IT Creations Partners

NVIDIA A40 GPU - System Overview

Performance

Nvidia A40 rear view

NVIDIA’s Ampere architecture CUDA cores offer double the processing speed for single-precision floating point (FP32) workloads. This GPU is also designed to take full advantage of PCIe 4.0 offering twice the throughput compared to PCIe 3.0. With better FP32 performance, the Nvidia A40 GPU can provide dramatically improved graphics for simulations and 3D computer-aided-design (CAD) and computer-aided-engineering (CAE) applications. Second-generation ray-tracing cores (RT cores) provide 2X the throughput enabling multi-tasking of key functionality with ray tracing, shading and denoising capabilities running in concert and providing complex renderings and simulations in a fraction of the time compared to the previous generation. AI and data science model training has also been enhanced with NVIDIA’s 3rd generation Tensor cores, which utilize accelerated FP32 capabilities by a factor of 5X. Available on previous generations NVIDIA’s Deep-Learning Super Sampling for enhanced rendering capabilities is also greatly enhanced. Nvidia’s virtual GPU (vGPU) software enables larger virtual workstations instances for remote users and high-speed remote design, AI and compute workloads.

Memory

The A40 GPU supports 48GB of GDDR6 memory and can be augmented with a second GPU using NVLink technology, which enables two GPUs to share resources and function as a single asset with a combined memory of 96GB. More memory enables this GPU to rapidly process massive data sets common when used for data analysis and simulation.

Nvidia A40 NVLink

Cooling and Power

Maximum power consumption on this card is logged at 300W and uses an 8-pin power connector. It is passively cooled depending on the server chassis in which it is installed to provide the necessary thermal control for operational performance. While the Nvidia A40 GPU is configured to support virtualization by default with the 3x DisplayPort 1.4 sockets disabled, they can be enabled via management software tools.

Summary

Offering top performance for simulation, ray-traced rendering, virtual production, the Nvidia A40 GPU has the unique distinction of being the most powerful data center GPU for visual computing. Offering 48GB of GDDR6 memory, plus next generation RT cores, CUDA cores, and Tensor cores, this GPU can deliver ray tracing, shading and de-noising capabilities running simultaneously for a quicker path to solution. Augmentation by NVIDIA’s NVLink bridge enables two GPUs to function in concert with shared memory and resources enabling up to 96GB of performance with reduced latency. It is truly and exceptional GPU bringing next-generation NVIDIA RTX technology to the data center for visualization and AI applications, including training and inference.

Check Availability

NVIDIA A40 GPU - Specifications

Memory
    GPU Memory: 48 GB GDDR6 with ECC
  • Memory Bus: 384-bit
  • Memory Bandwidth: 696 GB/s
GPU Engine
  • NVIDIA CUDA Cores: 9216
  • Boost Clock: 1740 MHz
  • Base Clock: 1305 MHz
  • Memory Clock: 1812 MHz/14.5 Gbps effective
Performance
  • FP16 (half) performance: 37.42 TFLOPS
  • Transistor Count: 28,300 million
NVIDIA NVLink
  • 112.5 GB/s - bidirectional PCIe Gen4: 64 GB/s
Display Ports
  • 3x DisplayPort 1.4
  • Supports NVIDIA Mosaic and Quadro Sync
System Interface
  • PCI Express 4.0 x16
Power Consumption

    Maximum Thermal Design Power (TDP)

  • 300W
  • Suggested PSU

  • 700W
  • Power Connectors

  • 8-pin EPS
Form Factor
  • 4.4" x 10.5" dual slot
Processors
  • Fabrication Process: 8nm
  • TMUs: 336
  • ROPs: 112
  • SM Count: 84
  • Tensor Cores: 336
  • RT Cores: 84
Engine
    vGPU Software Support

  • NVIDIA vPC/vApps
  • NVIDIA RTX vWS
  • NVIDIA Virtual Compute Server (vCS)

Check Availability

Get a quote for NVIDIA A40 GPU

If you know what you want but can't find the exact configuration you're looking for, have one of our knowledgeable sales staff contact you. Give us a list of the components you would like to incorporate into the system, and the quantities, if more than one. We will get back to you immediately with an official quote.

[email protected]