The DGX B200 is an integrated AI platform designed for companies of all sizes at every stage of their AI journey, supporting development-deployment pipelines. Equipped with 8 NVIDIA Blackwell GPUs interconnected via 5th-generation NVIDIA NV Link, the DGX B200 provides cutting-edge performance with 3 times the training performance and 15 times the inference performance compared to the previous generation. Utilizing the NVIDIA Blackwell GPU architecture, the DGX B200 is ideal for companies looking to accelerate AI innovation, as it can handle various workloads like large language models, recommendation systems, and chatbots.
The NVIDIA DGX SuperPOD with DGX GB200 systems is specifically designed for training and inference of AI models generating trillions of parameters. Each water-cooled rack connects 36 NVIDIA GB200 Grace Superchips (comprising 36 NVIDIA Grace CPUs and 72 Blackwell GPUs) through NVIDIA NVLink. Multiple racks can be connected via NVIDIA Quantum InfiniBand, allowing for scalability up to tens of thousands of GB200 Superchips.
AI supports critical use cases across industries, including healthcare, manufacturing, and financial services. As part of the NVIDIA DGX platform, the NVIDIA DGX BasePOD provides a reference architecture that enables companies to build and scale their AI infrastructure.
Learn More
The NVIDIA DGX SuperPOD is an AI data center infrastructure that allows IT to deliver consistent performance for all users and workloads. As part of the NVIDIA DGX platform, the DGX SuperPOD provides industry-proven results and leading accelerated infrastructure with agile, scalable performance for the most demanding AI workloads.
For further inquiries about our products, please feel free to contact us. Our experts are available to provide detailed guidance.