GB200 GPU

The NVIDIA GB200 is a data center GPU based on the Blackwell architecture, representing NVIDIA's current-generation offering for large-scale AI and HPC workloads. Built on the Blackwell architecture, the GB200 delivers 384GB of VRAM with 16,000 GB/s of memory bandwidth, positioning it as a high-capacity solution for memory-intensive applications. The GPU features a 2700W TDP and is designed for rack-scale deployments requiring liquid cooling infrastructure. Key technical specifications include 9,000 TFLOPS of FP16 performance and support for FP4 precision through its second-generation Transformer Engine. The GB200 incorporates fifth-generation NVIDIA NVLink technology enabling up to 130 TB/s of low-latency GPU-to-GPU communication within 72-GPU NVLink domains. The substantial 384GB memory capacity and high memory bandwidth make it suitable for applications requiring large model storage and high-throughput data processing. In cloud deployments, the GB200 targets large language model training and inference, massive-scale AI workloads, and high-performance computing applications that benefit from its extensive memory capacity and inter-GPU connectivity. The rack-scale configuration and liquid cooling requirements make it primarily suitable for specialized data center environments optimized for high-density compute workloads.