NVIDIA L40S product page specification table
spec-sheet
Fields
id
sources/acc2-nv-l40s-product-pagePublisher
NVIDIA
Kind
spec-sheet
Accessed
2026-10-03
notes
Retrieved via the markdown mirror at https://www.nvidia.com/en-us/data-center/l40s.md (the .html path yields no extractable text). Prints dense|sparse pairs separated by 'I' with '*' marking the sparse column: 'TF32 Tensor Core TFLOPS 183 I 366*', 'BFLOAT16 362.05 I 733*', 'FP16 362.05 I 733*', 'FP8 733 I 1,466*', 'Peak INT8 733 I 1,466*', 'Peak INT4 733 I 1,466*'. Footnote on the page: '1 Peak rates are based on GPU boost clock.' The headline 'Tensor Performance 1,466 TFLOPS' is therefore the SPARSE FP8 figure, while 733 is dense.
Notes
Retrieved via the markdown mirror at https://www.nvidia.com/en-us/data-center/l40s.md (the .html path yields no extractable text). Prints dense|sparse pairs separated by 'I' with '*' marking the sparse column: 'TF32 Tensor Core TFLOPS 183 I 366*', 'BFLOAT16 362.05 I 733*', 'FP16 362.05 I 733*', 'FP8 733 I 1,466*', 'Peak INT8 733 I 1,466*', 'Peak INT4 733 I 1,466*'. Footnote on the page: '1 Peak rates are based on GPU boost clock.' The headline 'Tensor Performance 1,466 TFLOPS' is therefore the SPARSE FP8 figure, while 733 is dense.
Cross-references
This record references no other records.