The NVIDIA V100 PCIe is the PCIe form factor variant of the Volta flagship, announced alongside the V100 SXM at GTC May 2017. It brings Tensor Core technology to standard server infrastructure, enabling organizations to deploy Volta's AI acceleration capabilities without specialized SXM baseboards.
The V100 PCIe is typically available in 16GB and 32GB HBM2 configurations. The 16GB version provides 900 GB/s memory bandwidth and was the initial release configuration. Both versions use the same GV100 die with 5,120 CUDA cores and 640 first-generation Tensor Cores.
The full-height, full-length dual-slot PCIe Gen3 x16 form factor has a 250W TDP, enabling air-cooled deployment in standard server chassis. For multi-GPU configurations, NVLink bridge supports connecting two V100 PCIe cards with 300 GB/s bidirectional bandwidth, providing GPU-to-GPU communication faster than PCIe.
The V100 PCIe was widely deployed in cloud computing environments and enterprise AI infrastructure between 2017-2020. Cloud providers including AWS (P3 instances), Azure (NC v3), and Google Cloud offered V100 PCIe-based instances that introduced many organizations to Tensor Core computing.