16-bit floating-point precision format (half precision). Commonly used in AI training and inference to reduce memory usage and increase throughput compared to FP32. Provides adequate precision for most neural network operations.
Brain Float 16 - A 16-bit floating-point format developed by Google with the sam...
Single-precision 32-bit floating point, the long-standing default for general-pu...
8-bit floating-point precision format introduced in Hopper architecture. Enables...
Specialized processing units in NVIDIA GPUs designed to accelerate matrix multip...
Browse the full glossary or dive into architecture guides
