NVIDIa DGX Systems

The NVIDIA DGX B200 is NVIDIA’s most advanced AI supercomputer to date, purpose-built for running massive generative AI and large language model (LLM) workloads. Powered by eight NVIDIA Blackwell GPUs, the B200 delivers ground breaking performance with support for multi-trillion parameter models, making it ideal for training, fine-tuning, and inference at scale.
 

• Ideal for training, inference, and RAG (retrieval-augmented generation) use cases
   
• Designed for seamless integration into DGX SuperPOD clusters
   

The DGX B200 sets a new standard in AI infrastructure, helping organizations unlock the next generation of breakthroughs in generative AI, scientific computing, and large-scale AI deployments.

The DGX H100 is NVIDIA’s powerful AI supercomputer built to accelerate training and inference of the most advanced AI models. Featuring 8× NVIDIA H100 Tensor Core GPUs, it delivers breakthrough performance for demanding workloads such as foundation models, generative AI, HPC, and large-scale deep learning.
 

• Designed for multi-node scale-out in DGX SuperPODs
   
• Full support for NVIDIA AI Enterprise software 

The DGX H100 enables enterprises and research institutions to build and deploy state-of-the-art AI models with speed, scale, and efficiency.

The DGX H200 is the next-generation upgrade to the H100 system, featuring 8× NVIDIA H200 GPUs with significantly increased memory capacity and bandwidth. It is specifically optimized for memory-intensive workloads such as large language models (LLMs), generative AI, and recommendation systems.

• Built for large-scale generative AI, retrieval-augmented generation (RAG), and LLM fine-tuning
   
• Compatible with DGX SuperPOD infrastructure
   
• Supports the latest NVIDIA AI software stack, including NeMo and TensorRT-LLM
   

The DGX H200 offers a substantial performance boost for data- and memory-intensive AI workloads, making it a powerful option for enterprises scaling up to the next level of AI innovation.

 The NVIDIA DGX B300 is NVIDIA’s latest flagship AI supercomputer, purpose-built to power the next generation of trillion-parameter generative AI models. Equipped with 8× Blackwell B200 GPUs and enhanced with Grace Blackwell Superchips, the DGX B300 is engineered for unmatched training, fine-tuning, and inference performance at extreme scale.

• Support for FP8, FP6, and new transformer engine enhancements for faster generative AI processing
   
• Integrated with NVIDIA AI Enterprise, NeMo, TensorRT-LLM, and CUDA software stacks
   
• Fully compatible with DGX SuperPOD and NVIDIA Spectrum-X networking for multi-node scaling
   

The DGX B300 delivers unprecedented AI performance, enabling organizations to train massive frontier models, power real-time inference, and drive AI breakthroughs across science, industry, and enterprise.