Not quite, but close. “tf32” is 18 bits, but with the same 10 bits of exponent that fp32 has. It’s the range of fp32 with the precision of fp16. It’s a shame to see such unoriginality in new number representations. I’d much rather see Posit hardware acceleration: https://web.stanford.edu/class/ee380/Abstracts/170201-slides...

TF32 is 19 bits, not 18 bits. There's an additional bit for sign.

https://blogs.nvidia.com/blog/2020/05/14/tensorfloat-32-prec...

No, tf32 has the same size exponent field as f32 but the mantissa size of f16, 10 bits.