOD is a service that has professional help to create a cluster of racks all interconnected as a cohesive unit. An AI ecosystem platform thrives with a high degree of parallel processing as the GPUs are interconnected with blazing fast communication by NVIDIA NVLink or AMD Infinity. Fabric. With the introduction of the GIGAPOD, GIGABYTE now offers a one-stop source for data centers that are moving to an AI factory that runs deep learning models at scale. The hardware, expertise, and close relationship with cutting-edge GPU partners ensures the deployment of an AI supercomputer goes off without a hitch and minimal downtime. Consulting Design with GIGABYTE servers, networking… Construction Deployment Validation GIGABYTE G Series Servers Built for 8-GPU Platforms One of the most important considerations when planning a new AI data center is the selection of hardware, and in this AI era, many companies see the choice of the GPU/Accelerator as the foundation. Each of GIGABYTE’s industry leading GPU partners (AMD, Intel, and NVIDIA) has innovated uniquely advanced products built by a team of visionary and passionate researchers and engineers, and as each team is unique, each new generational GPU technology has advances that make it ideal for particular customers and applications. This consideration of which GPU to build from is mostly based on factors: performance (AI training or inference), cost, availability, ecosystems, scalability, efficiency, and more. The decision isn’t easy, but GIGABYTE aims to provide choices, customization options, and the know-how to create ideal data centers to tackle the demand and increasing parameters in AI/ML models. NVIDIA HGX™ H200/H100 Biggest AI Software Ecosystem AMD Instinct™ MI300X Largest & Fastest Memory Intel® Gaudi® 3 Excellence in AI Inference Strong Top to Bottom Ecosystem for Success This comprehensive enterprise ecosystem provides the necessary expertise, resources, innovation, and interoperability to build, deploy, and maintain cutting-edge, large-scale AI infrastructure. GIGABYTE continues to expand its partners who are leaders in their respective fields, to ensure that rack-scale projects don’t get stuck at any phase. Customers will quickly reap the fruits of their investment through a developed and evolving ecosystem of partners. 1 2 3 4 5 6 Compute Networking Storage Management Cooling Rack / PDU Why is GIGAPOD the rack scale service to deploy? Experienced GIGABYTE has successfully deployed large GPU clusters and is ready to discuss the process and provide a timeline that fulfills customers requirements. Depth in Portfolio GIGABYTE servers (GPU, Compute, Storage, & High-density) have numerous SKUs that are tailored for all imaginable enterprise applications. Scale Up or Out A turnkey high-performing data center has to be built with expansion in mind so new nodes or processors can effectively become integrated. High Performance From a single GPU server to a cluster, GIGABYTE has tailored its server and rack design to guarantee peak performance with optional liquid cooling. Industry Connections GIGABYTE works closely with technology partners - AMD, Intel, and NVIDIA - to ensure a fast response to customers requirements and timelines. The Future of AI Computing in Data Centers The Ideal GIGAPOD for You GIGABYTE enterprise products not only excel at reliability, availability, and serviceability. They also shine in flexibility, whether it be the choice of GPU, rack dimensions, or cooling method and more. GIGABYTE is familiar with every imaginable type of IT infrastructure, hardware, and scale of data center. Many GIGABYTE customers decide on the rack configuration based on how much power their facility can provide to the IT hardware, as well as considering how much floor space is available. So, this is why the service, GIGAPOD, came to be. Customers have choices. Starting with how the components are cooled and how the heat is removed, customers can select either traditional air-cooling or direct liquid cooling (DLC). 1 2 Air Cooling Direct Liquid Cooling Ver. GPUs Supported GPU Server(Form Factor) GPU Serversper Rack Rack Power Consumptionper Rack RDHx 1 NVIDIA HGX™ H100/H200/B100AMD Instinct™ MI300X 5U 4 9 x 42U 50kW No 2 NVIDIA HGX™ H100/H200/B100AMD Instinct™ MI300X 5U 4 9 x 48U 50kW Yes 3 NVIDIA HGX™ H100/H200/B100AMD Instinct™ MI300X 5U 8 5 x 48U 100kW Yes 4 NVIDIA HGX™ B200 8U 4 9 x 42U 130kW No 5 NVIDIA HGX™ B200 8U 4 9 x 48U 130kW No Ver. GPUs Supported GPU Server(Form Factor) GPU Serversper Rack Rack Power Consumptionper Rack CDU 1 NVIDIA HGX™ H100/H200/B100AMD Instinct™ MI300X 5U 8 5 x 48U 100kW In-rack 2 NVIDIA HGX™ H100/H200/B100AMD Instinct™ MI300X 5U 8 5 x 48U 100kW External 3 NVIDIA HGX™ B200/H200 4U 8 5 x 42U 130kW In-rack 4 NVIDIA HGX™ B200/H200 4U 8 5 x 42U 130kW External 5 NVIDIA HGX™ B200/H200 4U 8 5 x 48U 130kW In-rack 6 NVIDIA HGX™ B200/H200 4U 8 5 x 48U 130kW External The Future of AI Computing in Data Centers none Discover the GIGAPOD From one GIGABYTE GPU server to eight racks with 32 GPU nodes (a total of 256 GPUs), GIGAPOD has the infrastructure to scale, achieving a high-performance supercomputer. Cutting-edge data centers are deploying AI factories, and it all starts with a GIGABYTE GPU server. GIGAPOD is more than just a bunch of GPU servers, there are also switches. Not to mention, the complete solution offers hardware, software, and services to deploy with ease. Applications for GPU Clusters Science & Engineering Research in fields such as physics, chemistry, geology, and biology greatly benefit with the use of GPU accelerated clusters. Simulations and modeling thrive with the parallel processing capability of GPUs. Generative AI Generative AI algorithms can create synthetic data that is used in training AI and it can help automate industrial tasks. This is all possible by a GPU cluster using powerful GPUs with fast Infiniband networking. Large Language Models (LLM) Training models that use billions of parameters while having sufficient HBM/memory is a challenge. And text-based data, as in LLM, thrives with a GPU cluster that has a single scalable unit with over 20 TB of GPU memory, making it ideal in scale. Related Products Resources
