Condition: Refurbished, fully tested and working perfectly. Restored to like-new condition. Original box and accessories not included.

Warranty: 1 year return-to-base warranty included. 14-day returns accepted.

The NVIDIA A40 brings the power of the NVIDIA Ampere architecture to the data centre for the most demanding visual computing workloads. Combining 10,752 CUDA cores with second-generation RT Cores and third-generation Tensor Cores, it accelerates everything from photorealistic rendering and virtual production to deep learning training and inference. Its 48GB of ultra-fast GDDR6 memory with ECC handles massive datasets, complex scenes and memory-hungry simulation with ease.

Built for round-the-clock server deployment, the A40 pairs a passively cooled, power-efficient dual-slot design with PCI Express 4.0 and NVIDIA NVLink support, letting two A40 GPUs connect for a combined 96GB of graphics memory. Secure and measured boot with hardware root of trust and NEBS Level 3 readiness make it a dependable foundation for enterprise and edge installations alike.

From powerful virtual workstations running NVIDIA RTX Virtual Workstation software to dedicated render nodes, batch rendering farms, data science and AI workloads, the A40 delivers state-of-the-art performance and true multi-workload flexibility for the modern data centre.

Manufacturer

NVIDIA

Part Number

900-2G133-0000-000

GPU Architecture

NVIDIA Ampere

CUDA Cores

10752

RT Cores

84 (2nd Generation)

Tensor Cores

336 (3rd Generation)

GPU Memory

48GB GDDR6 with ECC

Memory Bus

384-bit

Memory Bandwidth

696 GB/s

Bus Interface

PCI Express 4.0 x16

Interconnect

NVIDIA NVLink 112.5 GB/s bidirectional (two A40s combine to 96GB)

GPU Clock

1305 MHz base / 1740 MHz boost

Display Outputs

3x DisplayPort 1.4 (disabled by default, enabled via management software)

Video Engines

1x NVENC, 2x NVDEC (AV1 decode support)

Graphics APIs

DirectX 12.07, Shader Model 5.17, OpenGL 4.68, Vulkan 1.18

Compute APIs

CUDA, DirectCompute, OpenCL, OpenACC

Max Power Consumption

300W

Power Connector

1x CPU 8-pin

Thermal Solution

Passive

Form Factor

Full-height, full-length, dual-slot, 4.4in (H) x 10.5in (L)

NEBS Ready

Level 3

vGPU Software Support

NVIDIA vPC/vApps, RTX Virtual Workstation, Virtual Compute Server

Board Weight

990g