Skip to content
HardwareLadder
AI GPU ladder

NVIDIAAI GPU · Superchip

GH200

NVIDIA GH200 is a Hopper + Grace data-center accelerator with 144 GB HBM3e and 4.9 TB/s of memory bandwidth.

Ladder rank
#7
AI index (H100 SXM = 100)
120.9
Series
Hopper

Specifications

Dense throughput, no sparsity
Architecture Hopper + Grace
Form factor Superchip
Process TSMC 4N
FP16 / BF16 tensor (dense) 989.5 TFLOPS
FP8 tensor (dense) 1,979 TFLOPS
FP4 tensor (dense) —
Memory 144 GB HBM3e
Memory bandwidth 4.9 TB/s
Power 450–1000 W (superchip)
Launch Aug 2023

AI index

Tensor × bandwidth

100 × √((989.5 ÷ 989.4) × (4.9 ÷ 3.35)) = 120.9

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