NVIDIAAI GPU · PCIe (low profile)
A2
NVIDIA A2 is a Ampere data-center accelerator with 16 GB GDDR6 and 0.2 TB/s of memory bandwidth.
Specifications
Dense throughput, no sparsity| Architecture | Ampere |
|---|---|
| Form factor | PCIe (low profile) |
| Process | Samsung 8 nm |
| FP16 / BF16 tensor (dense) | 18 TFLOPS |
| FP8 tensor (dense) | — |
| FP4 tensor (dense) | — |
| Memory | 16 GB GDDR6 |
| Memory bandwidth | 0.2 TB/s |
| Power | 40–60 W |
| Launch | Nov 2021 |
AI index
Tensor × bandwidth100 × √((18 ÷ 989.4) × (0.2 ÷ 3.35)) = 3.3
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.