Chip comparison
NVIDIA B200 (Blackwell) vs NVIDIA L40S
Side-by-side specs
Straight from each vendor's datasheet. Winner in bold: higher throughput, memory, or transistor count; lower power draw or price. Sparse (2:4) throughput roughly doubles the FP8/INT8 numbers shown.
| Field | NVIDIA B200 (Blackwell) | NVIDIA L40S |
|---|---|---|
| Process node | TSMC 4NP | TSMC 4N |
| Transistors (B) | 208 | 76.3 |
| TDP | 1000 W | 350 W |
| Memory | 192 GB HBM3e | 48 GB GDDR6 |
| Memory bandwidth | 8,000 GB/s | 864 GB/s |
| FP16 (dense) | 2,250 TFLOPS | 362 TFLOPS |
| BF16 (dense) | 2,250 TFLOPS | 362 TFLOPS |
| FP8 (dense) | 4,500 TFLOPS | 733 TFLOPS |
| INT8 (dense) | 4,500 TOPS | 733 TOPS |
| Form factor | SXM6 | PCIe |
| Announced | 2024-03-18 | 2023-08-08 |
| Released | 2024-11-01 | — |
Robots running NVIDIA B200 (Blackwell)
No robots publicly running it yet.
Robots running NVIDIA L40S
No robots publicly running it yet.
Data centers with NVIDIA B200 (Blackwell)
Data centers with NVIDIA L40S
No data centers publicly running it yet.
Common questions
NVIDIA B200 (Blackwell) vs NVIDIA L40S: which is faster for training?
NVIDIA B200 (Blackwell) has 6.14x the dense FP8 throughput of the other (NVIDIA B200 (Blackwell): 4,500 TFLOPS; NVIDIA L40S: 733 TFLOPS). Real-world training speed also depends on memory bandwidth, interconnect, and software stack: see the spec diff for those inputs.
NVIDIA B200 (Blackwell) vs NVIDIA L40S: which has more memory?
NVIDIA B200 (Blackwell) carries more HBM (NVIDIA B200 (Blackwell): 192 GB HBM3e; NVIDIA L40S: 48 GB GDDR6). Memory capacity sets the largest model that fits on one chip without partitioning; memory bandwidth sets inference throughput.
NVIDIA B200 (Blackwell) vs NVIDIA L40S: which is more power-efficient?
Dense FP8 TFLOPS per watt: NVIDIA B200 (Blackwell) 4.50 (4,500 ÷ 1000 W); NVIDIA L40S 2.09 (733 ÷ 350 W). NVIDIA B200 (Blackwell) wins at the die level; system-level efficiency also depends on cooling and interconnect.
NVIDIA B200 (Blackwell) vs NVIDIA L40S: which is more widely used?
NVIDIA B200 (Blackwell): 0 robots and 3 data centers publicly running it. NVIDIA L40S: 0 robots and 0 data centers. Empty here means no operator has said publicly, not zero adoption in the market.
See also: every chip comparison · NVIDIA B200 (Blackwell) full page · NVIDIA L40S full page.