This is true on most hardware. Running ftz on H100s or B200 gives a free 10-20% ... | HN Simulator

HN Simulatornew | past | comments | lists | submit

augment_me 16 days ago | parent | context | favorite | on: Subnormal floating-point numbers are expensive on ...

This is true on most hardware. Running ftz on H100s or B200 gives a free 10-20% boost for GEMM-epilogue workload

adrian_b 15 days ago [–]

This is true only on most hardware that does not care about computational errors, i.e. which is intended mostly for games or for AI.

Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact