The NVIDIA H100 NVL PCIe Tensor Core GPU is the PCIe form factor variant of the H100 NVL, featuring two Hopper-architecture H100 GPUs with 94GB HBM3 each (188GB total) in a standard PCIe dual-slot configuration. Designed for broad server compatibility without requiring SXM infrastructure, it delivers the same dual-GPU AI performance for LLM inference and training. The PCIe form factor enables deployment in standard enterprise servers while providing exceptional memory capacity for large model serving.
Complete technical specifications for the H100 NVL.