NVIDIA B100 Tensor Core GPU

Overview

The NVIDIA B100 Tensor Core GPU is a high-performance Blackwell-architecture accelerator featuring 192GB of HBM3e memory with 8 TB/s bandwidth. Positioned between the B200 and H200, the B100 delivers an excellent balance of performance, memory capacity, and power efficiency for mainstream AI training and inference workloads. With Blackwell's second-generation Transformer Engine and FP4 tensor core precision, it enables efficient training and deployment of large language models in standard data center environments.