NVIDIA V100 Tensor Core GPU
The NVIDIA V100 Tensor Core GPU is a high-performance data center GPU built on the NVIDIA Volta architecture. Designed for AI training, deep learning, HPC (High-Performance Computing), and scientific simulations, it delivers exceptional compute performance with Tensor Cores, high-bandwidth HBM2 memory, and fast NVLink or PCIe connectivity.
Specification | V100 NVLink | V100 PCIe | V100S PCIe |
|---|---|---|---|
GPU Architecture | Volta | Volta | Volta |
Memory Capacity | 16GB / 32GB HBM2 | 16GB / 32GB HBM2 | 32GB HBM2 |
Memory Bandwidth | 900 GB/s | 900 GB/s | 1,134 GB/s |
Single Precision (FP32) | 15.7 TFLOPS | 14 TFLOPS | 16.4 TFLOPS |
Double Precision (FP64) | 7.8 TFLOPS | 7 TFLOPS | 8.2 TFLOPS |
Tensor Performance (AI) | 125 TFLOPS | 112 TFLOPS | 130 TFLOPS |
Interconnect | NVLink | PCIe | PCIe |
Interconnect Bandwidth | 300 GB/s | 32 GB/s | 32 GB/s |
Memory Type | HBM2 | HBM2 | HBM2 |
Maximum Power Consumption | 300 W | 250 W | 250 W |