DEPLOYDatabase

Chip comparison

NVIDIA GB300 (Blackwell Ultra) 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.

FieldNVIDIA GB300 (Blackwell Ultra)NVIDIA L40S
Process nodeTSMC 4NPTSMC 4N
Transistors (B)76.3
TDP1400 W350 W
Memory288 GB HBM3e48 GB GDDR6
Memory bandwidth8,000 GB/s864 GB/s
FP16 (dense)362 TFLOPS
BF16 (dense)362 TFLOPS
FP8 (dense)5,000 TFLOPS733 TFLOPS
INT8 (dense)733 TOPS
Form factorSuperchip (Blackwell Ultra + Grace)PCIe
Announced2025-03-182023-08-08
Released2025-11-01

Robots running NVIDIA GB300 (Blackwell Ultra)

No robots publicly running it yet.

Robots running NVIDIA L40S

No robots publicly running it yet.

Data centers with NVIDIA GB300 (Blackwell Ultra)

Data centers with NVIDIA L40S

No data centers publicly running it yet.

Common questions

NVIDIA GB300 (Blackwell Ultra) vs NVIDIA L40S: which is faster for training?

NVIDIA GB300 (Blackwell Ultra) has 6.82x the dense FP8 throughput of the other (NVIDIA GB300 (Blackwell Ultra): 5,000 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 GB300 (Blackwell Ultra) vs NVIDIA L40S: which has more memory?

NVIDIA GB300 (Blackwell Ultra) carries more HBM (NVIDIA GB300 (Blackwell Ultra): 288 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 GB300 (Blackwell Ultra) vs NVIDIA L40S: which is more power-efficient?

Dense FP8 TFLOPS per watt: NVIDIA GB300 (Blackwell Ultra) 3.57 (5,000 ÷ 1400 W); NVIDIA L40S 2.09 (733 ÷ 350 W). NVIDIA GB300 (Blackwell Ultra) wins at the die level; system-level efficiency also depends on cooling and interconnect.

NVIDIA GB300 (Blackwell Ultra) vs NVIDIA L40S: which is more widely used?

NVIDIA GB300 (Blackwell Ultra): 0 robots and 1 data center 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 GB300 (Blackwell Ultra) full page · NVIDIA L40S full page.