NVIDIA A40
NVIDIA Graphics Card
The NVIDIA A40 is a data centre GPU for professional visualisation and compute, built on the Ampere architecture with 10,752 CUDA cores and 48GB of GDDR6 ECC memory delivering 696 GB/s of bandwidth. It combines graphics and compute capability with full NVIDIA vGPU software support, making it a mainstay for virtual desktop infrastructure, cloud rendering and remote 3D design. The passively cooled dual-slot design suits high-density rack server deployment, and an NVLink bridge can combine two cards into 96GB.
Specifications
| Architecture | NVIDIA Ampere architecture (GA102) 3rd-generation Tensor Cores 2nd-generation RT Cores |
|---|---|
| Compute Cores | 10,752 NVIDIA CUDA Cores 336 Tensor Cores, 84 RT Cores |
| GPU Memory | 48GB GDDR6 with ECC 384-bit interface, 696 GB/s bandwidth |
| System Interface | PCIe 4.0 x16 |
| Interconnect | NVLink bridge support, two cards can be combined into 96GB |
| Virtualisation | Full NVIDIA vGPU software support |
| Power | 300W max power consumption |
| Cooling & Form Factor | Passive cooling (chassis airflow) Dual-slot full-height full-length, for rack servers |
| Use Cases | Virtual desktop infrastructure, cloud rendering, remote 3D design, professional visualisation, AI inference |