Skip to content
HardwareLadder
AI GPU ladder

NVIDIAAI GPU · Superchip (per GPU)

GB300

NVIDIA GB300 is a Blackwell Ultra data-center accelerator with 288 GB HBM3e and 8 TB/s of memory bandwidth.

Ladder rank
#1
AI index (H100 SXM = 100)
245.6
Series
Blackwell

Specifications

Dense throughput, no sparsity
Architecture Blackwell Ultra
Form factor Superchip (per GPU)
Process TSMC 4NP
FP16 / BF16 tensor (dense) 2,500 TFLOPS
FP8 tensor (dense) 5,000 TFLOPS
FP4 tensor (dense) 15,000 TFLOPS
Memory 288 GB HBM3e
Memory bandwidth 8 TB/s
Power 1400 W
Launch Aug 2025

AI index

Tensor × bandwidth

100 × √((2,500 ÷ 989.4) × (8 ÷ 3.35)) = 245.6

The AI index is the geometric mean of dense 16-bit tensor throughput and memory bandwidth, each relative to the H100 SXM. It is a spec-sheet comparison, not a measured benchmark.

Sources

2 references