TensorFloat-32

TensorFloat-32 (TF32) is a numeric floating point format designed for Tensor Core running on certain Nvidia GPUs. It combines the 8-bit exponent size of IEEE single precision with the 10-bit mantissa size of half precision.

Format

The binary format is:

• 1 sign bit
• 8 exponent bits
• 10 significand bits (also called mantissa, or precision bits)

The 19-significant-bit format fits within a double word (32 bits), and while it lacks precision compared with a normal 32-bit IEEE 754 floating-point number, provides much faster computation, up to 8 times on a A100 (compared to a V100 using FP32). ^[1]

Stored in the same space as FP32, it is not really a distinct storage format, but a specification for reduced-precision FP32 multiply–accumulate operations. FP32 inputs are rounded to TF32, multiplied to produce a 21-bit product (including the implicit msbit, this is an 11×11→22-bit multiply), and summed into a standard FP32 accumulator. ^[2]

See also

• IEEE 754

References

1. "NVIDIA TF32". 8 February 2023. Retrieved 23 May 2024.
2. Stosic, Dusan; Micikevicius, Paulius (27 January 2021). "Accelerating AI Training with NVIDIA TF32 Tensor Cores" . Retrieved 10 December 2025.