The NVIDIA V100 Tensor Core GPU was the first GPU to introduce Tensor Cores, revolutionizing deep learning training and HPC. Featuring 32GB of HBM2 memory with 900 GB/s bandwidth and 640 Tensor Cores on the Volta architecture, the V100 delivered a 12x speedup over its predecessor for deep learning training. It remains widely deployed in legacy clusters and is available through major cloud providers for continued AI and scientific computing workloads.
Complete technical specifications for the V100.