NVIDIA - DGX B200 - Blackwell AI Infrastructure Platform

The NVIDIA DGX B200 is a unified AI platform designed for the entire develop-to-deploy pipeline, providing a foundation for enterprise AI factories. Built on the NVIDIA Blackwell architecture, it integrates eight Blackwell GPUs with fifth-generation NVLink technology to deliver 3X the training performance and 15X the inference performance of previous generations. - 8x NVIDIA Blackwell GPUs with 1,440 GB total HBM3e memory and 64 TB/s bandwidth. - Delivers up to 144 PFLOPS FP4 Tensor Core performance (sparse). - Fifth-generation NVIDIA NVLink provides 14.4 TB/s aggregate bandwidth. - Dual Intel Xeon Platinum 8570 processors with 112 total cores and 4 GHz max boost. - 2 TB system memory, configurable up to 4 TB for massive datasets. - Advanced networking with 8x 400 Gb/s InfiniBand/Ethernet ports via ConnectX-7 VPI. - Dual 400 Gb/s NVIDIA BlueField-3 DPUs for infrastructure offloading and security. - High-density 10 RU rackmount form factor optimized for data center deployment. - Integrated NVIDIA AI Enterprise and NVIDIA Mission Control software stacks. - Storage configuration includes 2x 1.9 TB NVMe M.2 OS drives and 8x 3.84 TB NVMe U.2 internal drives.