Back to Glossary
GPU Cloud
Tensor Core
Tensor Cores are specialized GPU hardware units that accelerate matrix operations for AI workloads.
Definition
Tensor Cores are specialized hardware units within NVIDIA GPUs designed to accelerate matrix multiplication operations — the core computation in neural networks. Each Tensor Core can perform a 4x4x4 matrix multiply-accumulate (MMA) in one clock cycle. Modern Tensor Cores (H100, H200) support FP64, FP32, FP16, BF16, INT8, FP8, and INT4 precision. The H100 has 528 Tensor Cores delivering 3,958 TFLOPS FP8 performance. Tensor Cores are the reason GPUs are 10-100x faster than CPUs for AI workloads.
Related Keywords
tensor coregpu tensor corenvidia tensor corematrix multiplication gpu
Related Terms
NVIDIA H100 GPU
The NVIDIA H100 is a datacenter GPU built on the Hopper architecture, designed for AI training and inference at scale.
CUDA
CUDA is NVIDIA's parallel computing platform and programming model for GPU acceleration.
GPU Cloud
GPU cloud is a cloud computing service that provides access to graphics processing units (GPUs) on-demand for AI, ML, and HPC workloads.